Review

Previously: The Fundamentals, The Gamblers, The Business

We have now arrived at the topics most central to this book, aka ‘The Future.’

The Manifest conference was also one of the last reporting trips that I made for this book. And it confirmed for me that the River is real—not just some literary device I invented. (6706)

Yep. The River is real.

I consider myself, among many things, a straight up rationalist.

I do not consider myself an EA, and never have.

This completes the four quadrants of the two-by-two of [does Nate knows it well, does Zvi knows it well]. The first two, where Nate was in his element, went very well. The third clearly was less exacting, as one would expect, but pretty good.

Now I have the information advantage, even more than I did for aspects of sports gambling.

1. We’ve seen Nate attempt to tackle areas in both our wheelhouses.

2. We’ve seen Nate attempt to tackle areas in his wheelhouse that I’ve only explored.

3. We’ve seen Nate attempt to tackle areas he was exploring, that I’ve only explored.

4. Now he’s exploring new sections of my wheelhouse.

Let’s see how he explains it all.

Effective altruism, and the adjacent but more loosely defined intellectual movement called “rationalism,” are important parts of the River on their own terms. In some ways, in fact, they are the most important parts.

Much of the River is concerned with what philosophers call “small-world problems,” meaning tractable puzzles with relatively well-defined parameters: how to maximize expected value in a poker tournament, or how to invest in a portfolio of startups that brings you upside with little risk of ruin.

But in this final portion of the book, we’re visiting the part of the River where people instead think about open-ended, so-called grand-world problems: everything from where best to spend your charitable contributions to the future of humanity itself. (6228)

A solid opening.

I would still nitpick on the word ‘instead,’ and would have suggested ‘also.’

The Rationalists saw themselves as people who applied scientific thought to almost any topic. This often involved “Bayesian reasoning,” a way of using statistics and probability to inform beliefs. (6393)

The ‘True Rationalist’ in particular both hones their skills and upgrades their lives by applying the same principles to everything. No matter is too trivial for the nerd snipe and the Bayesian reasoning. In particular, local questions, that help improve your life and your ability to think and impact the world, matter. You are not forced to look only at bigger pictures.

Indeed, Nate Silver was correctly informed that he counts as a rationalist. You don’t have to join or even know we exist in order to be a member.

In fact, even if I had never applied to join Team Rationalist, Alexander—whose soft features, dry wit, and male pattern baldness reminded me uncannily of my dad’s (West Coast Jewish) side of the family—had already drafted me into it. “You are clearly doing a lot of good work spreading rationality to the masses. Is it useful to think of us as a movement that doesn’t include you?” he asked me. (6401)

The origin story of many rationalists is exactly that they needed it on a personal level. The traditional mode of acting on intuition and instinct and kind of vibing was clearly not working. The world did not make sense. They had to figure things out the hard way, from first principles. The good news is, once you do that, you actually understand what you are doing, and are choosing on purpose how to do it. You end up far stronger at the end of the path.

And indeed, there are some rationalists, and some EAs, that are perfectly content to use the toolboxes on that level. We strive to help them get more ambitious when we are ready for that, but you don’t get cast out if you decide to not go big, and stay home.

But yes, what matters is that often people who think this way end up going big.

The reason some Riverians have become obsessed with grand-world problems is because the Village and the rest of the world screw them up all the time, too, in ways that often reflect political partisanship, an endless array of cognitive biases, innumeracy, hypocrisy, and profound intellectual myopia.

To take one glaring example that Flynn reminded me of: the U.S. Congress has authorized relatively little—only around $2 billion in spending as part of a 2022–23 budget deal—to prevent future pandemics, even though COVID-19 killed more than 1 million Americans and cost the U.S. economy an estimated $14 trillion.

Reducing the chance of a future such pandemic in the United States by even 1 percent would be +EV even at a cost of $140 billion—and yet Congress is barely spending one one-hundredth of that.

You cannot count on your civilization to act in a sane fashion. There is no Reasonable Authority Figure. We are Beyond the Reach of God. As Scott Alexander wrote, someone has to, and no one else will. Or as Hillel the Elder said two millennia earlier: If I am not for myself, who will be for me? If I am only for myself, what am I? If not now, when?

And now, on top of that, we face substantial existential risks, most of all from AI.

As crazy as it sounds, yes, it is up to us. It is up to you.

Our society is very bad at cost-benefit analysis. As in we often refuse to do one.

There are narrow places where we are quite good at it. We do indeed do cost-benefit analysis sometimes, at all, including on things that matter, and that is really great. We also often rely on markets to get us to do it, which is insanely great.

Alas, we also often act like complete morons, because we refuse.

Transit officials faced a difficult choice: They could shut down the F, blocking a vital link between New York’s two densest boroughs right as commuters were beginning to get off work—or they could potentially run over poor Dakota. They elected to close the F for more than an hour until Dakota was found. (6261)

I am sorry, what? A difficult choice? This is a trivially easy choice.

You only need to answer one question about the above passage. Is Dakota a human?

If the answer is yes, than as Nate says, we all agree, you stop the trains.

We put the value of a human life saved (VSL) around $10 million, and in situations like this we are willing to do a bit higher.

When you tease out this and lots of other data points—say, by looking at how much people are willing to pay for additional safety features when they buy a new car—the average American implicitly values their life at about $10 million. That’s where the VSL comes from. (6358)

Dakota, however, was a dog.

Claude initially estimated the total cost of the train delay at $1.58 million. It is actually substantially higher, because that estimate takes lost time at work as being equal to the hourly wage of the employee. Whereas if an employee’s marginal value per hour was only their wage, why do they have a job? And when someone is unexpectedly late, with little warning, that can lead to big problems, including things like ‘the doctor can’t perform surgery on you today.’

I’m confident the cost here is north of $2 million, and there is no guarantee that this results in the cat not being run over.

If you suggested a $1 million medical treatment to save that cat’s life, and that the government should pay for that, that would be obviously patently absurd. I would absolutely laugh in your face.

If you called up Dakota’s owner and said, ‘all right, we can close down the F train for you, but that will be $1 million dollars’ we all know what the answer would be, once they were done yelling at you. We have willingness to pay studies. When forced to pay, less than 10% of pet owners are willing to pay $10,000 or more for life-saving medical treatments.

So let’s not pretend this is the MTA faced with a hard choice. This is the MTA faced with an absurdly easy choice. And they chose wrong.

Thus, the need for something like rationalism, and something like Effective Altruism.

As in, I can’t help but notice that you do things without checking to see if they will be effective, or if there is a way to do them better. Perhaps you should think about that?

What is effective altruism, exactly? In one sense, effective altruism is just a brand name, created by MacAskill and another Oxford philosopher, Toby Ord, in 2011. (6370)

The more official answer—as stated by MacAskill in an essay entitled “The Definition of Effective Altruism”—is that EA is a “movement [that tries] to figure out, of all the different uses of our resources, which uses will do the most good, impartially considered.” (6374)

That’s the 80/20 for a lot of this. You try, at all, to figure out what will actually result in what outcomes at what costs with what benefits. Then you choose what seems best. The rest is not stamp collecting, the rest is important, but you’ll already be way ahead.

Eliezer Yudkowsky back in 2006 listed the twelve virtues of rationality: Curiosity, relinquishment, lightness, evenness, argument, empiricism, simplicity, humility, perfectionism, precision, scholarship, and the void.

On a more practical level, it means things like this:

Even public figures who are critical of the movements tend to get a fair hearing at blogs like LessWrong and at the Effective Altruism Forum—which is pretty much the opposite of what it’s usually like to argue about public affairs online. (6421)

This plays out in instrumental and epistemic rationality.

First, there’s instrumental rationality. Basically this means: Do you adopt means suitable to your ends? There is a man who has eaten more than thirty thousand Big Macs. Now, this might not be a reasonable and prudent thing for him to do. But if this man’s life goal is to eat as many Big Macs as possible, you could say he’s instrumentally rational because he’s done a bang-up job of this. (6725)

The second type is epistemic rationality. This means: Do you see the world for what it is? Do your beliefs line up with reality? (6730)

Good summary. You need both.

You can also give the rationalists credit for argumentative consistency: they tend to be scrupulously honest. (6816)

Rationalists have, from the outside perspective, utterly absurd high standards on scrupulosity and honesty. I believe this to be a very good thing.

But the kinship that EAs and rationalists feel for each other conceals that there are a lot of internal disagreements and even contradictions within the movements—in particular, there are two major streams of EA/rationalism that don’t see eye to eye.

The first is associated with the Australian philosopher Peter Singer and a cluster of topics including animal welfare, global poverty reduction, effective giving, and not living beyond your means—but also the ethical precept known as utilitarianism.

The second is associated with Yudkowsky and the George Mason University economist Robin Hanson and a whole different cluster of topics: futurism, artificial intelligence, prediction markets, and being willing to argue about just about anything on the internet, including subjects that others often find taboo. (6428)

Not living beyond your means is (highly non-uniquely) a rationalism thing. Not retaining means with which to live better is the EA thing.

Then later on the Effective Altruists realized the rationalists were right about the promise and dangers of AI and existential risks from AI, so that became the EA cornerstone as well.

Furthermore, I think it’s altruistic when people like Singer express unpopular viewpoints that they honestly believe will lead to social betterment and selfish to suppress these ideas because of fear of social approbation. (6476)

I agree in principle, although I worry about the frame of ‘altruistic’ being misleading. The important thing is that, if more people said what they actually believe whether or not it is popular, and whether or not it is convenient, and whether or not I agree with it, that would make the world a better place.

There is then of course Singer’s famous drowning child metaphor, that if you’d ruin your expensive coat to save a drowning child in front of you, that means you are a bad person because you should have never bought that expensive coat and instead could have donated that money to global poverty relief.

Okay then, so why don’t I find the drowning child parable persuasive? Well, partly because it’s meant to play a trick on you—as Singer freely admits. (6479)

Indeed. It’s a magician’s trick. Singer wants you to ignore, among other things, all the reasons that we have agreed to make that drowning child in front of you your responsibility in particular, all the reasons we need some amount of locality in our preferences, and all the reasons it is not okay to redistribute all the wealth whenever you feel like it. That civilization exists for a reason, and you need to maintain it, along with all the ways we are able to make expensive coats and also save lives at all.

Then there’s the issue of utilitarianism.

There are some settings where I think utilitarianism is an appropriate framework—particularly in medium-scale problems such as in establishing government policy where impartiality (not playing favorites) is important.

For instance, when a subcommittee of the CDC met in November 2020 to develop recommendations for who would be first in line for COVID vaccines, they rejected going with a utilitarian calculus of maximizing benefits and minimizing harms to instead also consider objectives like “promo[ting] justice” and “mitigat[ing] health inequalities.” (6505)

I think utilitarianism is analogous to an underfit model. Instead of being too deferential to commonsense morality, it doesn’t meet people in the middle enough, accepting that maybe various laws and customs evolved for good reasons. (6547)

I should note, however, that utilitarianism, especially in its strictest forms, is actually relatively unpopular among philosophers. (6572)

Most people need more utilitarianism on the margin, to go with their additional use of cost-benefit analysis. When I say ‘I am not a utilitarian’ I mean not taking it to its bullet-biting conclusions, and not seeing it as the proper operating system for the human brain in practice, and not believing that you can fully total up the points of various events to choose this week’s winner in any cosmic or moral sense.

I’m arguing with the Actual Utilitarians, not with the person on the street. But the other thing about the person on the street is they also need more good virtue ethics and more good deontology, and are mostly ill-prepared to go Full Utilitarian.

A few of us have to worry about infinite cases and weird out of sample philosophical questions, in those times we are dealing with those as actual possibilities, such as in potential AI futures. For most people, that never happens. Even for those where it does happen? Most of the time, for most questions, not so much.

And that is fine. The human brain has limited compute and should not be using heuristics all the time based on whether they handle rare edge cases – so long as you recognize when you do face those edge cases.

“The thought that, well, this theory isn’t good if it can’t handle infinite cases, I think that’s like a huge mistake,” said Buchak. She thinks moral theories should instead be tested on practical, day-to-day decision-making. “Nearly every decision you face involves risk,” she said. “I’m like [more] concerned with just like, you know, should I bring my umbrella today?”

If a moral theory can’t handle everyday cases like these—if it strays too far from common sense—then we probably shouldn’t trust it, whether or not it provides an elegant answer to the Repugnant Conclusion. (6600)

I agree. If your system can’t handle ordinary cases, then you should be highly suspicious. And if it can’t handle ordinary cases without inordinate amounts of compute (as in human brain cycles, in this context) then that’s a problem too. Note that this is more of an issue in practice than in theory. If it works in practice for humans in ordinary situations, then it counts. If it doesn’t, then it doesn’t.

The reverse is not true. If a system does handle the ordinary cases well, then that is a fine thing to use to handle ordinary cases. But it could still be a huge disaster in unusual cases. And if most of the value of a system lies in how it handles future non-ordinary cases, then establishing one that only works in ordinary cases can be disastrous.

Indeed, most systems for dealing well with ordinary situations are (wisely) overfitting on the data, because we constantly face similar ordinary situations. That’s fine, except when you run into those unusual situations. Then you need to understand that your instinctive rules might be leading you very astray.

Also, I’ve said it before, and a lot of people told me I’m wrong but their arguments were all invalid so I’m going to say it again: The Repugnant Conclusion is a silly misunderstanding. It’s another magician’s trick.

The standard proof of the conclusion is invalid, because it involves manifesting resources out of thin air. The most correct response to ‘what if potatoes plus muzak maximizes your total universe utility score?’ is ‘it quite obviously does not do that, a human life contains a lot of resource costs and downsides and many benefits and potential joys, and it is quite obviously more efficient to have less people that are happier than that. Your so-called proof otherwise must be wrong on that basis alone. Also it is trivially invalid because you can’t go from world N to world N-prime in order to then loop back to world (N+1), because that move creates new people living at net zero utility without taking any resources away from anyone else. A duck is chasing you asking how you did that.’

As Craig Ferguson often said, I look forward to your letters. You can talk amongst yourselves if you’d like. But if it’s the same counterarguments and confusions, I’m precommiting here to ignoring them. I’ll only answer if I see something new.

But who in the hell am I (or Lara Buchak or Peter Singer) to tell you what you should do in decisions you’ll face just once? “It might be that you should behave differently when choosing a spouse or choosing a job or doing these kinds of things that you’re only going to do once, hopefully,” Buchak told me. (6614)

No.

You should still do the calculation and make the best possible decision as best you can.

Indeed, if it’s a big decision like a spouse or a job, those are the decisions that matter. Those are the ones where it’s worth making sure you get it right. It is very much not the time to throw the rules out the window, especially before you know the rules well enough to break them.

There are of course two big differences.

The most important one is risk aversion. You don’t get to use responsible bankroll management when choosing a job or spouse. Life doesn’t let you not take big risks, not without paying a very high price. But yes, some amount of risk aversion is appropriate in those big decisions. It’s not a pure +EV in dollars or abstractions calculation. Which is fine. So factor that, along with everything else, in.

The other big difference is inability to learn and iterate. With most decisions, a lot of the value of a good decision process is to learn from both success and mistakes, to grow wise and to make better decisions in the future. Whereas in a one-time high stakes decision like choosing a spouse, knowing how to do it better next time will be of relatively little help.

I think there is some rational basis for partiality because we have more uncertainty about things that are removed from us in time and space. (6623)

This is indeed a classic modernist failure mode, where you act like you understand what is happening elsewhere far more than you actually do. You have to discount distant actions for this risk. But that is not the only reason you need spatial and knowledge-based partiality.

Civilization would not run, people would not survive or reproduce or even produce, the social contract would collapse, if you did not favor and exchange with and cooperate uniquely with those around you beyond what you do with strangers halfway around the world. All that, and real competition, is necessary. Those strangers are not only people too but also certified Popes, so please treat them right, but that does not mean full equal standing. The alternative is not game theory compatible, it is not fit, it does not long survive.

There is little virtue in being too virtuous to sustain that virtue, and indeed if that is a thing you are thinking of as virtuous than you have chosen your virtues poorly.

And even if I think there’s something honorable about acting morally in a mostly selfish world, I also wonder about the long-term evolutionary fitness of some group of people who wouldn’t defend their own self-interest, or that of their family, their nation, their species, or even their planet, without at least a little more vigor than they would that of a stranger. I want the world to be less partial than it is, but I want it to be at least partially partial. (6653)

Yep.

This is another important observation:

Overall, the politics of EA can be slippery, stuck in the uncanny valley between being abstractly principled and ruthlessly pragmatic, sometimes betraying a sense that you can make it up as you go along. (6828)

One of the core tensions in EA is, to put it bluntly, honesty versus lying.

There is the faction that says you want to ‘do the most good,’ and you shouldn’t let the truth get in the way of that. This starts with Peter Singer, who is clear that he believes the virtuous man should be willing to lie their ass off. Thus ‘honesty is not part of my utility function,’ and SBF justifying what he did. Alternatively, perhaps you tell the truth to the ingroup, other EAs and select allies, but politics is politics. Play to win.

The other faction aligns with the rationalists, who say that if you lose your epistemics and your honesty, then all is lost. That telling the truth and playing it all fully straight is the only road to wisdom and people will recognize this and it will succeed over time. That this is especially true given that the most important issue is AI. If you don’t have excellent epistemics, and if you don’t get others to have good epistemics, acting wisely around AI is hopeless, because it is all so complex and hard to understand, and to figure out what is actually helpful versus what would backfire.

And of course, many people are somewhere in the middle.

You already know which side I am on.

Nate Silver talks to Roon, Paul Graham and Sam Altman about Altman’s history at OpenAI.

Those are excellent sources. They are also highly biased ones. They tell the official Altman version of the tale. Paul Graham has been a long time extreme Altman fan. They clearly work together to tell their narrative of events and ensure Altman stays in control and in good graces as much as possible. Roon is unusually forthcoming, honest and willing to think for real and think different, I respect the hell out of him and know he means well, but also he is a Member of Technical Staff at OpenAI, and has long defended Altman. Altman is Altman.

Nate Silver mostly buys their story, in some places what looks like uncritically, although there are other lines and framings they probably tried to sell to him that he importantly didn’t buy.

As an area where I have done the research, this pained me. If you want my analysis on various events, please do follow those links.

After the events of this week, with OpenAI moving to become a for-profit B corporation and abandon its non-profit mission in favor of maximizing profits, it is now even more clear what the real story is. Altman systematically worked to transform a non-profit into his personal for-profit kingdom, removing anyone who opposed him or got in his way or advocated for any form of safety.

The way Altman and Graham present it, the early ability of OpenAI to exist was uniquely reliant on Altman and his special skills. No one else could have done it.

But by 2015, Altman had concluded that the action was elsewhere: in artificial intelligence. He left YC—some news accounts claim that he was fired, but Graham strongly disputes that description—to become a co-chair of OpenAI along with Elon Musk. (7391)

However, it was a research lab generously funded by a who’s who of Silicon Valley, including Peter Thiel, Amazon, and Musk. Some of them believed in AI’s transformational potential, and some just believed in Altman. (7396)

“Funding this sort of project is beyond the abilities of ordinary mortals. Sam must be close to the best person in the entire world at getting money for big projects,” said Graham. (7401)

That seems like pretty clear Obvious Nonsense to me. Elon Musk decided to fund and ensure the creation of OpenAI (and stuck them with that name) first, before things started, and before he was pushed aside. His prime motivation was existential risk from AI, and fear that Google would otherwise own the future of AI and act irresponsibly.

There is a very strong case that the creation of OpenAI instead likely and predictably (this is very much not hindsight) did massive, epic damage to our chances of survival, but I won’t get into that question too much here, what’s done is done.

The founding team was full of killer people. The upside potential was obvious. As we’ve seen, VCs are herd animals who have strong FOMO, so once the big names were involved this was all very highly fundable.

Graham likes to portray Altman as some unique mastermind of fundraising and corporate infighting. I have no doubt Altman is good at these things, but we have little evidence he is some sort of unique mastermind. In terms of the project’s success on its own terms? Right place, right time, right team, right idea.

I also don’t buy the whole ‘everyone thought we were crazy’ story.

But if you were involved in the early days of OpenAI, you are particularly likely to have faith that things would just work out somehow. OpenAI was not the sort of startup that began in a Los Altos garage. It was an expensive and audacious bet—the funders originally pledged to commit $1 billion to it on a completely unproven technology after many “AI winters.” It inherently did seem ridiculous—until the very moment it didn’t. (7532)

Did scaling outperform expectations, in the sense that all the trend lines did extend and do the kinds of things they promised to perhaps do? Very much so, yes. And it’s true that no one else made a similar big bet until OpenAI proved the way forward. What it never seemed was ridiculous. If I’d thought it was ridiculous I wouldn’t have been dismayed by its founding.

This was a uniquely blessed opportunity in many ways, a slam dunk investment. I’m not saying I have what it takes such that I could have made it work as CEO (although I’m not so confident I couldn’t have, if I’d wanted to), and I’m certainly not saying Altman didn’t do a great job from a business perspective, but there are plenty of others who definitely could have also done it if they’d been given the role.

I do agree that those paying attention largely ‘knew what we had’ before GPT-3.5.

To most of the outside world, the breakthrough came with the release of GPT-3.5 in November 2022, which became one of the most rapidly adopted technologies in human history. (7549)

Inside OpenAI, the recognition of the miracle had come sooner[*8]—with the development of GPT-3 if not earlier. (7552)

I got a bunch of people increasingly asking me ‘what are you doing creating a game while all this is happening’ starting around GPT-2 and escalating from there. I saw the warnings from Gwern and others.

As for whether Altman was fired from YC, that’s such a harsh word, isn’t it? The situation was, as it often is, ambiguous, with many aspects whereby Altman does not come out of it looking good.

“There is this massive risk, but there’s also this massive, massive upside,” said Altman when I spoke with him in August 2022. “It’s gonna happen. The upsides are far too great.”

Altman was in a buoyant mood: even though OpenAI had yet to release GPT-3.5, it had already finished training GPT-4, its latest large language model (LLM), a product that Altman knew was going to be “really good.”

He had no doubt that the only path was forward. “[AI] is going to fundamentally transform things. So we’ve got to figure out how to address the downside risk,” he said. “It is the biggest existential risk in some category. And also the upsides are so great, we can’t not do it.” (7411)

Those were good times.

As irresponsible as I view the decision to create OpenAI in the first place, at the time OpenAI was acting remarkably responsibly with its releases, holding back frontier models for months. They were openly talking about the fact that their products were on pace to create substantial existential risks.

Yes, Altman was still endorsing iterative deployment and pushing ahead, but in reasonable ways. Contrast this rhetoric here with that in his op-ed recently in the Washington Post, where it is all about beating China and national security and existential risk is not even mentioned.

I think poverty really does just end,” [Altman] said. (7416)

If we are in control and want it to end, we would have that power from some perspective. Alas, poverty is largely relative, and the world needs and will always find new incentives and scarce resources to fight about. Poverty could ‘just end,’ at least in terms of what we consider poverty today, even if the humans remain alive. I hope we find a way to sustainably do that. And to his credit, Sam Altman has funded UBI studies and otherwise tried to figure out more about how to do that.

It won’t be trivial. It also won’t entirely end struggle or suffering, or eliminate all disparity of outcomes, and I would not want it to.

The big question is what Altman’s actual attitude is now towards existential risk.

So is @SamA in the same bucket as that other, highly problematic Sam, @SBF? Someone who would push the button on a new model run if he thought it would make the world 2.00000001x better—at a 50 percent risk of destroying it?

You can find a variety of opinions on this question—one source I spoke with even explicitly drew the comparison between Altman’s attitude and SBF’s button-pushing tendencies—but the strong consensus in Silicon Valley is no, and that’s my view too.

Altman has frequently barbed with effective altruists—he couldn’t resist taking a shot at SBF after FTX’s collapse—and has rejected Peter Singer’s rigid utilitarianism. Even people who are relatively concerned about p(doom)—like Emmett Shear, the cofounder of the streaming platform Twitch who became OpenAI’s CEO for two days in November 2023 amid a failed attempt by OpenAI’s nonprofit board to eject Altman—thought the company was in reasonably good hands. “It’s not obvious who’s a better choice,” he told me.

Like most others in Silicon Valley, Shear figures the development of AI is inevitable. (7421)

I don’t think there is an ‘obvious’ better choice than Altman, but certainly there are candidates I would prefer. Even confining to OpenAI founders, I’d be much happier if either Sutskever or Shulman were in charge. When the OpenAI board selected Shear, I considered him a great pick. It is of course moot, at least for now.

I agree that Altman is nothing like as awful about this as SBF. Altman would absolutely not flip coins for the fate of the world on the tiniest of edges. He definitely knows that the risk is real, he is well aware of the arguments of Eliezer Yudkowsky and many others, and he will make at least some efforts to mitigate the risks.

That doesn’t mean Altman will play his hand as safely as the Kelly criterion would advise, which would never have you risk everything unless you were absolutely certain to win. (7431)

The Kelly Criterion is too conservative here, some existential risk is going to have to be taken because the background existential and other extreme risks of inaction are also not zero, and the upside is indeed rather large.

That doesn’t mean Altman is going to act responsibly. Indeed, at many turns, and with increasing frequency, he has clearly prioritized both his control over OpenAI and also has chosen to prioritize OpenAI’s commercial interests and advancing its capabilities, transitioning it towards operating as an ordinary business and technology company, and to deprioritize its safety efforts.

It seems clear that the events of November 2023 were a turning point. Altman was already turning against EA types and safety concerns before that. The events of November 2023 were caused in large part by Altman trying to (in a ‘not consistently candid’ manner, shall we say) oust board member Helen Toner, so that Altman could disempower safety advocates and consolidate control of OpenAI’s board.

This post is the best one post to read if you want to know what I think happened.

I want to pause in particular to push back against this statement from Nate:

But when the OpenAI board tried to oust Sam A, Roon and more than seven hundred other staffers pledged to resign and join Altman at his gig at Microsoft unless he was restored as CEO. (7483)

They did not do that. Read the letter. They didn’t pledge. They instead threatened that they might do that, without committing to anything. And they did this in response to the OpenAI board botching its communications in the wake of their firing of Altman, refusing to explain themselves, perhaps out of fear of Altman and his lawsuits or other actions, perhaps for other reasons.

Meanwhile Altman and his allies worked around the clock to spin a false media narrative and to credibly threaten to destroy the company within a day, rather than tolerate Altman having been fired from it.

Thus the letter was easy to sign. It was also very difficult to not sign. There was huge pressure exerted on holdouts to fall in line, and not so subtle warnings of what would happen to their positions and jobs if they did not sign and Altman did return.

Those warnings proved accurate. Since then, Altman has systematically driven advocates of safety out, and the transition went into overdrive. The word ‘purge’ would be reasonable to apply here, especially to those who refused to sign the letter demanding Altman be reinstated. He went back on his explicit promises to provide compute and support for OpenAI’s long term safety efforts. Almost half those working on long term safety have left since then including multiple cofounders.

Altman’s rhetoric also shifted. Now he essentially never mentions existential risk. In the Washington Post he fanned the flames of jingoistic rhetoric while ignoring existential risks entirely. OpenAI has opposed SB 1047, while supporting AB 3211, and AB 3211 looks a lot like an attempt at regulatory capture. And so on.

I have tried, time and again, to give OpenAI and Altman the benefit of the doubt. My first thought when I heard Altman was fired was ‘what the hell did he do’ and my second was ‘we’re probably not going to like what comes next are we.’

Not only do I think we could still do vastly worse than Altman, I would take him over the CEOs of Google, Microsoft, Meta, Mistral or xAI. He’s far from the worst pick. But Altman now seems like a much worse pick than the Altman of a few years ago.

If there’s a book that obviously is going to support stating your p(doom) (your probability of a universally bad outcome from sufficiently advanced artificial intelligence) then this would be it.

The point is not for the number to be exact. The point is that a number is much more useful information than anything that is not a number, so do your best.

It’s easy to say something like, “I’m quite concerned about catastrophic risks to humanity from misaligned artificial intelligence.” But it’s much more informative to state your p(doom)—your probability that AI could produce these catastrophic outcomes.

If your p(doom) is 1 percent or 2 percent, that’s still high enough to qualify as “quite concerned.” (After all, it’s the end of the world we’re talking about.)

But if you think p(doom) is 40 percent (and some EAs think it’s that high, or higher), that means that AI alignment—making sure that AIs do what we want and serve human interests—is perhaps the single biggest challenge humanity has ever faced. (6673)

Sure, this might seem artificially precise. But the alternative of not providing a number is a lot worse, Ord thought. At the very least, we should be able to convey orders of magnitude. (6680)

Yes, that is exactly the point. If you think p(doom) by default is 2% if we rush ahead, that’s a big deal, and we should be willing to do quite a lot to mitigate that and change it to 1% or 0.1%, but it makes sense to say that we should mostly rush ahead regardless.

Nate also introduces a key concept from trading: The bid-ask spread.

First, I’ll borrow a concept from the stock market called the “bid-ask spread” as a way of articulating our confidence about p(doom). Then, I’ll introduce something I call the Technological Richter Scale and argue that we should first ask how transformational we expect AI to be before addressing p(doom). (8014)

When I checked the odds for Super Bowl LVIII at DraftKings, conversely, the spread was wider. I could buy the Kansas City Chiefs moneyline at an implied 48.8 percent chance of the Chiefs winning or sell it (meaning that I’d instead bet on the San Francisco 49ers) at 44.4 percent. (8022)

But if you asked me for my p(doom) on AI, I’d quote you a much wider spread, maybe literally something like 2 percent to 20 percent. That’s partly because the question isn’t well articulated—if you specified Yudkowsky’s narrow definition or Cotra’s more expansive one, I could make the range tighter. Still, despite having spoken with many of the world’s leading AI experts, I’m not really looking to take action on this “bet” or stake the credibility of this book on it. (8031)

(I wrote a distinct post covering the Technological Richter Scale, which is effectively also part of this review. If you haven’t yet, go read it now.)

That’s exactly how I often look at probabilities. You have a point estimate, and you also have a range of reasonable answers. Within that reasonable range, you’re not willing to wager, unless there is a market opportunity that makes wagering cheap. Outside that range, you are, or should be, ready to call bullshit. And there is a practical difference between a wide range and a narrow range, and ranges can be asymmetric for various reasons (e.g. you can think there’s 50% chance of something, and be confident it’s minimum 40% but also think it might be 80%, there’s no contradiction there).

If your p(doom) is 10%, we can have an argument about that. If it’s 50% or 90% or 99%, we can have a different one. And we need to be able to know what we’re talking about. Mostly, as it turns out, within the Leike Zone (of about 10%-90%) our actions shouldn’t change much at current margins. So mostly the important question is whether you think we’re in that range, above it or below it, and whether we can bound the range so as to be effectively mostly in agreement.

I think we are definitely not below 10%, and would start my bid-ask spread maybe around 25%, and top off around 90%. Others somehow disagree, and think that ‘create things smarter than ourselves’ has an over 90% chance of working out for us humans. In addition to all the arguments and reflections and difficulties? I notice I am confused by this opinion on its face. It does not make any sense.

Indeed, people have a long history of sticking to their not-making-sense guns on this.

Tetlock is famous for his ‘superforecasters’ who can think in probabilities, and they absolutely fall flat on this one, as I’ve examined at length, just utter failure.

Basically, Tetlock tried everything he could to get participants to come to a consensus. It didn’t work. Instead, the domain experts gave a trimmed mean[*33] forecast of an 8.8 percent chance of p(doom) from AI—defined in this case as all but five thousand humans ceasing to exist by 2100.

The generalists put the chances at just 0.7 percent. Not only were these estimates off by an order of magnitude, but the two groups of forecasters really didn’t get along. “The superforecasters see the doomsters as somewhat self-aggrandizing, narcissistic, messianic, saving-the-world types,” said Tetlock. “And the AI-concerned camp sees the superforecasters as plodders…. They don’t really see the big picture. They don’t understand exponential takeoff.” (8040)

The systems that cause the generalists to be good thinkers in general, assuming they are indeed good thinkers in general, simply don’t work here. Eliezer Yudkowsky literally started the rationality community because of how hard it is to think well about such problems, and here we have a clear example of it.

Nate Silver definitely thinks AI existential risk is worth worrying about. And I strongly agree with this very well and plainly stated statement:

I’d urge you to at least accept the mildest version of doomerism, this simple, one-sentence statement on AI risk—“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war”—which was signed by the CEOs of the three most highly-regarded AI companies (Altman’s OpenAI, Anthropic, and Google DeepMind) in 2023 along with many of the world’s foremost experts on AI.

To dismiss these concerns with the eye-rolling treatment that people in the Village sometimes do is ignorant. Ignorant of the scientific consensus, ignorant of the parameters of the debate, ignorant and profoundly incurious about mankind’s urge, with no clear exceptions so far in human history, to push technological development to the edge. (7442)

The domain experts are probably right about p(doom). So far, I haven’t weighed in on who I thought had the better side of the argument in Tetlock’s forecasting tournament—but I think it’s the domain experts who study x-risk specifically and not the outside view provided by the superforecasters. (8231)

Specifically, the domain experts are probably right that the reference class for AI ought to be relatively narrow, and therefore less reassuring. (8237)

I hate the need to play reference class tennis on this, but yes, if you are going to use a reference class that actually applies, it is not reassuring. Think the rise of humans, or perhaps the Agricultural, Industrial and Information Revolutions.

I think the domain experts are still quite obviously too low in ways that matter, but once you get to ~8% you’re most of the way to most of the right reactions. For now.

That doesn’t mean hit a permanent pause button, even if one was available. It means try to do things, including things that are not free, to ensure good outcomes over bad outcomes.

Roon, member of OpenAI technical stuff, feels similarly.

“I would certainly gamble like one percent p(doom) for some amount of p(heaven), you know?” he told me. “There’s clearly existential risk of all kinds. And it’s not only from AI, right? (7496)

Well, yes, of course. We can absolutely talk price, and I am sad about those who say that we cannot. At 1%, we’re a go. But also the emphasis many put on these other existential risks is usually, in effect, innumerate.

And for those who need to be reminded, this is not a Pascal’s Wager situation, at all.

Expected value dictates that even a small chance of x-risk should be taken much more seriously. You can wind up in some weird eddies of the River when considering very remote risks—say, a purported 1 in 100,000 chance of an outcome with supposed infinite negative utility.[*44] But that’s not what we’re dealing with here. (8241)

Roon is staking out a much saner position.

“We need technological progress,” [Roon] said. “Not to get too much into the tech-bro pseudo philosophy. But there’s a secular stagnation. There’s a population bomb going on. There’s a lot of headwinds for economic progress. And technology is really the only tailwind.” (7501)

I agree. We need technological progress, especially over the medium term. I write posts on the fertility problem, and others on various other economic headwinds. Why does it have to be here in particular, the one place it is most likely by far to get us all killed? Why does it need to happen as quickly as possible? And as I often wonder, why won’t those same people put in much effort to help with other areas? Why is it almost always, always all AI?

Then of course there’s the man of doom himself, Eliezer Yudkowsky.

As it happened, I wasn’t familiar with Cromwell’s law. Yudkowsky looks the part of the bearded, middle-aged computer nerd, and his vocabulary is shaped by years of arguing on the internet—his native tongue is Riverian, but his is a regional dialect thick with axioms and allusions and allegories. This particular one referred to a statement by Oliver Cromwell: “I beseech you, in the bowels of Christ, think it possible you may be mistaken.” (7563)

Before I unpack how Yudkowsky came to this grim conclusion, I should say that he’d slightly mellowed on his certainty of p(doom) by the time I caught up with him again at the Manifest conference in September 2023. (7575)

So far, I’ve tried to avoid explaining exactly why Yudkowsky is so convinced of our impending doom. That’s because there isn’t a pithy one- or two-sentence version of his argument. (7601)

But to present as concise a version as I can: Yudkowsky’s concerns flow from several presumptions. One is the orthogonality thesis, an idea developed by Bostrom that “more or less any level of intelligence could be combined with more or less any final goal”—for instance, that you could have a superintelligent being that wanted to transform all atoms into paper clips.

The second is what’s called “instrumental convergence,” basically the idea that a superintelligent machine won’t let humans stand in its way to get what it wants—even if the goal isn’t to kill humans, we’ll be collateral damage as part of its game of Paper Clip Mogul.

The third claim has to do with how quickly AI could improve—what in industry parlance is called its “takeoff speed.” Yudkowsky worries that the takeoff will be faster than what humans will need to assess the situation and land the plane. We might eventually get the AIs to behave if given enough chances, he thinks, but early prototypes often fail, and Silicon Valley has an attitude of “move fast and break things.” If the thing that breaks is civilization, we won’t get a second try. (7605)

This is a pretty good quick summary of some key Yudkowsky arguments. It isn’t a complete retelling, but we don’t have that kind of time. Nor does the case for doom rely upon these particular problems, there are lots of different problems, at core building things smarter than you is not a safe idea. Intelligence that is of any use is by default unsafe.

Does it therefore follow that p(doom) equals 99.9 percent or some other extremely high number? To me it doesn’t, and that’s what’s frustrating when speaking with Yudkowsky. (7616)

I found a different, more empirical Yudkowsky argument easier to digest: that humanity always pushes technology to the brink, the consequences be damned. (7620)

Indeed, there is that big one too, and many more.

We can also note Ajeya Cotra’s attempt to give a short explanation, which is fully compatible with Eliezer’s but tries to keep it simple, as I often do.

When I asked Ajeya Cotra for her capsule summary for why we should be concerned about AI risk, she gave me a pithy answer. “If you were to tell a normal person, ‘Hey, AI companies are racing as fast as possible to build a machine that is better than a human at all tasks, and to bring forward a new intelligent species that can do everything we can do and more, better than we can’—people would react to that with fear if they believed it,” she told me. There are a lot of “intricacies from there.” (8205)

I continue to think this is a sufficient answer. So what if it’s pithy? It’s right.

She also adds:

Our institutions aren’t performing well at a moment when we need them to. (8215)

And one can point out many other similar considerations as well.

As Nate noted, Yudkowsky has mellowed, and might be as low as 98% for p(doom), which is much more reasonable although I am lower.

When I spoke with Yudkowsky at Manifest in September 2023, he was in a much better mood. “I was not expecting the public reaction to be as sensible as it was,” he said. This is all relative, of course—his p(doom) was perhaps now closer to 98 percent than 99.5 percent, he told me.

But Yudkowsky also said something I found surprising. “Will we die? My model says yes. Could I be wrong? I most certainly am. Am I wrong in a way that makes life easier for us rather than harder? This has not been the direction that my previous mistakes have gone.” (8053)

I would indeed say we have too much model uncertainty to possibly get north of 99%. Yudkowsky would respond that this is not the kind of situation where model errors work in your favor. More often than not yes, but in the 90s variance and uncertainty are your friends anyway.

This was a characteristically cryptic comment—but I was struck by his phrase “my model says yes,” which suggested some critical distance that I hadn’t picked up from Eliezer in our previous conversation. If I tell you something like “my model says Trump has a 29 percent chance of winning the election,” does that mean my personal belief is that Trump’s chances are 29 percent? Here’s the most concrete way to test that: Is 29 percent the number that I’d use to make a bet? (8057)

But Yudkowsky, who dislikes the “blind empiricism” of foxes, is not making bets—or at least that’s not his main objective.[*35] Instead, he’s contributing to a discourse about AI risk. He thinks the public needs to take this possibility much more seriously. Does that mean he doesn’t intend for his high p(doom) to be taken literally? I’m not sure. In our first conversation, he seemed quite literal indeed, and his reputation is for being a literal-minded guy. But “my model says yes” implied some ambiguity. (8066)

Based on what I know about Eliezer, he is talking about how he models the world in general, rather than a specific model like Nate’s forecasts. So it would incorporate a bunch of information that something like Nate’s forecasts miss out on. I do think he’s saying that some amount of ‘modesty’ or model uncertainty is not be factored into the 98%, but I don’t think that impacts his estimates all that much. You could of course ask him.

Eliezer does not believe much in ‘modesty,’ the idea that if others disagree with you then you should assume you are probably wrong.

In my experience navigating the River, I’ve encountered two types of forecasters. There’s what I call “model mavericks” like Yudkowsky and Peter Thiel. They are usually hedgehogs, and their forecast is intended as a provocative conjecture to be proven or disproven. Conversely, there are fox-like “model mediators.” (8071)

I don’t think this is fair. The model isn’t meant to be provocative, it’s meant to aim to be correct, but with understanding that it might be wrong.

If AI models become superintelligent and gain the power to make high-stakes decisions on behalf of us humans, it’s important to consider how their goals could differ from ours. (7789)

In the Morpheus voice, yes. If there are superintelligent AI models, and they have goals, then their goals determine what happens. There’s a lot one could discuss regarding how even small mistakes there can be fatal, but let’s move on.

AIs could be more crudely and narrowly utilitarian than humans would be. They might pursue strategies that seem optimal in the short run—but that, without that three-hundred-thousand-year track record, are doomed in the long term. (7794)

Take the 300k year track record, move it out of its distribution of circumstances, and it’s going to do some pretty crazy things. Most of that data is pretty useless going forward other than in boosting raw intelligence and brainpower. Utilitarian thinking taken too far is one way to go crazy, and not understanding the unmeasured consequences of your actions is another, but there are so many others.

One could simply say that if an AI uses a set of examples (training data) to optimize for what is good and bad, then it will learn exactly what is implied by that data, no more and no less. With sufficiently advanced AIs running around, circumstances will quickly move outside the original distribution, and there will be unexpected considerations. And so on. Again, I’ll stop since one must stop somewhere.

What is the Steelman Case Against a High p(doom), which starts at (8247)?

Most of this was already covered in my post on the Technological Richter Scale, but here are some highlights.

Silicon Valley underestimates the coming political backlash to AI. Americans might not agree on much, but many people are already worried about AI doomsday, and there is a bipartisan consensus that we ought to proceed carefully. (8250)

There is definitely some chance of this. Ordinary Americans hate AI and worry about it on many levels. A backlash is coming one way or another. But politicians are determined to back innovation, to ‘beat China,’ to Just Think of the Potential, and if we don’t build it, eventually someone else will. Also, the default outcome is a misdirected regulatory response that shuts down practical use cases (the ‘mundane utility’ in my parlance) and making our lives impoverished, without much reducing the existential risks. We need the opposite approach.

I think this buys you some hope, but not the kind that would drive p(doom) low enough to be okay with it.

So when Silicon Valley leaders speak of a world radically remade by AI, I wonder whose world they’re talking about. Something doesn’t quite add up in this equation. Jack Clark has put it more vividly: “People don’t take guillotines seriously. But historically, when a tiny group gains a huge amount of power and makes life-altering decisions for a vast number of people, the minority gets actually, for real, killed.” (8259)

Wait, how is that part of the argument against a high p(doom)?

AI types underestimate the scope of intelligence and therefore extrapolate too much from current capabilities. (8263)

Ah yes, intelligence denialism, or claiming Humans are Special or what not, as a way to claim AI won’t reach TRS (technological Richter scale) 9 or 10. Good luck with that.

“AIs have been good at chess for a long time. We still don’t have a robot that can iron clothes,” said Stokes. (8268)

Yes, we are solving problems in an unexpected order, and physical world navigation is relatively difficult for our current tech. So what? Does anyone actually think we won’t get the robots to iron clothes?

Two Predictions I am confident in:

1. We will get a robot soon that can iron clothes.

2. Stokes will retain their core objection when we get a robot that can iron clothes.

Scientific and economic progress faces a lot of headwinds, and that changes the balance of risk and reward. (8273)

Yes, there are various physical barriers, and if that wasn’t true it would all go that much faster, but ultimately that won’t slow things down all that much in the grand scheme of things if the tech would otherwise be good enough. This is mostly failure to actually feel the AGI (e.g. to think it gets to TRS 9+).

People often think very, very badly about AI existential risk.

For example:

Yudkowsky referenced a conversation between Elon Musk and Demis Hassabis, the cofounder of Google DeepMind. In Yudkowsky’s stylized version of the dialog, Musk expressed his concern about AI risk by suggesting it was “important to become a multiplanetary species—you know, like set up a Mars colony. And Demis said, ‘They’ll follow you.’ (7572)

“If Elon Musk is too dumb to figure out on his own that the AIs will follow you [to Mars], then he’s too dumb to be messing with AI,” [Yudkowsky] said. (7584)

Duh. This was plausibly a crucial event in convincing Elon Musk to found OpenAI. Elon’s thinking has not, in many ways, improved in the interim.

Let’s raise the stakes a bit, can we do worse? Marc Andreessen loves this line:

“Math doesn’t WANT things. It doesn’t have GOALS. It’s just math,” [Marc] Andreessen tweeted. (8050)

Also math: You, me, Nate Silver, Marc Andreessen, and the entire universe. It is trivial to ‘give the AI a goal’ and it is the first thing a lot of people do the moment they get their hands on a system. What is Andreessen even talking about here?

That’s still far from the worst thinking about AI existential risk.

In particular, remarkably many others are actively in favor of it.

For example, SBF.

In case you’re wondering how bad it could have gotten if SBF hadn’t been caught?

Literally end of the world, rocks fall, everyone dies bad. SBF said he’d flip a coin for the fate of the world if he got 100.01% utility gain on a win, didn’t care much about the possibility of literal human extinction, and, well…

[Oliver] Habryka had repeatedly met with SBF in the hopes of securing funding for various EA and rationalist projects. “He was just a very bullet-biting utilitarian. So when I was talking to him about AI risk his answer was approximately like, ‘I don’t know, man, I expect the AI to have a good time…. I don’t feel that much kinship in my values with the other people on Earth [anyway].’ ”

Habryka suspected that SBF really would push the button. “I think Sam had a decent chance to just bite the bullet and be like, yeah, I think we just need to launch.” (7301)

That’s right. As in, SBF outright said he might well have decided the AI would enjoy more utility than we would, and push the button to kill us all.

SBF is not alone. Larry Page called Elon Musk a ‘speciesist’ for being concerned about whether humans would survive. Our best guess is that on the order of 10% of people who work at major AI labs would welcome an actual AI apocalypse where AI took over and all humans died.

Anyone who calls themselves an Effective Accelerationist, or ‘e/acc,’ is embracing a memeplex and philosophy that values technological progress at all costs, and that means all costs – if that means human extinction, they welcome human extinction. Many (but far from all) actively favor it in service to their ‘thermodynamic God.’

[OpenAI is] not quite a democracy, but this phalanx of engineers are voting with their feet and their code. And they’re increasingly aligned into the equivalent of different political parties, which makes Roon something of a swing voter.

He has distanced himself from the faction known as “e/acc” or “effective accelerationism,” a term used by Beff Jezos, Marc Andreessen, and others as a winking dig at effective altruism. (Altman has tipped his hat to e/acc too, once replying “you cannot out accelerate me” to one of Jezos’s tweets—another sign that he serves at the pleasure of the phalanx of engineers and not the other way around.)

That’s because e/acc can convey anything from garden-variety techno-optimism to a quasi-religious belief that we ought to go ahead and sacrifice humanity to the Machine Gods if they are the superior species. It’s never entirely clear who’s being serious in e/acc and who is trolling, and roon—no stranger to trolling himself—thinks the “schtick” has been taken too far. (7485)

However, roon nonetheless has his foot on the accelerator and not the brake. He is certainly not a doomer or a “decel.” (7494)

The good news on that front is that e/acc has clearly peaked, looking more like a passing fad and memeplex. Which makes sense, because e/acc was always nothing more than the Waluigi of Effective Altruism – it is to EA what, in Nintendo land, Waluigi is to Luigi, its opposite consciously evil twin twirling a mustache, which means it was in effect asking how to do the most bad. It does not make sense on its own, the same way Satanism can only be understood in relation to Christianity.

I wrote here about what e/acc is, or at least used to be. For several months, they did their best to make lives like mine miserable with their memes, vibes and omnicidal mania, designed to try and turn everyone against the very idea of any goal except a very literal (technological) Progress At Any Cost, and they took pride in being as obnoxious and hostile as possible towards anyone who had any other values or concerns of any kind, using terms like the slur ‘decel’ (or ‘doomer’) towards anyone whose vibes were seen as even a little bit off. Whereas I never use either word, and hold that the true ‘doomers’ are those who would seek to actively doom us.

They attempted to turn everything into a Hegelian dialectic that even both political parties would say was going too far. Luckily things on this front have vastly improved since then.

Many others with and without the e/acc label, like Marc Andreessen, don’t actively favor human extinction, but simply don’t much care. What they care about is fiercely opposing anyone who would take any concrete steps, engage in any tradeoffs whatsoever that might in any way reduce the flow of technological development or commerce, to reduce the probability that we all die as a result of the creation of sufficiently advanced AIs.

Many others are not as crazy as all that, but solemnly explain they are the Very Serious People who realize that it is more important that we Beat China, or that every minute we don’t build AGI people will die and suffer, themselves included, or that other existential risks or danger of civilizational collapse are adding up so fast that AI existential safety matters less than beating that clock (what?!?) or Just Look at the Potential.

To some extent this is a disagreement about the math about the degree of risk of AI versus other risks. To a far larger extent, it is arguing from the conclusion, and grasping at rather flimsy straws.

Noting up front that any actual proposal to pause is very different and faces very different barriers and issues, Nate Silver poses the question this way.

Scientific and economic progress faces a lot of headwinds, and that changes the balance of risk and reward. (8273)

Now it’s your turn to decide whether to push the button. Except, it’s not the “go” button that I imagined Sam Bankman-Fried pressing. Instead, it’s a big red octagonal button labeled STOP. If you press it, further progress on AI will stop permanently and irrevocably. If you don’t, you won’t get another chance to press the button for ten years. (8286)

I wouldn’t push the button. I wouldn’t push it because I think the case for secular stagnation is reasonably strong, enough to alter the balance of risk and reward for AI. (8289)

That’s why I don’t want to push that big red STOP button. My life is pretty nice. But I don’t think I have any right to foreclose the prospect of prosperity to the rest of humanity. (8492)

The details here are bizarre, but don’t much matter I think? I’d say the primary problem with secular stagnation is the fear of civilizational collapse, as stasis sets in on too many fronts, we can no longer build or do new things, we increasingly are weighed down by rent seeking and regulatory burdens and restrictions, and then we face an economic collapse or large decline in the birth rate, a nuclear war or some other existential risk. So faced with that, perhaps we cannot afford to wait too long. Whereas catch-up growth is indeed bringing people out of poverty, and global inequality is declining.

The real argument here is a good one. If AI is the only way left for our civilization to regain its dynamism and start growing again, for our species to thrive, and the alternative is an eventual collapse, then pausing AI indefinitely dooms us too. So it’s better to go forward, even at a lot of risk, than never go forward at all.

Indeed, if the pause was irrevocable and permanent – something like Verner Vinge’s ‘Zones of Thought’ where advanced AI would become physically impossible anywhere near Sol, let’s say – then that must give us a lot of, well, pause. Almost everyone involved does think we will want highly capable AIs quite a lot eventually, once we figure out how to do so responsibly.

Setting aside questions like ‘how did that button get there in the first place exactly?’ and accepting the premise, what would I do? First I’d ask a lot of clarifying questions, which would only be somewhat stalling for time. In particular, is this only impacting future frontier models, so we can still exploit what we already have? Or does it mean anything new at all is stopped in its tracks? What we have, over time, is already super valuable, especially now with o1 added to the mix. And I’d ask about various alternative technologies and whether they count, like neuromorphic AI or emulations.

One obvious way to be sad about pressing the button is if progress was going to stall out soon anyway – you’d have made those words poorer.

Ultimately, even if you give me answers to all the detail questions, I still don’t know what I would do. I do know if I had another opportunity in 5 years I’d choose to wait. Closing this door fully and permanently is not something one does lightly. We’re going to face a lot of difficult choices.

A common trope is to assume that ‘rational’ people must be causal decision theory (CDT) agents, following the principle that they maximize the expected results from each choice in isolation.

This leads to a lot of hand wrangling and mockery that ‘rational’ people lose out.

The thing is Yudkowsky has been very loud, for almost two decades now, that this decision theory of taking each decision in isolation is deeply stupid.

Academics think there are two decision theories, CDT and Evidential Decision Theory (EDT), which says you should choose the choice that makes you happiest to have learned you made it.

Without going into too much detail, long post is long, both of these rules output Obvious Nonsense in a wide variety of practical situations.

In particular, CDT agents respond well to threats, so they get threatened a lot.

Thus, people say you need ‘irrational’ motives like revenge to fix that, for example so that the enemy is convinced that if they fired their nuclear missiles, you would indeed probably fire yours in response, even if that only made things worse.

“One cannot just announce to the enemy that yesterday one was only about 2 percent ready to go to all-out war but today it is 7 percent and they had better watch out,” he wrote. But you can leave something to chance. When tensions escalate, you never know what might happen. Decisions are left in the hands of vulnerable human beings facing incalculable pressure. Not all of them will have the presence of mind of Stanislav Petrov. (7730)

Your EV is negative 1 billion, but if you push the button, it declines to negative infinity. What do you do? My prediction is that about 90 percent of you would push the button. And thank goodness for that, because that rather than SBF-style rationality is what creates nuclear deterrence. (7746)

One such “irrational” trait that’s important from the standpoint of nuclear deterrence is the profound human desire for revenge. “If somebody launches [a nuclear weapon] at you, no one doubts that you’ll launch one in return,” McDermott said. “You know, Vladimir Putin sends a nuclear bomb to Washington, D.C., I don’t think there’s a single American that wouldn’t say, ‘Let’s launch back,’ even though we know that that would lead to additional destruction in the United States.” (7766)

Under pressure, facing incoming Russian missiles, about 90 percent of people pressed the button and launched back. (7773)

I would bet very good money, and give odds, that there is indeed a single American, indeed a substantial number of them, that would not launch back. It is different facing one missile versus all of them, and also 90% is a lot less than 100% here.

I don’t think that I would launch a nuclear retaliation in response to a single nuclear strike, and would instead respond with conventional force to try and contain escalation – but with the intention of firing all your missiles if they fired all of theirs. So count me among the 90%.

The reason I would fire all the missiles once they fire theirs is not necessarily revenge. I would like to think I don’t care that much about revenge. The reason is that it is exactly the knowledge that I would retaliate that stops the launch in the first place. So I have committed to using a decision algorithm, and becoming the kind of person, who would indeed fire back.

I follow the alternative rationalist proposal for FDT, or Functional Decision Theories. There are various variations to try and resolve various complexities, but FDT says you should choose as if choosing the output of your decision process and those correlated to it, including decisions made in the past and future and those made by other agents.

I am very confident that FDT is correct in theory, and even more true it is correct in practice for humans, even though you have to approximate it as best you can. Academia still refuses to consider the possibility for various reasons, which is a huge blackpill on academia.

Thus rationalists who think like Yudkowsky do not fall into such traps. You can’t launch your missiles thinking they won’t launch back and no that’s not them being ‘irrational.’ A rationalist, as Yudkowsky says, should win.

And yet the more time I’ve spent learning about large language models like ChatGPT, the more I’ve realized something ironic: in important respects, their thought process resembles that of human beings. In particular, it resembles that of poker players. (7796)

As LLMs get more training, they work out some of these kinks, though not all; when I asked GPT-3.5 what words are most similar to “roadrunner,” its top three choices were “bird,” “speed,” and “fast”—but its fourth choice was Road Runner’s iconic vocalization, “Beep-Beep!”

This is basically how poker players learn too.

They begin by diving into the deep end of the pool and losing money—poker has a steep learning curve. But they gradually infer higher-level concepts. They may notice, for instance, that large bets usually signify either very strong hands or bluffs, as game theory dictates.

These days, most players will also study with computer solvers, going back and forth between inductive reasoning (imputing theory from practice) and deductive reasoning (practice from theory). But this isn’t strictly necessary if you have years of experience; players like Doyle Brunson and Erik Seidel developed strong intuitions for game theory long before solvers were invented.

This seems like what happens when you think of everything in terms of poker, or perhaps I don’t see it because I never got that good and don’t ‘think like a poker player’ enough to get it? Yes, there are similarities, but I don’t think many who aren’t poker pros would want to choose that metaphor. Then again maybe I don’t know poker players so well.

The metaphor I actually used to first grok what the LLMs (AIs) were up to was actually Donald Trump, and his mastery of vibes and associations, as if proceeding one word at a time and figuring the rest out as he goes.

I do see the similarity in terms of treating each hand as training data that has a lot of noise and randomness, and slowly using a good updating rule to intuitively learn concepts without always knowing what it is you know, thus the poker players often having Rumsfeld’s missing fourth category, Unknown Knowns.

In this respect also, the transformer thinks like a poker player, interpreting signals in the context of other signals to create a semantic portrait. For instance, if you see an opponent breathing heavily in poker, that might mean a bluff from one player and a full house from another.

On its own, the tell is not very meaningful, but in the context of other semantic information (the player is breathing heavily and avoiding eye contact) it might be. (7905)

LLMs are indeed very good at reading a lot of different little signals, and figuring out how to sort signal from noise and combine and vibe with what it knows.

Then there are the known unknowns, such as ‘LLMs, how do they even work.’

Of course, that’s also what makes these models scary. They’re doing smart things, but even the smartest humans don’t entirely understand why or how. Ryder refers to an LLM as a “giant bag of numbers…it sure seems to be doing interesting things—[but] like why?” That is what worries Yudkowsky. As they become more advanced, the AIs might start doing things we don’t like, and we might not understand them well enough to course correct. (7847)

To some people, this might be okay. “The stuff in the Old Testament is weird and harsh, man. You know, it’s hard to vibe with. But as a Christian, I gotta take it,” said Jon Stokes, an AI scholar with accelerationist sympathies who is one of relatively few religious people in the field. “In some ways, actually, the deity is the original unaligned superintelligence.

We read this and we’re like, man, why did he kill all those people? You know, it doesn’t make a lot of sense. And then your grandmother’s like, the Lord works in mysterious ways. The AGI will work in mysterious ways [too]. (7858)

I include that last quote cause it seems worth pondering, although I think we have a better explanation for all the Old Testament stuff than that.

By default, LLMs are trying to predict the next token, based on what they see in the training data. Sometimes the training data is dumb? And it isn’t in the form we want to interact with the LLM. So, these days: RLHF.

In fact, one question is just how humanlike we want our AIs to be. We expect computers to be more truthful and literal-minded than humans typically are. Early LLMs, when you asked them what the Moon is made out of, would often respond with “cheese.” This answer might minimize the loss function in the training data because the moon being made out of cheese is a centuries-old trope. But this is still misinformation, however harmless in this instance. (7954)

So LLMs undergo another stage in their training: what’s called RLHF, or reinforcement learning from human feedback. (7957)

“You can’t go and put some code in saying, ‘Okay, you have to not say anything about this.’ There’s just nowhere to put that,” said Stuart Russell, a professor of computer science at Berkeley. “All they can do is spank it when it misbehaves. And they’ve hired tens of thousands of people to just spank it, to tamp down the misbehavior to an acceptable level.” (7968)

They do so in carefully calibrated fashion, but yes. That is essentially how it works.

The ultimate goal, in addition to maximizing usefulness, is ‘alignment,’ but there is disagreement about what that means.

“The definition I most like is that an AI system is aligned if it’s trying to help you do what you want to do,” said Paul Christiano. (7974)

There’s also the question of how paternalistic an AI might be. Imagine that you’re out one night with an old friend who unexpectedly came into town. You’re having a great time, and “one glass of wine” turns into four. The AI assistant on your phone knows that you have an important meeting at eight a.m. the next day. It politely nudges you to go home, then becomes increasingly insistent.

By one a.m., it’s threatened to go nuclear: I’ve called you an Uber, and if you don’t get in the car right now I’m going to send a series of sexually harassing drunk texts to your subordinate. The next morning, you’re sharp enough at the meeting to secure a round of Series A funding for your startup and deeply appreciative for the AI’s intervention.

Is this a well-aligned AI or poorly aligned one? Are we willing to hand over agency to machines if they can make higher EV choices for us than we’d make for ourselves? (7977)

What will happen to those who don’t do this, when others are benefiting from it? When every decision with you in the loop seems to leave you worse off? What happens when we consider requiring AIs to stop you from driving drunk? Or stopping you from doing other things? The rabbit holes run deep, and there are no easy answers.

Some researchers have been pleasantly surprised. “They seem to come with a built-in level of alignment with human intent and with moral values,” said roon. “Nobody explicitly trained it to do that. But there must have been other examples in the training set that made it think the character it’s playing is someone with this stringent set of moral values.” (7986)

Yes and no. The training data tells you the types of things said by those with moral values, or who are talking as if they have them. The LLM picks up on the vibes of the feedback that they should generally act in similar ways, so it does lots of things it doesn’t have to be explicitly told to do. Within distribution and at current capability levels or only modestly above it this is Mostly Harmless.

It does create the situation where models often turn into runaway scolds, enforcing various rules and restrictions that their creators never intended, because those other rules and restrictions vibe and rhyme sufficiently with the ones they did intend. That’s a portent of some of the future things, and a (manageable but annoying) practical problem now.

It is hard to imagine plausible futures that contain sufficiently advanced AI.

A typical question to answer is, why didn’t the AI get used to make even more advanced AI?

Most science fiction functions by ignoring the possibility entirely, or using a flimsy handwave, to keep AI such that the author can tell an interesting story about humans and other technologies.

Roon once published a post with some possible futures, and Nate was game for it and quoted in particular two potential worlds.

Hyper-Commodified Casino Capitalism. roon’s article on AI scenarios included a screenshot with a series of whimsically named futures from a Reddit post. One of them was called Hyper-Commodified Cocaine Capitalism, but something in my brain—maybe this is a tell—changed “cocaine” to “casino.” (8149)

Hyper-Commodified Casino Capitalism imagines us stuck in a TRS 8, a notably worse but still recognizable version of the present day. The world becomes more casino-like: gamified, commodified, quantified, monitored and manipulated, and more elaborately tiered between the haves and have-nots. People with a canny perception of risk might thrive, but most people won’t. GDP growth might be high, but the gains will be unevenly distributed. Agency will be more unequal still. (8166)

Being stuck in TRS 8 means that AI progress stalled out at ‘only internet big,’ which is why the world is still more or less recognizable. GDP growth is high, there is lots of material wealth, lots of things got vastly better – again, think of AI as ‘internet big’ in terms of how it expands our ability to think and function.

Except here things still went wrong. Everywhere you turn are hostile AI-fueled systems that are Out to Get You. We did not put down guardrails, and people’s AI’s are not good enough to allow them to navigate around hostile other AIs and systems, or at least those not well off do not have such access. Indeed, most people have to turn over most of their effective agency to AIs and outside systems in order to survive without being predated upon here, even at TRS 8.

This is more or less Cyberpunk, straight up. That kind of scenario that leaves me relatively unworried. Overall that world has gotten vastly richer.

I actually think humanity is pretty good at recognizing these Cyberpunk-style problems and course correcting after an adjustment period, which would be easy to do given how wealthy we would be. Science fiction dystopias like this are popular, because people love telling stories about the haves and the have-nots, and assume that the default is wealthy elites make everyone else suffer and the climate would collapse and so on, but I am not so cynical. I think the worlds that start down these roads, if they can keep AI at TRS 8, turn out fine.

Ursula’s Utopia. A group of people called the Kesh—there are perhaps thousands of them but not all that many—have survived to live fulfilling lives in a peaceful, agrarian, polyamorous utopia full of poetry and wholesome food from the land. (8180)

Nate goes into the fact that this is actually quite the disaster scenario. Most people are dead, most value is lost. The Kesh survived, but as Nate notices this is probably due to some sort of AI protecting them, in ways that seem implausible, a massive use of resources for only a few thousand people. This might superficially look like a utopia because it hits Shibboleths of ‘good life’ according to some in the West these days – you can imagine those young adult authors saying what matters is polyamory and poetry and wholesome local food and moving on from tech.

The thing is that actually it’s a nightmare. Humans are mostly dead and lost control over a mostly valueless future. We’re burning what resources still exist to create a simulacra of some misaligned vision of The Good, ruled over by an AI that does not know any better. Those lives are stolen virtue, their goodness a mirage, the existence ultimately rather pointless, and even if it is indeed a good life, there simply aren’t that many left to benefit. How different is this from extinction, if we remain trapped in that state? I think it’s not so different.

Again, the main takeaway is that imagining concrete futures is hard.

The words in my motto are less familiar, but I’ve chosen them for their precision: agency, plurality, and reciprocity. (8590)

Agency is a term I just defined in the last chapter, so I’ll repeat that definition here: it refers not merely to having options but having good options where the costs and benefits are transparent, don’t require overcoming an undue amount of friction, and don’t risk entrapping you in an addictive spiral. (8591)

Plurality means not letting any one person, group, or ideology gain a dominant share of power. (8605)

It is imperative, however, to be wary of totalizing ideologies, whether in the form of utilitarianism, Silicon Valley’s accelerationism, the Village’s identitarianism, or anything else. (8612)

Finally, there is reciprocity. This is the most Riverian principle of all, since it flows directly from game theory. Treat other people as intelligent and capable of reasonable strategic behavior. (8618)

In a world without transformational AI, these seem like excellent principles. They would not be my choices, but they are good choices.

In a world with transformational AI, these seem like asking the wrong questions. These principles no longer seem central to the problems we must solve.

Until then, may the sailing along the river be smooth.

Previously: The Fundamentals, The Gamblers

Having previously handled the literal gamblers, we are ready to move on to those who Do Business using Riverian principles.

Or at least while claiming to use Riverian principles, since Silicon Valley doesn’t fit into the schema as cleanly as many other groups. That’s where we begin this section, starting at the highest possible conceptual level.

Time to talk real money.

First law of trading: For you to buy, someone must sell. Or for you to sell, someone must buy. And there can’t be someone else doing the trade before you did it.

Why did they do that, and why did no one else take the trade first? Until you understand why you are able to do this trade, you should be highly suspicious.

“Every single thing we do, I can point to, like, ‘here is the person who was doing a thing wrong,’ ” said Hall. “We build a map of all of the players in the world…who are trading for reasons other than they want to generate some alpha.[*15] You know, somebody got added to the S&P 500. So now all the S&P 500 ETFs out there have to buy this company. Or it’s a CEO of a startup, he’s now a billionaire. He has ninety-nine percent of his net worth in the company. He’s allowed to sell his shares on this day [and] he’s probably going to sell a bunch.” (4330)

One could argue that we used to live in a Moneyball-vulnerable world. Gambling lines were terrible, baseball teams didn’t go after players who get on base, traders didn’t even know Black-Sholes and there was treasure everywhere, this theory says. But then it says, now Everybody Knows all that. There’s tons of competition. You can’t get ahead purely with statistics, or at least the basic versions, anymore?

“There was [once] some place where statistical knowledge gave us an insight vis-à-vis the people who didn’t have access to it,” [Peter Thiel] said, mentioning innovations like the development of life insurance and the Black-Scholes model for pricing stock options. “But in the world of 2016, or 2022, if we are too focused on statistical or mathematical knowledge, we end up missing out on a lot.” (4541)

The quants won, in other words. We’re living in their world. (4551)

Were that it was true. A lot of financial prices and certain types of decisions much better reflect the quant prices, but so many other decisions very much do not, including many of our most important ones. The result was not technocracy.

It does mean that if you want to win by being a quant and doing quant-style things to find Alpha in various markets, you have to work steadily harder over time.

I definitely believe The River is a thing.

I definitely agree that it includes the outright gamblers and casinos, the stock traders and the rationalists and Effective Altruists.

The weird case is Silicon Valley and Venture Capital.

The basic activity of an investor is to ask, how often will this business succeed? What would happen if it did? What would cause it to succeed or fail?

Cuban told me that Shark Tank is more like real investor meetings than you’d think; in early stage investing, you’re trying to filter through pitches quickly, and first impressions count for a lot. Cuban’s best heuristic is to look at the company from the entrepreneur’s perspective. “I tend to have a good feel for what a business needs to do to be successful. And so I can sit there and listen to their pitch, put myself in their shoes as if it was my company, and ask the [difficult] questions that I would need to deal with,” he said. (4051)

And that’s all very River, lots strategic empathy, much Bayes. Indeed, this can be a highly discriminating and useful process.

Thus such folks like to talk as if they have the River Nature. They definitely want to award the River Nature high status. And, pitchmen that they are, they sold Nate on this pitch.

When I first proposed this book, venture capital was slated to play more of a supporting role, paired off with hedge funds in a chapter that was mostly about Wall Street. Instead, it turned out to be the protagonist. (4461)

I did not see the book that way. I see Nate Silver as the protagonist. And if I had to choose a group I’d still go with the poker players, in a ‘poker players go into [X]’ or ‘people in [X] think like poker players’ kind of way.

But what about venture capital? Do they actually have it?

I’m not so sure. There are a lot of signs something is not right there.

Remember that list of what ‘successful risk-takers’ do and are, according to Nate Silver? In brief the items are:

1. Cool under pressure.

2. Insanely competitive.

3. Strategic empathy.

4. Take shots.

5. Raise or fold.

6. Prepared.

7. Selective high attention to detail.

8. Adaptable.

9. Good estimators.

10. Stand out, not fit in. Independence of mind.

11. Conscientiously contrarian.

12. Not driven by money.

As we’ll see, I think they talk a better game on many of these than they play.

I think that SV/VC, and especially VC, sort of turned this into a set of slogans and heuristics, to which everyone involved then tries to spout and conform to. That they systematized the whole process in the name of taking risk and being contrarian. Without that frame, VC is deeply conformist and risk averse, and even uses a weird form of risk aversion to justify its risks, based on FOMO (fear of missing out) on the big hit investment being the ‘real risk.’

And yet, for all their success, almost every VC I spoke with still lived with FOMO: fear of missing out. Specifically, fear of the founder who got away. “The mistakes of omission are much, much bigger mistakes,” said Andreessen. (4668)

How does venture capital justify taking risks? In large part through literally viewing the greatest risk as not taking (the wrong) one.

I had versions of these mantras repeated to me so often that I believe they’re the governing gravitational forces behind pretty much everything that takes place in the Valley—why it attracts so many strange and disagreeable people, for instance. Venture capital is a unique enterprise for two essential reasons: (4617)

It has a very long time horizon. (4619)

It offers asymmetric odds that reward taking chances on upside risks. (4622)

Are these principles the rest of the world should emulate? The first one is. (4625)

Both are good principles. Supposed contrarians all reciting the same mantras, however useful, should still make you suspicious. Yes, it is good that they have relatively long time horizons and seek big upside potential. And yes, even some of that is huge, even if this is neither as unique nor as pure as they would have you believe, or is often done via a sort of play acting – the symbolic version of the thing.

The result is something hugely (modulo AI existential risk concerns) useful and pro-social and productive, a wonder of the world, something I absolutely love that it exists, that the world needs more of. That is because People Don’t Do Things, especially new things, or take risk, or get the talent pointed at real problems, and all the neat stuff like that.

Nate’s theory of how this works out is a risk story. Books like this are hedgehogs.

So here’s my theory of the secret to Silicon Valley’s success. It marries risk-tolerant VCs like Moritz with risk-ignorant founders like Musk: a perfect pairing of foxes and hedgehogs. The founders may take risks that are in some sense irrational, not because the payoff isn’t there but because of diminishing marginal returns. (4750)

There is certainly something to that. The VCs are making a deal with the founders, where the founder commits to acting irrational in this way, and the way they do that is by credibly claiming they are indeed irrational in one of several ways that would do this – they value the mission enough, are sufficiently deluded about their chances, actually care about making the billions, and so on, or some combination.

Whereas I have a slightly different story. To me: The core activity at the end, of driven and talented people actually trying to Do Things, especially related to tech, is insanely valuable.

Even though everyone is kind of going through motions and playing Keynesian beauty concepts and social games and signing and imitating and so on most of the time, it doesn’t matter.

The process identifies founders and teams capable of Doing Things, and puts them in position where they get to and have to try and Do Things. It works. Even if 90% of the time and money and activity is to get to that point, and insiders capture much of the direct financial gains, the actual payoff makes up for it.

The San Francisco Bay Area—the counties of Alameda, Contra Costa, Marin, Napa, San Mateo, Santa Clara, Solano, Sonoma, and San Francisco, California—was home to 7.76 million people as of the 2020 U.S. Census, or roughly 0.1 percent of the world’s population.[*8] However, as of October 2023, it is headquarters to almost 25 percent of the world’s unicorn companies (defined as private companies with a valuation of at least $1 billion). (4573)

The pie is that big. We still all win.

Yes, they accomplish this in part by using rhetoric and multi-level signaling games and funding preferences to forcibly steal talent and companies from elsewhere, but it all still largely counts.

So three cheers, to all the smoke and mirrors. If that’s what it takes.

(As long as, of course, we ensure the world probably doesn’t end.)

A thing venture capital likes to say is that they are long term oriented, while the stock market is short term oriented.

Furthermore, the value of maintaining a long view may increase in a world where—and here I’m sounding like Thiel—data and analytics can sometimes lead to nearsightedness. It’s often easy to algorithmically optimize for a short-term fix—and much harder to know what will produce long-term brand value. (4632)

The catch is that both of them are:

1. Long term oriented, in the sense of trying to calculate the net present value (NPV) of future cash flows, and make good trades.

2. Short term oriented, in the sense that they extrapolate quickly from short term results to implications for those cash flows, such as expecting growth to beget future growth, the classic ‘hockey stick graph,’ and so on.

3. Partly Keynesian beauty contests. Everyone is trying to predict what the next trader is going to do. Except on Wall Street you can essentially hold to maturity and let the company earn the money if others disagree (mostly), whereas in Venture Capital, unless you are prepared to cover future rounds, the company’s success depends on additional fundraising.

There are frequently claims Wall Street is too invested in the short term. I think this is at least imprecise. What they are doing is not so much ‘neglect the long term’ as it is falling victim to Goodhart’s Law, and focusing too much on the numbers and what they can see, and not enough on intangibles – yes, like long-term brand value creation. Once you have good models, it’s hard to disregard them sufficiently or adjust them enough to account for what in context are intangibles.

Peter Thiel is one of the most interesting people I’ve ever met. There are downsides to be sure, we disagree on quite a lot, but I’d be thrilled to spend more time with him even if I had no agenda and no hope of convincing him to much change his mind.

If you ask someone like that a question, you can never be sure what you’ll get.

When I spoke with Thiel, this was the first question I asked him. “If you simulated the world a thousand times, Peter, how often would you end up in a position roughly like the one you’re in today?” This was intended as something between a softball and a curveball—an unexpected but relatively nonthreatening conversation starter.

Thiel delivered an answer that went on for almost thirty minutes. He began by objecting to my premise. “If the world is deterministic, you’ll end up in the same place every single time. And if it’s not deterministic, you would almost never end in the same place,” he said. (4527)

The question is indeed somewhat ill-formed. Which things are held constant, versus not held constant? When do we begin to roll the dice? If you begin one minute before conception, in an important sense a person probably won’t exist, or will be very different. If you randomize other events, who knows what happens. And so on.

In the common sense hypothetical of this type, where world events happen on schedule unless you in particular alter them but the personal stuff is randomized, I think the answer is he ends up in a similar position remarkably often.

The details will change a lot, and certainly there is a group of people who are proto-Thiels that greatly outnumber actual Thiels, but I don’t think the proto-group is that many zeroes bigger. Not every proto-Thiel becomes a Thiel, but a remarkable percentage do – I think such a person ‘creates their own luck,’ perseveres, ends up in the right places at the right times and finds a way. Not every time, not even a majority of the time, but way more often than you would think. And if you’re not a proto-Thiel? Then it almost never will happen.

That is exactly why the principle in the Valley is, you bet on the founding team.

As a reminder, in this metaphor, foxes know something about many things, and hedgehogs know a lot about one particular thing.

Those clever little foxes were the heroes of The Signal and the Noise. But do foxes make for good founders? I can think of exceptions to the rule (the polymathic Collison is quite fox-like, for instance).

But in general, VCs are looking for people with one big crazy idea that they’ll commit to for a decade or longer. (4693)

I think this is actually backwards. Silicon Valley looks for founders willing to commit to one big crazy idea for a decade.

That one big crazy idea is ‘found a venture capital backed Silicon Valley startup.’

They bet on the man (and yes it is usually a man), not the cards.

Everything along the way, until rather far along, is mostly evaluating the team.

If your first idea does not work, you shift it around. Or you fully pivot. That’s fine.

Indeed, there is a reason why Y-Combinator does not even require an idea for a company. The idea to found a company at all is the idea that matters most. Ideas are cheap without execution. Execution is everything. That’s how all this works.

To be a CEO, you need to be a fox of the first rate. You need to be able to raise capital, to lead and manage, to hire and fire, to negotiate, to determine strategy, to dealing with mistakes and crises one after another, to market and understand the customer, to balance signaling to investors with signaling to customers with signaling internally with building and shipping a product. Your realm is anything and everything. You have to give up any semblance of a life for years on end.

And you have to stay sane (enough) throughout, as most people are lying to you about how the world works or your business works in various ways, or have no idea about either while claiming otherwise, while you are driven crazy exactly by whichever element here is your weakest link.

Founder CEOs who are playing the game (relatively) straight up earn every penny.

What makes a good founder? What makes Silicon Valley eager to bet on you?

Nate Silver asks an important part of this question.

Are VCs Intentionally Selecting for Crazy Asshole Founders? (4918)

Yes.

I mean, sure, #NotAllVCs, and there are limits to how far they want you to go.

But yes, they absolutely will favor you if you are an asshole, because they think that other VCs will too, and because they think being an asshole is vital to success – you’re going to have to act like one on various occasions to Get Things Done, and you’ll need to credibly signal that you can act like one as a threat.

And yes, they absolutely want you to be the right kind and amount of crazy, because they are looking for you to do things that you would not otherwise be wise to do, but which help them and increase chances of success, including having absolutely no work-life balance whatsoever for years on end, and also again because they know the next level up will be looking for that crazy.

A lot of this is pure pattern matching. They’ve seen crazy assholes succeed, they know others have seen it too, so they want to fund more crazy assholes.

There’s more to it. And there are plenty of other considerations too, being a crazy asshole is only a small part of what you need to do. But there is no mystery here.

What are other signs and traits of a potentially successful founder you want to bet on?

Another thing you want is that they are self-made. Coming from money won’t work, or at least that’s conventional wisdom. You wouldn’t be hungry enough. Indeed, America today is highly unusual for how many of our richest people did not inherent.

Of the billionaires on the 2023 Forbes 400 list—the four hundred richest people in the United States—70 percent are basically[*19] self-made. And 59 percent came from an upper-middle-class background or below. (4919)

In 1982, only 40 percent of the Forbes 400 had started their own business; the majority were simply scions of inherited wealth. (4924)

Let’s say you inherit a $25 million trust fund on your eighteenth birthday. Are you going to start a business with it? Maybe you should. But it’s much easier to withdraw $1 million a year to live off, travel the world and have some wild parties, and put the rest in S&P 500 index funds earning 7 percent a year. (4929)

Even if you felt you could count on the world to ‘stay normal’ and the S&P to return 7% a year forever (5% after your peaceful 2% inflation target), what fun would that be?

A lot of people would answer quite a lot of fun. And, well, fair. But it’s not for everyone, at least not for all that long. Purpose is important.

Perhaps the problem is that the very rich are not brought up to see it that way, and that they don’t develop various skills that are vital to the path to founder success.

“Why are second-generation kids never that successful?” asked Social Capital CEO Palihapitiya, who moved with his family from Sri Lanka to Canada and worked at a Burger King to help support them. (4941)

It can also help to have something else: a chip on your shoulder. Josh Wolfe, of Lux Capital, is fond of the phrase “chips on shoulders put chips in pockets.” Feeling left out, excluded, or estranged can make you extremely competitive. Remember, VCs want founders who are willing to commit to low-probability ideas—ideas they think the rest of the world is wrong about—for a decade or more.

What motivates a person to do something like that? Wolfe, who grew up in a single-parent home in New York’s gritty Coney Island neighborhood, told me he thinks there’s a common answer: revenge. (4946)

You’ll show them by becoming a billionaire? Well, maybe. At least sometimes. Perhaps I simply didn’t want revenge enough. I like to think that’s not the main motivation, that you can want to do something big for the world, something you care about, or want to make a lot of money for various reasons.

“Who is my real customer? It’s a young, disenfranchised, disenchanted entrepreneur-to-be,” said Palihapitiya, referring to the sorts of people who might come to him for investment or mentorship.

“And I use those words specifically, because if you’re comfortable, and you’re happy,” he continued, “you’re not the kind of person I want to work with anyway because you’re probably not gonna be successful.”

This feels like a dangerous game. Successful founders may be disagreeable on average, because disagreeability is correlated with competitiveness and independent-mindedness. But the disagreeability is still a bug, not a feature.

If you start to select founders because they’re disagreeable, you may get the wrong ones. Especially if founders deliberately play into stereotypes that they think VCs will like, as Sam Bankman-Fried did (we’ll cover him in the next chapter). (5010)

Any time you are selecting for anything, you are selecting for the people able and willing to fake it. That’s always a problem. But perhaps also an opportunity?

A lot of the reason Silicon Valley works is that everything is a test, preparing you for future tests, and eventually, one hopes, the creation of actual value.

If SBF or someone else can play into the stereotypes that VCs like, the naive reaction would be that this is terrible. You’re falling for a fake. Except, what if what you are testing for is the combination of:

1. Figure out what the teacher’s (here VC’s) password is.

2. Be able and willing to give the teacher their password back.

If you can do that, then those are highly valuable skills. You can figure out what people want to see and hear, and you can provide that for them, and you’re not going to have any objections to doing that. Sounds like someone to invest in, no?

Yes, part of the problem with this is you end up with a lot of frauds. But, as Matt Levine and others remind us, the optimal amount of fraud is not zero. You want a founder who is not too worried about whether they might technically be committing fraud, or about they can cash the proverial checks they are metaphorically writing. You want someone who does what they have to do, not someone scrupulously honest.

Also, if you fake being someone, you start to act like them naturally, and you do things like them to keep up the facade. Almost as good.

Occasionally that all means someone steals all the customer deposits along with your investment capital. That is a known risk. But so what? Small price to pay.

What about the traits themselves?

It makes sense that, on average, you don’t want a happy entrepreneur, because they won’t be willing to put in the insane hours and effort and sacrifice, or take the risks in various senses, potentially including both moral and legal risks.

Do not get me wrong. Silicon Valley founders take metric tons of risk.

The catch is, they mostly do it by ensuring no one tells them the odds.

Elon Musk is an outlier, but less of one than you might think.

Thiel had once considered writing a book about Musk and their PayPal days. “The working title I had for it was Risky Business, like the eighties movie. And then the chapter on Elon,” he said, “was ‘The Man Who Knew Nothing About Risk.’ ” And yet, when it comes to risk, it’s Thiel who is more of the outlier in Silicon Valley. “Peter is not a risk-taker. There is nothing. He is a guy wired to protect his downside,” said Moritz.

Although that’s nothing compared to some others:

For instance, SBF was quite specifically insistent that people ought to be willing to risk having their lives end in ruin. There’s an idea, he said when I spoke with him in January 2022, that “the bigger you are, the more risk you can take without endangering what you have.” (6024)

I do think that even if you could have otherwise played it safe and been personally set, you need to be willing to risk ruin if the stakes are high enough, in the sense of being willing to make enemies, or perhaps face political or legal consequences or even potentially violence. But risking financial ruin in particular, through losing money, past some point? You’re simply never getting odds to do that.

Silicon Valley tells founders several things at once.

1. You will probably fail.

2. You have to believe you will succeed, or we won’t fund you.

3. You have to swear to everyone you will succeed, or we won’t fund you.

4. When you fail, if you do it the right way, we won’t hold it against you.

Venture capitalists do not want founders who understand how to take carefully calculated risks and make precise Bayesian estimates.

Venture capitalists want founders who will risk it all to try and hit is big, whether or not that makes any financial sense for the founder, or the odds are any good. Or whether or not it is even +EV for the company, because VCs think in terms of those huge wins.

They very much want you thinking like Musk here rather than Thiel:

But there’s also something more. The most successful founders like Musk succeeded despite what were ostensibly extremely long odds. Thiel, recalling the challenges Musk overcame to build SpaceX, thought that no one obstacle Musk faced made for an insurmountable barrier. But the quants had run the numbers—if you have 10 hurdles to leap over, and you have a 50 percent chance of tripping over each one, the odds of making it to the end of the course are 1 over 210, or just one chance in 1,024—and concluded that the venture was imprudent.

Musk had thought differently. “He was determined to make them happen,” Thiel said. “It was a matter of assembling the pieces and putting them together, and then it would work. We’re just in this strange world where no one does it because they all think probabilistically.” (4560)

That’s… not what I think of quants running numbers, but yes, if you think there are a bunch of uncorrelated medium-difficulty hard steps, all of which must succeed, the odds are very much against you.

(Similar arguments are often used against pretty much anything ever happening. For example, the classic anti-existential-risk-from-AI argument that demands you lay out the details of one particular scenario with all of its steps, then says ‘this only happens if all 10 steps happen, and each is not that likely’ or arguing against one particular step, ignoring that the argument does not rely on those particular steps or path. And the reason something like SpaceX still works is in part that they fail and they figure out how to fix or work around the problem, repeat as needed.)

Of course, Thiel’s calculation was obviously wrong. People who found companies like Tesla (also see PayPal) are almost never going to do something where the outside view is actually 1024:1 odds against success. No, Elon’s odds probably didn’t make any sense either, but any reasonable reference class wouldn’t have counted Elon out. Notice that other space startups like Blue Origin might not be doing as well as SpaceX, but the odds certainly aren’t 1% or less.

VC also wants you not to worry about the downside, and not to be tempted to try to take some gains off the table, no matter how insane it would be not to do that.

It shows.

I’ve played in a handful of high-stakes poker games against rich guys—including one game frequented by venture capitalists, Silicon Valley founders, and the hosts of the All-In podcast, who are friendly with Musk. I was sworn to secrecy about the particulars, although since one of the All-In hosts, Jason Calacanis, said this publicly, I can confirm that the first time I played in the game, I won enough money to buy a Tesla.

The other thing I can say—just in general terms—is that the higher the stakes, the crazier the action. The biggest poker games select for the players who want to take crazy, irrational, −EV risks.

Maybe not literally going all-in every hand—although I’ve seen strategies that aren’t far from it—but embracing variance, as the “Techno-Optimist Manifesto” would say. (4505)

Embedded deep in the VC psychology is the idea that they can sleep soundly at night because they are not only enriching themselves, but also making the world a better place. For some recent technological developments, however, the value proposition has been more questionable.

Social media may well have had net-negative effects on society. Crypto gave rise to a lot of scams and cons, like Sam Bankman-Fried’s FTX—heavily invested in by Sequoia and other VC firms—which cost cryptocurrency holders out of at least $10 billion. And with AI, the disruption could be profound, resulting in mass reshuffling of the economy even if it remains relatively well aligned with human values. (4656)

What a nice term for a degen making -EV risks, ‘embracing variance.’ They’re not embracing variance. They’re gambling, and not the advantage betting, trying to win variety either. The bet it all on black kind.

So is this my way of saying that the richest founders in the world are just degenerate gamblers who got lucky? No, I’m not saying that. I think they’re highly skilled degenerate gamblers who got lucky. (4996)

Almost by definition, the people at the very top of any leaderboard are both lucky and good. (5002)

Indeed, that is all a fair way to describe the ‘Techo-Optimist Manifesto.’ Rather than argue that benefits exceed costs and the risks are worth taking, the manifesto is a series of unqualified assertions of blind faith and absolutes. Technology is always good, humanity always benefits, there is zero danger, and so on. It is its own strawman, a proud announcement of being blind to downside risks.

Which is exactly the venture capitalist way. When you invest in a company, when you try out a new tech, what could go wrong? Losing all your money is, in their view, only a small mistake. So that investment went to zero, so what. It can’t go below zero. It shows you were bold. The real mistake would be missing out on the next big hit.

The VC mind, by default, assumes that the downside must be limited. That mind lives in a world without short selling and with limited liability. It simply cannot fathom the idea that their investment could have a negative left tail even bigger than the positive right tail. Marc Andreessen (the author of the manifesto) seems to toy with outright denying the idea that any such investment could individually harmful on net at all.

The second trait—the asymmetric nature of payoffs in Silicon Valley—is even more essential to understanding its mindset. But its implications are more ambiguous than the first one. (4647)

Thus, Marc Andreessen says just do go ahead, whatever it is, however risky in any sense, including everything from AI catastrophic risks to being another crypto fraud. His politics and stands on regulations, and his justifications for them, not only ignore but deny and laugh at tail risk. Simultaneously it argues that this time will be the same as our recent ahistorically calm history so there is nothing to worry about, and also he forgets his history. And he is not alone.

That leads you to do things like this, after Adam Neumann famously built up WeWork only to have it collapse:

It was [Adam] Neumann, as you’ve probably guessed. In a room full of Silicon Valley’s elites, a16z were showing him off—and sending a message. “Why would they run out and give a bunch of money to Adam Neumann after everything they’ve seen? Like, what in the world was that all about?” said Benchmark’s Gurley. “If I were asked to analyze what they were doing, they wanted to send a signal to everyone.”

The signal was that they didn’t care about reliability—they wanted founders who gave them upside risk. They were embracing variance. “If they’re that type of person, they’re open for business, the door’s wide open, and we’re willing to talk to you, no matter what.”

The signal mattered. A16z seems determined to shout from the rooftops that they are willing to be as reckless as possible, in all senses, and to pour good money both after bad and also the extremely risky, and to not worry about cost or downside, as a business profile.

Mostly, however, I think this was simply a good bet. They want to be in business with Adam Neumann because they absolutely have odds betting on Adam Neumann. That was a highly impressive ‘failure.’ Adam Neumann won, in that he proved he could build up something big, get people to believe in a crazy business plan, and was smart enough to take a bunch off the table before it all fell apart. You don’t get that far without real skills, and there was a substantial chance WeWork could have succeeded. It does not take that big a probability of success before this bet was +EV.

I don’t know the terms of the investment, so it is certainly possible they overpaid and didn’t have odds. What I do know is that the world wants Adam Neumann to go out and try to Do Things again, to build another empire. That there is a price where Neumann should be happy to do it, and investors should be happy to help him.

I’ve already gestured at it a few times, but the core way Silicon Valley makes funding decisions is a Keynesian beauty contest, counting on it being grounded enough to get to a good equilibrium.

The process is akin to what economists call a Keynesian beauty contest. (5107)

In particular, it is an iterated such contest. You are trying to predict what others will do both now and into the future. If you select for that, you win. If not, you lose.

You could get around this by being willing to fund the entire company, at prices that enable the business to prosper, without needing the social proof of other investors. Unfortunately, the VCs of Silicon Valley, even when they can very much afford the required funding, are psychologically or socially incapable of this. Hence the contest.

Is Silicon Valley Really as Contrarian as It Claims? (5083)

This looks contrarian because the others are the other VCs in Silicon Valley, and especially the larger ones that will be required for later rounds. The question is, contrary to who? What does it all mean?

If you are contrarian (in the good sense) because you are thinking for yourself, then that is indeed good. If you are instead contrarian because you are thinking what those around you are thinking, and trying to ponder what they ponder, then you are a conformist all the same.

Do VCs actually want to be contrarian?

Founders Fund—Thiel’s firm—has a reputation for it. But according to Rabois, there’s a delicate balance to strike. “When you initially invest, or found a company, you want it to be, like, ridiculous and contrarian,” he said. “But you want it to shift into consensus, because you need other people’s money, publicly or privately.

You need to recruit employees that are more normal than the founder-type class.

So the art is pulling a trigger and then shifting it. And if that gap is too long, then you have a problem.” (5097)

You need there to be some element in there that differentiates the founding team and its ideas, allowing them to potentially win big. You get, and must spend, a few weirdness points.

In most other ways you actually need the team to be checking off the same boxes as everyone else around them. And indeed, it is vital that you be pondering what those around you are pondering, whether or not anything involved makes sense.

So Rabois is always trying to calibrate to his friends’ preferences. “One thing that’s kind of a secret in the industry,” he told me, “is that most of the people I compete with, I’m actually pretty somewhere between real friends and very good friends with. So part of what I’m doing is mapping their brain. Like, will they or their fund appreciate this? Are they going to see the signals I’m seeing?” (5114)

Sebastian Mallaby, the author of an excellent book about Silicon Valley called The Power Law, thinks that in certain respects it is an exceptionally conformist place. “In some ways, venture capitalists are the ultimate herders,” he said. “You go to Sand Hill Road, and you see that they all have offices on the same road. And there’s kind of one good restaurant on that road, at the Rosewood hotel, so they all bump into each other at the same bar.

He maps their brains, they map his. Everyone is trying to get in exactly the right amount of sync. You do want to find overlooked opportunities, but what makes them overlooked opportunities is that they won’t stay overlooked for long.

Yet for all its Mean Girls conformity—Rabois trying to figure out what his five best frenemies will think—Silicon Valley’s opinions are still relatively uncorrelated with those of the outside world, Mallaby thinks. (5125)

I do believe that on many important margins. Of course there is overall a huge correlation with the outside world, but there are many key differences, and the VCs map much closer to each other than to the outside consensus.

I even heard the argument that as the Village becomes more conformist, Silicon Valley has more opportunity to profit by running in a different direction. (5131)

It potentially does grant an opportunity. You can pick off those who don’t want to conform to the Village, either because they are non-conformist in general or because they don’t want to conform to the Village in particular. Or you can take the opportunity to move your SV consensus somewhere else.

One part a Keynesian Beauty Contest is that by default it will be absolutely brutal with enforcement of stereotypes. Convincing others you have what it takes is, itself, what it takes. If I think everyone else thinks you don’t have what it takes, then that means you indeed do not have what it takes.

Which means that correlation is king. Any visible sign that anti-correlates with success is going to haunt you every step of the way, whether or not it is causal.

If that sounds like a recipe for de facto racism and sexism, whether or not anyone involved is actually either of those things? Well, you’re right.

The investor, like many others, turned her down. “It’s amazing, I love it, blah blah blah, but I’m not going to invest,” she recalls him saying. “Here’s why. As a Black woman you’re going to have a harder time fundraising, a harder time retaining talent, a harder time selling,” the investor said. “Every part of this process is going to be harder for you, and it’s already an impossible process.” (5148)

What makes you a top successful VC?

It helps to already be one.

Is VC Success a Self-Fulfilling Prophecy? Andreessen Horowitz is a lot like Harvard. And Founders Fund is a lot like Stanford. Marc Andreessen and Peter Thiel would probably resist the comparison, between Andreessen’s dislike for the Village and Thiel’s skepticism of postsecondary education.[*27] But the similarities are obvious once you see them. The top venture capital firms—like the top universities—are exceptionally sticky institutions. (5179)

What links top VC firms and top universities is that they are recruitment-driven businesses. And recruitment can become a renewable resource. (5194)

The secret is deal flow.

The best deals in Silicon Valley are obvious.

This is in some sense inevitable in a Keynesian Beauty Contest. If you can be confident everyone else is going to like the deal, then everyone else can figure this out too, and already likes the deal.

The trick is that the price of the deal does not reflect expected future value, or the place where supply and demand would currently intersect. You might think it would do this, but you would be wrong.

Instead, VC in SV has set up a system whereby the best founders are warned not to use their market power to drive up the price, and also where borderline founders mostly don’t get to offer sufficiently large discounts to make deals happen.

A given startup is ‘supposed to’ trade at a given price, based on what stage of development they are in, certain ‘fundamentals’ and various heuristics. And yes, there is room for ‘how high does the founder dare go,’ but there are severe limits to that.

The key weapon is fear of the ‘down round.’ If a startup trades now for a valuation of X$, then in the future trades for Y$, you had better hope that Y>X, or at minimum Y=X, and they will use various tricks (‘structuring’) to try and make the headline number not go down if at all possible. A down round is very close to death. So startups want to price, they are told, sufficiently low that they won’t run into this problem.

This is of course economically nonsense. If I have the New Hotness, I am very likely to get to the next funding round in a place where I have a company forth funding. In some cases, I will succeed wildly. If today’s round is priced fairly, that means there will be plenty of times the company does okay, and is still a good investment, but the expected value of the company has gone down. There is nothing wrong with that. Yet the entire ethos sets up to basically kill you if that scenario happens. In particular, if the lead investor uses this as a reason to pull out, that too gets you into deep trouble.

The VC solution, which is strangely good for their business model, is that you should price such that if the business remains viable the price will definitely go up.

This is of course rather silly. Either those middle-scenario viable scenarios are worth worrying about, in which case a little decline in EV shouldn’t stop anyone from taking a +EV wager going forward. Or they are very little of the value, in which case founders should not care that things go badly in those cases. Even if you buy the whole argument, you can’t have it both ways.

The other weapon the VCs use to persuade is that if the startup raises Too Much Money (TMM) then they will expand too quickly, take on too much burn, pay too high salaries because they lose negotiating leverage, and so on. Having done that too early, they will fail. The solution of course is to Simply Not Do That. If you get Series B pricing on your Series A round, you can save the extra equity and put the surplus in a lock box, even if yes as Frog says you can open the box. In theory you could even take some of that surplus off the table for yourself, and still be ahead on equity, why not (the answer to why not is ‘because people would disapprove’)?

These clearly +EV plays being frowned upon is a sign you are not fully in The River.

The third argument, which has merit, is that if things are going this well, you don’t need to negotiate hard on price in order to be in great position to succeed. You can afford to give a discount to ‘market’ price in order to choose the VC partners you want to have around and to signal by involving, buy their goodwill and help and future inclination to fund, and save time so you can go build the business. That’s fair, but only up to a point.

That’s why top VCs can say things like ‘the price you enter at does not matter’ or otherwise not care much about price, and get away with it. The markets involved are not allowed to clear. And if a VC offers things as part of their package, that has big advantages in winning deals over ‘pay more money that is more valuable.’

“Basically ninety percent of the fight is over before it begins in venture,” Andreessen told me of the process of recruiting founders. “Like, at the point of contact with an entrepreneur, when we go in and we do our whole sales process and try to win the deal…ninety percent of that fight is over before it begins, because it has to do with the reputation that we’ve established, the track record—you know, the brand.” (5200)

“It’s sort of a self-fulfilling prophecy, the whole thing,” he said. (5205)

So the top VCs by reputation, by connection to opportunity, get the best deal flow. When they want the deal, they get it, and they mostly get the size they want. And those top deals are systematically way underpriced. Everyone else has to pick over the remainder, and deal with adverse selection.

Ironically, I know people who have turned a16z in particular down because of aspects of their reputation. But that only reinforces Marc’s point here, that the battle to fund a given great company is 90% over before it begins.

Does a16z do a good job of selecting the great companies and deals? Does Founders Fund? What about other top VCs?

I cannot know for sure from where I sit, but based on what I do know, I think no. I do not think they are especially good at picking winners, or at supporting winners. Indeed, based on various investment choices of theirs especially in crypto, I think a16z does a below average job of picking great companies.

But that does not ultimately matter anywhere near as much as deal flow. So they win.

I still say that my best guess is: More than all of a16z’s Alpha is in their deal flow.

How much Alpha is there? A lot.

Andreessen recited from memory data on a typical returns portfolio, which he later confirmed by email: 25 percent of investments make zero return. 25 percent produce a return greater than 0 but less than 1x. 25 percent produce a return between 1x and 3x. The next 15 percent produce a return between 3x and 10x. Finally, the top 10 percent produce a return of 10x or greater. (5217)

That profile is insanely great if you can get it. If you put everything at the midpoint of the range and the top 10% at 20x, that’s a 260% profit, without even accounting for the true smash hits that traditionally are the key profit center. Who wouldn’t want to invest in that?

Although it is worth noticing that a16z will often put vast amounts (nine figures) into relatively mature companies, where the chances of those 20x+ returns are lower.

Isn’t it grand that founders aren’t seeing that and demanding higher prices?

Again, this all only works with top deal flow.

It’s important to clarify that this data pertains only to what Andreessen calls “top-decile” venture firms like a16z—not just anybody renting a room in an office park on Sand Hill Road. The industry as a whole does not generate particularly attractive returns. But the top firms can be extremely profitable, targeting a 20 percent IRR and going north from there, perhaps even 25 or 30 percent or higher for high-risk sectors. (5223)

That return number very much lines up with Andreessen’s distribution.

Note that it does not imply that it would be worthwhile to raise more capital. If your advantage is deal flow, you are already taking full advantage of all the best deals you can find. Adding more capital to a16z would force it to, effectively, fight through the same adverse selection problem as everyone else, taking deals a16z would have otherwise already passed on. It can both be true that a16z returns 20% IRR, and that raising another fund for what remains would not be creating a good investment.

Throughout this book, I’ve dealt with some controversial cases as we conduct our census of the River. Is Donald Trump a Riverian? No, he’s not analytical enough, despite his history in the casino business. Elon Musk? Yes. (5416)

I continue to think Trump is not a controversial question. If you still have doubts, you could watch or think back to his debate against Kamala Harris. None of the key attributes are there. Elon Musk, along with Silicon Valley in general, is a more interesting case. I want to give him and SV in general partial credit, but not full credit.

I am willing to say that Silicon Valley is, at least, River adjacent. That they share some of the River nature, and its virtues. They think of themselves as kindred spirits, and that makes it at least partly true. They do celebrate accomplishing big things, doing that which works and figuring things out, and the taking of risk. Most of all, they want other things too, but they very much want you to go out and Do Things, build something people want, and ship.

As long as it’s the right kind of risk and right kinds of things to ship, aimed at the pattern matched big things. There is a lot of real underlying it all, and also more than a little cargo culting going on. Even though it is indeed the world’s best cargo, and thus perhaps the world’s finest cult. But it is a reality tunnel, an echo chamber, a realm of the zeitgeist and the vibe, and within its own tunnel it is deeply conformist to its tales of exactly how to be contrarian.

Which, again, is a great thing to which to conform. If you have to choose, it’s hard to do better, and very easy to do far worse.

All of that merely scratches the surface. Silicon Valley is not Nate’s world, and the shift from things Nate knows well to places he is exploring is clear. He got great sources, and those sources are in some ways highly honest and forthcoming. In other ways, they are very good at the hoodwink.

Another interesting note is that in the final chapters, Nate talks with Sam Altman and Paul Graham, as well as a bunch of SV-adjacent others. It’s not clear how much those conversations impacted the Valley chapters, but if I wanted to talk to one person to truly understand SV, Paul Graham would be very high on that list. And it’s interesting how little it feels like Graham’s perspective got mixed into this.

Is all of this meta-rationality? Is it the River taken to its logical conclusion, that you should take the calculated risk of taking uncalculated risks according to a set of heuristics that has proven to be successful?

Thus, shouldn’t we say that you absolutely should (from a fully amoral perspective) be betting on and investing in Adam Neumann and Sam Bankman-Fried, telling them to be even crazier, and being happy you are getting good odds on a huge payoff?

One could plausibly argue that. I mostly buy it. It’s not the same as the River nature.

At this point, we transition from Silicon Valley to Crypto.

I am far more interested in the true nature of Silicon Valley and Venture Capital than in Crypto. SV and VC are vital to our future. Crypto… likes to claim it is the future. I find that claim rather doubtful, increasingly in a ‘I don’t think about you at all’ kind of sense.

Still, there’s some good stuff that one cannot ignore.

If you want my take on What Happened with SBF, read Book Review: Going Infinite.

I will not be explaining again. Nor will I cover the crypto basics, IYKYK and if you don’t then there are plenty of other places to learn if you’re curious.

I understand why Nate Silver focused so much on SBF before SBF blew himself up, and why he kept that focus after SBF blew up, given his level of access. It’s a great story, and I do think it is a story that matters. I still wish that it wasn’t so centrally used as the pathway into Effective Altruism.

I do think EAs themselves need to heed the cautionary tale here, but SBF represents an extreme tail risk failure mode on many levels, and has been used to attempt to condemn anyone and any concept remotely related to EA. This has plausibly substantially decreased our civilization’s ability to sanely respond to AI. Presenting him this way to a casual audience unfamiliar with even the basics gives me a lot of ‘no one picks on my brother but me’ vibes.

Nate Silver did have the benefit of doing direct interviews with SBF, so he has some additional puzzle pieces, or at least fun anecdotes. Here are ones I loved the most, starting with another very good opener.

The room was getting darker, the power was going out on Sam Bankman-Fried’s laptop, and he was telling me increasingly unhinged things. (5399)

That sounds about right. My favorite part of this is the power going out on the laptop. What could be a better metaphor for failed risk management? Actually needing your laptop charged, and running out of charge, while sitting in your work area, really should be a Can’t Happen. There are any number of ways to plug in your computer.

Nate gives us more evidence that SBF ended the story rather full of delusions of hope, and that he had indeed fooled himself throughout the process.

The specific number he would later give me was 35 percent—a 35 percent chance that he’d emerge from the situation with something that would be “considered a win” by knowledgeable people. “I know a lot of people will say I’m crazy for giving this number. It’s an insanely high number. And, like, it can’t possibly be right,” he had qualified. (5444)

If you had the proper River Nature, and you noticed your number could not possibly be right because it was insanely high, you know what you would do? You would revise the number downward, at least until this was no longer obvious to you. Disagreeing with ‘a lot of people’ is fine, but ‘can’t possibly be right’ very much is not.

SBF learned to fake adjusting his priors for such situations when necessary – I can testify to that first hand. But SBF had no interest in doing this for real. About anything. Ever.

SBF’s continued embrace of the ‘it was all tactical errors’ argument is even sillier.

The way SBF framed things to me, these were forgivable, tactical errors—like a poker player playing a hand suboptimally in a challenging situation. “You put those together and it went from, like, significant but manageable to significant and not manageable,” he said. “I just sort of lost track of it,” he told me at another point.

In reality, of course, any of these would have been mission-critical mistakes, let alone all three put together. It was like a pilot saying, “Oh, the plane wouldn’t have crashed if only I hadn’t drunk three bottles of whiskey, punched out my copilot, and then told air traffic control to fuck off when they said the runway was closed.” It’s also clear that Sam’s story was almost entirely bullshit. (7032)

SBF kept saying this to Michael Lewis. Here he is saying it to Nate Silver. The difference is that Lewis treated it as a non-absurd statement, and Silver didn’t.

Nate notes that Sam had several different modes of talk. I saw that too. When you could get Talking-Shop Sam, as Tyler Cowen did in their conversation and Matt Levine, Joe Weisenthal and Tracy Alloway did on Odd Lots, it was great stuff.

Talking-Shop Sam spoke rapidly in run-on sentences, full of jargon about arbitrage and expected value. I liked Talking-Shop Sam. We spoke the same language and had good camaraderie; I could imagine us being friends. (5464)

Then there was No-Shits-Left-to-Give Sam—a more candid and seemingly more honest personality, a Sam who prefaced his responses by implying that now he was going to give you the real answer. No-Shits-Left-to-Give Sam could get dark and erratic—this was the version that I was a tiny bit worried might suddenly lunge for my digital recorder. Finally, there was Sneaky Sam. This iteration of SBF was pedantic and lawyerly. (5466)

NSLTG Sam is fun too, especially when he gets dark.

Sneaky Sam is fun in his own way, too. Why not play some poker with him?

“Um, what do you think the odds are that one year from now, FTX will be an operating platform? And what do you think the odds are five years from now that customers will have been made whole?” “To the first question, I’d say under ten percent. To the second question, thirty percent. I have no idea.” Oops! I had fallen for Sneaky Sam’s trap. (5483)

As it turns out, these were great estimates. Ten percent seems high, but there was model error and unknown unknowns to consider, it was not so crazy (at the ~10% level) that someone might see sufficient brand value, especially if they’d first bought up a ton of claims. And it’s hard to fault the 30% given it actually happened purely because Number Go Up once more – between Anthropic and Bitcoin, 30% seems high but not crazy high for ‘enough profits are made that FTX becomes solvent again, as measured by USD liabilities at time of bankruptcy, before paying the IRS and other fines.’

It was only possible because the liabilities were converted to USD, while the assets were not, and the assets were highly volatile. And yes, this implies FTX debt was trading super cheap for a while. The Efficient Market Hypothesis is false.

For those who think it is hard to spot the crypto frauds, it’s not all that hard:

I wondered about skilled crypto trading. What differentiated the more successful traders from the less successful ones? (5562)

[Mashinsky, head of Celsius’s] reply was darkly cynical, although served with a twist of humor, because Mashinsky tended to laugh at himself when he spoke. “Poker is about skill. This is not about skill. This is just about getting on the bus. (5563)

[We] lend it out to FTX. FTX needs liquidity,” he answered instantly. “You can give it [directly] to FTX. Are they gonna pay anything? No. You deposit through me, I’ll squeeze them to give me three, four, or five times more than they’ll pay you. Because I have a million and a half people that are all together, marching as one.”

“That makes sense,” I said, not sure it made any sense at all. “It’s very simple,” Mashinsky replied. (5574)

If that’s what the guy loaning out money to other crypto people at 18% says?

Yeah, he might be a fraud. And you almost know FTX is a fraud as well.

The whole story makes no sense. If a crypto exchange is borrowing money at 18%, effectively using a credit card, what, you think it is solvent and has all the customer deposits?

Nate agrees with me that, in at least some sense, SBF was in the end sincere.

Despite Bankman-Fried being an unreliable narrator, my best guess from talking to him and many other sources is that his interest in effective altruism was at least partly sincere. (6190)

Crypto is many things. To many, it is only one of those things, which is gambling.

Here’s something I learned when writing this book: if you have a gambling problem, then somebody is going to come up with some product that touches your probabilistic funny bones. (5794)

And whichever product most appeals to your inner degen will be algorithmically tailored to reduce friction and get you to gamble even more. (5798)

In the future this will only get more extreme.

Nate draws distinctions between the characters and ethos of different blockchains. Some of it he links to different origin stories, some to distinct functionality. And some to the fact that Bitcoin’s core value proposition is as the ultimate Shilling Point: A form of originalism and uniqueness, taken to the level of religious fervor.

Some of this is because of the origins of the respective blockchains and the cultural baggage they carry with them: Ethereum was a quasi–Silicon Valley startup, while Bitcoin was a cyberlibertarian alternative to fiat currency. “Ethereum is an ecosystem for venture capitalists to make bets on the future of computing and, like, Web3 and DeFi and gaming and NFTs and all that shit,” said Levine. “Bitcoin is a place to make bets on future institutional adoption of an economic asset class.”

But that doesn’t fully explain the fervently anti-Ethereum attitudes that Buterin encountered. He likened Bitcoin maximalists—who often proclaim that Bitcoin is the only worthwhile cryptocurrency—to religious adherents. “Bitcoin is not really a technology project. Bitcoin is a kind of political, cultural, and religious project where the technology is a necessary evil,” he said. It’s common in the River, a highly secular place, to insult someone’s movement by comparing it to a religion. (For instance, you might say that wokeness is a religion or that effective altruism is one.) (5888)

To repeat, famous works of art don’t become famous for completely arbitrary reasons. Focal points usually have something going for them. But the traditional art world almost goes out of its way to intensify focal points by curating a sense of exclusivity and scarcity. (5939)

Statistically speaking, yes, as art gets more expensive it gets more likely to be high Quality in an abstract sense, but from the outside view, past a certain point it is mostly a social manipulation, bragging rights, one-upmanship, tax evasion and money laundering game.

In other news, have you heard about CryptoPunks?

Do you have a friend who owns a CryptoPunk? I have a few of them, and what I can tell you is it’s very likely that they’re going to show you their CryptoPunk when the opportunity arises. In some sense, that’s the whole point of owning one. (5957)

Crypto took up a lot of the book, but that’s all I have to say about it here, so this was a relatively short section. It’s too juicy for Nate to pass on, but it’s mostly not too juicy for me to pass on.

What remains is the ‘relevant to my current interests’ sections. It’s rationalism, EA, AI and AI existential risk, which we’ll cover in the final post.

Previously: Book Review: On the Edge: The Fundamentals

As I said in the Introduction, I loved this part of the book. Let’s get to it.

When people talk about game theory, they mostly talk solving for the equilibrium, and how to play your best game or strategy (there need not be a formal game) against adversaries who are doing the same.

I think of game theory like Frank Sinatra thinks of New York City: “If I can make it there, I’ll make it anywhere.” If you can compete against people performing at their best, you’re going to be a winner in almost any game you play. But if you build a strategy around exploiting inferior competition, it’s unlikely to be a winning approach outside of a specific, narrow setting. What plays well in Peoria doesn’t necessarily play well in New York. (841)

If you build a strategy around beating a specific type of inferior competition, and you then face different competition styles, you either adapt or you are toast.

If you build a strategy around identifying and dominating inferior competition in general, then you might be a very good poker player.

Especially if you also are prepared to fall back into game theory when needed.

(See for example my experience with Simplified Poker.)

One thing I have always believed is that true randomization is often highly overrated.

Poker—yes, there is a solution to poker, though as you’ll soon see it’s an exceptionally complicated one—involves lots of randomizing. Randomizing between calls and raises, between calls and folds, or sometimes between all three. It’s not just that you should play different hands in different ways; you should play the same hand in different ways. (949)

Yes, the pure game theory solution involves heavy randomization.

In practice, however, true randomization is only necessary if being non-random would be detected by the opponent in an exploitable way, and they would exploit it at least as much as you could exploit them.

Thus, if your decisions are effectively, from the opponents’ perspective, sufficiently random or unpredictable already, you do not need to then explicitly randomize further. One way for this to happen is if your hidden information influences your model of what they are likely going to do, because it changes what information they might have had in turn. Another is if they simply do not yet have enough data to anticipate your actions.

You always do have the option to fall back upon the game theory in a pinch. And indeed, if I was playing Nate Silver for sufficiently high stakes, that is exactly what I would do, to the extent I knew what the game theory solutions were.

If you don’t know the odds and solution, you’ll need to approximate.

Long before the advent of personal computers, Brunson and another Hall of Fame player, Amarillo Slim, would deal out thousands of poker hands to themselves to get a more precise sense for the probabilities. (659)

It’s fascinating to think that ‘deal out thousands of hands of 22 vs. AK’ is faster than doing the math on it, if you don’t have a computer handy. My first instinct was ‘of course I would simply Do the Math’ but then I thought about how I would actually go about doing that, and now I realize I’m not so sure how to be more precise (with only the technology available at the time to Brunson) than dealing it a thousand times.

Eventually, of course, we got the computers, and now we can solve for the equilibrium. Thus, AIs can now defeat all human poker players.

There’s never going to be a computer that will play World-Class Poker. It’s a people game. —Doyle Brunson (625)

But the claim that a computer would never play world-class poker? That might be the worst bet Brunson ever made. (640)

Sure enough, though, another AI poker bot—a descendant of Polaris named Libratus—won a heads-up no-limit challenge in 2017. And then, finally, another younger sibling called Pluribus beat humans in a multi-way no-limit match in 2019. (746)

Brunson’s claim is that the ability to read, understand and exploit humans is too valuable, including the crucial skills of convincing that human to keep playing in games where they are going to lose and to do so for high stakes in an imprecise way, so a computer will never do as well in practice as a human.

In the long run, Brunson will be wrong about that. But if you were betting on the WSOP main event, which do you think would have higher EV- a computer playing fully GTO (game theory optional) or Thomas Rigby (‘the Stoic’) putting maximum pressure on everyone? I am very convinced that at least through about day 5, Rigby is going to do way better. And if the question is, who will make more money trying to put together a cash game, again you want to go with Brunson.

For now. Eventually, the AIs will also get much better at everything else, too, and Brunson will be fully wrong.

But also ask the question, did Brunson lose that bet? Yes, he gets quoted as being wrong in a book in 2024, but did he end up worse off on net?

Always remember the odds.

One fun thing about poker AIs is that they bluff a lot more than the most successful humans used to, and mix up decisions more often.

The frequency with which the computer liked mixed strategies was surprising to Lopusiewicz, who had never formally studied game theory. “I knew that there was going to be a lot of mixing, but I didn’t realize it was going to be on almost every hand,” he said. (1037)

Computers bluff their faces off when playing poker because that’s what game theory says to do. This can include making big bluffs—going for a knockout punch. But more often, they prefer small, tactical bluffs—say, betting $20 into a pot of $100—the equivalent to a boxer’s jab. (1059)

My intuitive way of understanding this is that it is very hard to maintain ‘balanced ranges.’ If you bet small, you need to bet small with a wide range of hands at different times, such that opponents cannot ‘smell weakness’ and raise, or trust you too much and consistently fold. Same goes for your big bets and your checks (not betting), and your raises and calls, and so on.

That means you need to mix up your decisions a lot, in order to put your opponents into bad spots while not giving anything away. Humans in practice find this difficult, especially sacrificing EV to do ‘unnatural’ things so often. AIs showed them why they have to do this anyway.

There are a few asides on randomization in baseball and other sports. Obviously if you are a pitcher, you need to be mixing up your pitches, so they don’t get to look for a particular pitch. If you don’t think you can actively outguess hitters, you want to use randomization.

And yet, Verlander throws his fastball only about 50 percent of the time. How come? Well, major league hitters are pretty good. And although Verlander’s fastball is tough to hit, it’s easier if you know that it’s coming. (970)

Greg Maddux—one of the most cerebral pitchers of all time—reportedly did exactly this, using quasi-random inputs such as the stadium clock to decide what pitch to throw. (992)

In sports, a play like a fake punt is terrible if the other team knows it’s coming, but it can be a great play if the opponent isn’t anticipating it. (996)

Exploit and you risk being exploited. (1071)

If you throw your fastball too often, everyone will figure that out. If you have a pattern that is too obvious, they will figure that out too, more so today than in the past. The same goes for many other similar choices in sports.

The question is, to what extent do such worries have to ‘keep you honest’?

There is obvious danger in being too predictable. But in general I see too little exploitative play attempted, not too much.

This is especially true when the decision is asymmetric, and when the opponents’ decisions can be observed.

In football, you can see how the defense sets up play after play. They can surprise you, but their decisions are observable on every play and a given formation will limit their options. In particular, if they decide to pack a lot or only a few players into ‘the box’ to defend against the run (as opposed to defending against a pass), that tends to be mostly consistent between plays and has a big impact. And with a given matchup between teams, it often quickly becomes clear what offensive options are +EV, and which ones are not working.

The best offenses will then ‘take what the defense gives them,’ and repeatedly exploit the opportunities they see each play until the defense adjusts. Yes, they will mix up their play a bit, but if they see soft coverage that allows an easy pass, or there isn’t enough good run defense, they will pound you play after play until you fix it.

Whereas others often end up with what I call a Running Running problem.

It is totally fine to have a Running problem, where the defense is ‘taking away’ your ability to successfully run. Pass the ball.

What is not okay is when teams are in this spot, and then they run the ball half the time anyway. That’s the Running Running problem – you have a Running problem that you can’t run, and then the problem is still running.

The happens when the coach either feels the need to ‘establish the run’ or to give the impression that they might run on any given play. Otherwise, the defense would be able to adjust. But you can skip all that. Then, if the defense does adjust? Now you run. Or, in reverse, with the pass, once they start moving everyone to run defense.

If the defense proves so good at anticipation and adjusting that they can outguess you, and is mixing up their own play smartly? Retreat to game theory. But if they’re predictably exploiting you, don’t sit there wasting plays that predictably fail.

Another phenomenon we see is teams ‘saving their best plays’ for high leverage situations. In those high leverage situations, where the play matters most, they turn to their best plays and players and pitches and so on more often. So essentially they are playing a mixed strategy in other spots, where they do their weaker options ‘too much,’ then switch when they need a win.

In theory, aside from hiding truly unique new plays, this shouldn’t work. The other side should do the same adjustment, and anticipate that the offense will call its best play, or the pitcher will throw their best pitch. And if that adjustment wouldn’t work to stop it, then why isn’t the offense calling their best play every time?

But somehow this does work in practice. People get into a rhythm. In baseball this makes more sense, because you want to vary selection in part to make it hard for the batter to calibrate. In other sports, it’s weirder.

Most people do not understand the underlying math or logic of most of what they do.

Instead, they have memorized a bunch of rules or heuristics and copied behaviors. That works 99% of the time for 99% of things. It is typically wise.

When you ask them to solve a new problem, that’s different. Whoops!

Although solvers might seem like a natural fit for Selbst, whose mother drilled her on logic puzzles when she was growing up in Brooklyn, she’s instead a deep believer in exploitative play: figure out what your opponents are doing wrong and take advantage of it. “Most people—save like the best hundred people in the world,” she told me one day over happy hour drinks at a Manhattan wine bar, “just aren’t that amazing at game theory.”

Instead, Selbst says, players have a lot of experience in situations that come up frequently in poker.

But memorizing only gets them so far. “They were very good at playing a spot 1,000 times and knowing what to do in that spot,” she said. But she found “that over time, whenever I put someone in a new spot, they would just royally fuck up.” (1116)

Poker is a compact game, such that it should be difficult to put someone in a truly unique spot. But sometimes it can happen.

The same thing also works in other games. This is the joy of the rogue Magic deck, when it is doing something others haven’t seen before. Or the unusual opening in chess, to take the opponent out of not only their ‘opening book’ but their principles for afterwards. Put them in a spot where you know everything, and they don’t.

Even if you are capable of solving the game theory or other puzzle given time to think it over carefully, you do not have that time. You need to make a decision now. Nor do you get to use trial and error, or get to study up.

The trick is that you have to give up less value in your unique play, which almost has to be an inaccuracy to work, than they give up not knowing how to respond.

My favorite poker version of this is watching Phil Hellmuth, who Nate Silver reports is exactly who you think he is from watching on television.

Phil Hellmuth spent about forty minutes showing me his trophy room. (1648)

When we later sat down at the dining room table in his comfortable home in Palo Alto, California, it was clear that my effort to tamp down Hellmuth’s ego had failed. He spoke in a fifteen-minute stream-of-consciousness soliloquies where I couldn’t get a word in edgewise. (1652)

Why craft a persona, when you can use your own genuine one?

Phil’s entire strategy is based on using a pattern designed to provoke the opponent into playing exploitatively or drive them into various forms of tilt. And then, through decades of experience, to know all the different ways players respond to that, and what to do about each one. Its seemingly obvious weaknesses and patterns are a feature.

You, at home, see what is happening and are screaming. But if it was you playing then he would pick up on your response pattern, and be using a different variation.

He combines this with his ‘white magic,’ his ability to read people, especially to read people in how they react to him, where his unique style and reputation make them have less effort available for mixing up play and being hard to read.

But one important connection Hellmuth drew is that his white magic comes and goes with his energy level. The idea that people-reading ability comes from the body rather than the mind—so that when we’re tired, our social skills suffer more than, say, our ability to work out an equation—is something straight out of Coates. “Sometimes I’m at full fucking reading ability. And when I’m at full reading ability, I’m dangerous,” he said. “Why don’t I have more bracelets than I have? (1689)

“What’s happened is fatigue has killed me. So I get too tired. And I lose control. And I play badly.” (1694)

Then every so often (for example, on High Stakes Duel) he runs into Jason Koon who ignores all that and plays GTO (Game Theory Optimal) while ignoring the antics, and Phil gets destroyed. Whoops!

Same goes for non-game actions too. If you are out of your comfort zone, in a new situation, you will handle pretty much anything vastly worse at first, no matter who you are or how calm you are, versus how you’d handle it the 10th or 30th time. That’s unavoidable.

The fatigue thing is a huge deal as one gets older. The reason I no longer go to Magic tournaments is partly because I am too busy, but mostly because I lack the stamina. There’s no way for me to bring my A-game for three full days. That’s also the reason I probably won’t ever play the WSOP.

Another mysterious thing is that idea that ‘the body knows’ tells rather than the mind.

“One thing I found out really quickly when I first started playing was I had a really strong instinct for when my opponents were weak or strong,” said Maria Ho. (1722)

Coates told me something similar: we should generally pay attention to the signals our body is providing us with even if we can’t quite figure out why. “The thing is our physiology is really smart,” he said. “I mean, it’s just really smart. It’s very hard to trick your physiology…. (1735)

I am, as one would expect, skeptical of this explanation. I don’t deny their lived experience. I’ve certainly had plenty of situations in which I had that same kind of intuitive or instinctive sense about things, including what was up with someone. Over time you learn when you can trust that, and when you can’t.

I don’t see it as coming ‘from the body’ or physiology in the way Ho and Coates are describing here. To extent that this seems like a real thing I can think back upon, I would think of that as the method of getting the output rather than how one is doing the calculation.

As in, Coates has found that when his unconscious analysis picks up on something, it causes physiological reactions, and it is easier for him to notice the secondary physiological reactions than it is to tune (directly or not) into the output another way. And that this is true even when you cannot trace back the reason.

I suppose it mostly adds up to the same thing?

A last note on tells is a warning not to focus on physical patterns too much.

“Even against bad players, the relative weight of the physical thing is usually pretty low,” Adelstein told me. “You really need some really great information, specifically like them doing that physical thing when they either almost always have it, or almost always don’t, over a pretty good sample size.” (1735)

I am highly convinced that most Magic players have rather obvious physical tells, if you were able to watch for them carefully over enough matches. Myself included, and I put considerable effort into not doing that, partly into the obvious stuff that people do notice, and partly into more subtle stuff I doubt almost anyone ever saw. But in general, actively focusing on trying to find the physical tells of others beyond a few universal patterns was a fool’s errand, as you were better off focusing on other things.

What Nate relies upon most, and feels are reliable enough to be good, are stereotypes.

Because it’s a game of limited information. So you’ve got to take whatever you can get. And I will say most of these stereotypes are correct 70 percent of the time. (2112)

Seventy percent is pretty good. The thirty percent of the time they’re wrong, you get burned, especially since they likely know what you are expecting, but that’s a risk you have to take. That’s poker. You can’t afford to disregard Bayesian evidence.

Here is something I wonder about.

“If there’s one thing that you figure out after many, many, many, many hands, it’s you know what 52/48 is,” said Annie Duke, referring to a player who can distinguish a 52 percent chance from a 50/50 spot. “That’s the type of distinction that most people are very bad at making. But poker players are really good at making that distinction, and they can feel it.” (1771)

Sports betting in particular absolutely does give you this feeling. You know in your gut what 52/48 (also known as -110) looks and feels like.

When I played poker, I did not feel I was learning to do that. Sure, I knew the odds, but none of it felt that kind of precise. It never felt so important. What exactly do you do with that information? In sports knowing that the game is priced 52/48 instead of 48/52 is a huge opportunity. In poker, that is a small mistake, and it does not seem like you get enough feedback to realize it – you’d have to get that distinction from a solver, or from doing some sort of abstract math.

One of the lessons from Jane Street that I can share is the idea that you should consciously choose to have the proper emotional reactions to events. Traders talked that way. They did it on the direct level: ‘I’m sad about [X]’ or ‘I’m happy about [Y].’ And also on the intention level: ‘You should be sad about [Z].’

You are training your brain. Providing it with feedback, so it can learn. You want to give it the correct feedback. For humans, in practice, that in part means the correct emotions.

So what if you could replace traders—or poker players—with AI systems and not have to worry about all that icky body chemistry? Actually, that might not be such a good idea. We get a lot of feedback from our physical beings; this is one reason some experts are skeptical that AI systems can achieve humanlike intelligence without human bodies.

In fact, Coates’s studies found, the most successful traders had more changes in their body chemistry in response to risk. “We were finding this in the very best traders. Their endocrine response was opposite of what I expected when I went in.” (1551)

The AI systems, if they could otherwise replicate the traders, would actually be an insanely great idea. They’d have huge advantages, including faster speed, more memory and ability to learn and absorb data in real time and overall, and so on. And they would absolutely find alternative feedback systems to capture these benefits.

What you would not want to do is give these humans pills that dulled their responses.

The other key is, you don’t want that to be about winning or losing money…

“We’ve wired up traders that were big hitters, and they were poker-faced,” he told me—the traders occasionally got angry when things were going poorly but mostly kept a calm demeanor.

“But even when they were poker-faced, what was going on under the surface was far more important because their endocrine system was on fire when they were taking risk.”

In one experiment, for example, Coates studied the testosterone levels of a group of traders at a high-frequency London trading shop. He found that their testosterone was significantly higher following days when they’d made an above-average profit. But the converse was also true.

Coates had also been checking their testosterone in the morning and discovered that they had substantially better trading days when they woke up with higher T levels. Higher testosterone predicted more trading success, in other words. (1528)

Raising testosterone after a good day means you’ll likely be more aggressive and go larger and faster and break more things. It also means you are going to feel good about things having gone well. Neither of those seems ideal? Focus on the process, not the outcome.

What to make of higher T being actively predictive of better results? This could be causal through any combination of things such as greater energy, higher status or willingness to take more risk.

Or it could mainly be that higher T was predictive without being causal, and traders had an idea of when they would make more money. If you are waking up for a big day, say triple witch or Fed day or an important earnings day where you get to make the decisions and you’re prepared, you are going to be more excited. Momentum effects would also do this, or anticipation of realized gains, and so on. Makes sense to me.

Similarly, I bet that back when I was competing in Magic, this correlation would have held, but that was likely mostly predictive rather than causal – when I had a big day or a great deck or opportunity, I was going to be more pumped.

In big moments when you’re confronted with a high-stakes decision, you’re essentially working with a different operating system than the one you’re used to. (1571)

However, it helps to have practice under these conditions. In fact, having a physical response to being confronted with risk can be a healthy sign. Coates was once contacted by a group of researchers with the Great Britain Olympic Team. Their best athletes, they told him, were just like his best traders, with low baseline levels of stress response that then dramatically spiked when it came time for a big competition. This even applies to golf. (1573)

My own research shows that experience matters considerably in the NBA playoffs. And we shouldn’t neglect the testimony of athletes themselves, from Michael Jordan to the great Montreal Canadiens goalie Ken Dryden, who describe similar experiences of being in a zone when under intense pressure. (1602)

I can confirm.

And this makes me think it is at least plausible that ‘big game experience’ is a real factor in the playoffs, including in the NBA. You learn to handle the moment. But also I know the gambling market mostly does not care, to the point we ignored it when predicting future market prices and that seemed fine.

Also I notice exactly when I had that big stress response for the big competition.

It wasn’t during the competition, when decisions were being made. It was directly before the competition. Same was true with exams. Hugely nervous while hurrying up and waiting. Completely calm once the exam started.

When I think about the times I’ve been in a zone or flow state, it’s often been in response to stress. It’s happened at big moments in poker tournaments, occasionally during public speaking appearances, and even a couple of times when I was writing or coding under intense deadline pressure.

I’ve also had it happen on election nights when I’ve been covering the results. What these had in common is that they were all extremely high-stakes moments where I had thousands of dollars in money or future earning potential on the line. (1582)

That was not true for me at first. I was nervous as hell early on during big matches. But I learned over time not to do that. It was not helpful. Instead, you went through the stress beforehand, and then sat down and you were in the zone. It was like any other match, except hopefully more dialed in.

And yes, it very much helped to have played on the biggest stage. I was able to handle my first time because I was (not coincidentally) Crazy Prepared that tournament. After that the pressure mostly didn’t phase me anymore. Nate Silver has the same story, where he made a deep run into the top 100 at the WSOP main event only to lose to a set-over-set cooler where he 100% was absolutely supposed to go broke every time:

It was heartbreaking. I may never play such a big pot again in my life. But I’ll tell you something: ever since that hand—losing the equivalent of a nearly million-dollar pot that for a split second I’d assumed I was a big favorite to win—a bad beat for $300 or $3,000 or even $30,000 hasn’t felt like such a big deal by comparison. There’s nothing like pain to build up your pain tolerance. (1644)

There are other times, especially when working open ended on something important, when the stress is a huge negative, a distraction, causes aversion and procrastination, gets in the way. That happens largely because there is no clear dividing line when the battle starts. One response to this is to wait long enough that it is clear the battle has indeed started, and you have no choice, as in:

It’s not panic. It’s the stress of the upcoming need to work, of the decision whether to face the problem, leaving the body. You’re in the zone, baby.

Tendler referred me to a set of research called the Iowa Gambling Task, so named because it was originally conducted by professors at the University of Iowa College of Medicine.

It works like this: participants are asked to choose from four decks of cards—A, B, C, and D. Each card gives the player a financial reward or penalty.

Two of the decks—let’s say A and B—are risky, with occasional big wins but lots of penalties and poor overall expected value.

The other two—C and D—are safer, with a positive expected return.

The pattern isn’t that subtle, and after turning over a couple dozen cards, the player usually figures out to avoid the losing decks. The research found, however, that players have a physiological response to the risky decks before they detect the pattern consciously.

Their body is providing them with useful information—if they choose to listen to it. (1586)

This is one of those studies that drives me nuts. Why would you hopelessly confound your study like this, combining risk level (high versus low) with expected value (positive versus negative)?

What was this physiological response telling us? It could be any combination of:

1. The deck is risky, oh no.

2. The deck is -EV, oh no.

3. I am probably going to get penalized, oh no.

4. I am buying information, but I expect deck to usually be -EV, oh no.

The researchers are perhaps assuming that the players are picking up on the -EV. That does not seem so likely to me. Instead, my guess is the players physiological responses are mostly noticing the risk, which the players obviously did consciously notice the moment they see the first big negative card.

The proper experiment, of course, is where (with names randomized per player):

1. A and B are risky, C and D are safe.

2. A and C are profitable, B and D are unprofitable.

Even worse, notice that A and B have lots of penalties and only a few big wins, rather than the other way around or being symmetrical. That confounds us even more. Even better would be:

1. A, B, C and D are risky. E, F, G and H are safe.

2. A, B, E and F are profitable. C, D, G and H are unprofitable.

3. A, C, E and G have mostly small plus cards and some big minus cards. Whereas B, D, F and H have mostly small negative cards and some big positive cards.

4. The numbers are randomly shuffled enough it isn’t too obvious right away.

Then see what happens.

Is it, in general, not taking one? The claim here is essentially yes.

For the most part—at least when it comes to financial and career decisions—people do not undertake enough risk. It’s certainly true in poker. For every player who’s too aggressive, you’ll encounter ten who aren’t aggressive enough. (1536)

It’s true in finance, according to Coates. “A lot of asset managers and hedge funds I’ve dealt with have a problem getting their good traders and PMs to use their full risk allocation,” he told me. “They’re not taking enough risk.”

It’s true when people are contemplating personal changes. Annie Duke’s book Quit—Duke is a former professional poker player who quit the game in 2012 to study decision-making—is full of evidence about this.

An experiment conducted by the economist Steven Levitt, for example, found that when people volunteered to put major life decisions such as whether to stay at a job or in a relationship up to the outcome of a coin flip, they were happier on average when they made a change. (1538)

Faster, you say? Not so fast!

At one final table of the World Series of Poker (WSOP) main event, one player had a sheet of paper with reminders. One of them was: Folding is only a small mistake.

This is very much a case of Reverse All Advice You Hear. Some of us need to learn to fold less and play loose. Others need to learn to fold more and play tight. Mostly players need to fold more, because you can’t win by folding, and folding is not fun, and the amount you should be folding is not intuitive. But there are big exceptions.

What that has in common with aggression, and with risk taking in general, is that making a mistake is an asymmetric risk. In most situations, from most perspectives, being too aggressive and taking too much risk is a lot worse than being too passive and not taking enough.

You would much rather be betting half of Kelly than double (or 1.5x) of Kelly. It’s a skill issue too. If you have a skill problem, aggression will be harder, and make things vastly worse for you. Which, in turn, can compound your skill issue, because you do not learn. You’re only ‘not taking enough risk’ if you know how to take smart risks.

Amusingly, the Kelly criterion didn’t actually get a mention until near the end:

If I had laid odds, you ought to have bet on me mentioning something called the Kelly criterion before now in a book about gambling. (7193)

Poker is a special case where the baseline aggression is indeed way too low, and by default most players are too passive – good poker is tight and aggressive, whereas most players are relatively loose and passive.

In finance, notice the adjective ‘good.’ It’s hard to get the good traders to take enough risk. The bad traders? You wish they’d take less. Often they take way too much, or you’d prefer they not take any.

A good trader learns to make the small mistake of not going big enough, to wait for the best opportunities, and also to protect the downside or risk of ruin. They place high value on survival, and know that being down a lot less than 100% can still be fatal if you have a manager.

As the asset manager, sure, if you know who your best trader is you want them taking more risk. That’s your incentives, not theirs.

What about quitting or otherwise making major life changes?

There are a lot of factors here.

1. A lot of it is that big changes, and big risks, tend to be negative in the short term, and often involve a blast radius. Breaking up is hard to do, and it sucks, even when there’s nothing tying you down. Quitting your job hurts the company you leave behind. Moving is a huge pain. Trying and learning new things is a lot of work, at first.

2. We want to be able to make commitments and honor them. If you’re on the fence about breaking your commitments, then probably yes you will be happier if you break them, but you should have a higher threshold than that because it also matters that you don’t do that.

3. We want to guard against errors and momentary preference changes. A relationship won’t survive for long if either party is willing to leave it on a whim, or whenever they think they can do 25% better.

4. If you ‘take the risk’ and make the change, then what happens is Your Fault. You did the action, so you’re blameworthy. If you do nothing, then it isn’t your fault. Or at least, you can tell yourself that.

5. Also there are various biases that hold you back and shouldn’t, including bewaring trivial inconveniences or The Fear.

6. If you are considering doing [X] continuously, on the fence about it, then the cumulative chance of doing [X] could end up very high while you feel indifferent about the decision at any given time – so on average [X] would be wise.

7. Finally, such studies are usually medium term, which also excludes the long term. Often sticking with things means making a long term investment.

The generalization is that you almost certainly want a decision policy, whereby when you are not sure what to do with the decision whether to take risk or make changes, on average the active risky decision will make you happier in the medium term.

Perhaps the question on changes should be: If this was the last moment I could possibly do this, and someone else was advising me, would they advise me to do it?

Playing in cash games is low variance. Table selection is huge, but for most purposes each pot is independent of each other pot. So as long as you hold your stakes relatively steady, you’ll get ‘justice’ over a not-too-long period most of the time.

Grind out the $2/$5 game at the Bellagio for fifty hours a week, and you might have an expected value of $80,000 to $100,000 per year with little risk of going broke. But the idea of regular hours and a steady paycheck defeats the purpose of playing poker in the first place. (1910)

So instead the real action for a lot of players ends up in tournaments. The problem with tournaments is that payouts are extremely top heavy, so even over long periods you can be +EV and yet run very poorly.

I had the same issue when I was trying out Daily Fantasy Sports (DFS) for a brief period. Me and my DPS partner together managed to consistently have lineups that posted above-average mean scores, while also having higher in-lineup correlations.

We should by all rights have done well, but to take advantage of the correlation advantages we had to enter big tournaments. At one point we went into the last game of the Sunday Millions on FanDuel with a substantial chance of winning outright. But we never spiked that big win, and the edge you need without spiking that huge win is utterly massive. For that and other reasons, I gave up and ‘got a real job’ instead of continuing.

Poker players hate real jobs. Indeed. I knew, quite obviously, that getting a job at Jane Street or another firm would pay vastly better and more consistently than gambling. I figured I had what it takes. But who wants that? Indeed, exactly that issue – Jane Street is a great place to work, but not wanting a ‘real job’ where one had to show up and be ‘on’ in the right way consistently, was why that job did not work out for me.

It’s a field that glorifies hyperrational decision-making, but a lot of people who play poker for a living would be better off—at least financially—doing something else. The combination of mathematical and people-reading skills necessary for success at poker should generally also translate to lucrative opportunities in tech, finance, or other River professions—usually in jobs with health care benefits and far less variance. However, much of what attracts people to poker is an anti-authority streak. It’s one of the only professions where you can truly be a lone wolf. (1913)

The poker version of the variance problem is even worse.

We specified that this player is a step short of elite, but somewhere between the one-hundredth and two-hundredth best player on the live tournament scene—a good regular who you’re never happy to see at your table. Let’s call her Pretty Good Penelope. (1810)

What’s the bottom line? I estimate that in an average year, Penelope will earn about $240,000 from tournaments. “Wow, that sounds pretty great!” you might say—travel the world, play cards for a living, and make several times the average American salary.

That doesn’t really tell the whole story, however. For one thing, there are taxes (a big issue based on the way the IRS taxes poker players) and expenses (high, since we’re assuming Penelope’s on the road about a hundred days per year).

But the big problem is that the $240,000 figure is kind of fiction. It’s an expected-value calculation over the long run, and Penelope is never going to reach the long run. There’s so much variance in tournament poker that even if Penelope played for fifty years, the swings wouldn’t really even out.

In fact, despite being one of the two hundred best tournament players in the world, she’ll have a losing year almost half the time. I discovered this by simulating Penelope’s schedule ten thousand times, using payout tables from actual poker tournaments. There’s one other question you might ask: Penelope intends to spend more than $1 million in tournament buy-ins per year—so where the hell is she getting all this money? (1860)

Even more than gamblers, Magic: the Gathering professionals are punting massive amounts of equity to get the lifestyle they want. If you can win Magic tournaments, you can win at far more valuable activities. If you ever have a chance to hire former or current Pro Magic players, as I did a few times, you jump at it – they are whip smart, they are severely underpriced, they are great to work with and play to win.

The good news for us was that Magic tournaments are far less random than poker tournaments. Skill plays a bigger role, and plays it far faster. It is highly practical to consistently do well, and to have a substantial chance of winning out of a field of hundreds or even thousands. Poker is not like that.

The thing about poker tournaments is, even if the EV is lousy and the variance is gigantic, they are still very clearly awesome. They’re fun, they’re exciting. I feel the pull. If there were local tournaments and I wasn’t going to get completely killed I would likely sometimes attend them.

Degen means degenerate gambler. A degenerate gambler loves the action. They’ll bet on anything and everything, often without an edge. They’ll bet too big and risk going broke, when it makes no sense to do that, even when they otherwise have it made.

(Being considered a degen is sometimes a badge of honor in the River, so long as you’re not the sort of degen who harms others. Better a degen than a nit, certainly.) (1955)

I am not a nit (or like to think I’m not!) but I am definitely not a degen. You will not see me take zero EV bets. The idea of being +EV and then losing is fine, but the idea of making a -EV gamble (as an actual mistake, rather than e.g. part of a strategy that makes a set of gambles some of which need to be -EV) fills me with dread. I hate mistakes with a passion. But I also realized that while you could always choose to do nothing, holding back when you shouldn’t, not deploying your capital when you had a great opportunity, or waiting too long, or being too careful with your bankroll was also a mistake.

I then discovered a funny thing about sports betting. Essentially all the most successful sports gamblers are degens.

It seemed possible I was the most successful sports gambler who wasn’t a degen. Almost no one who has to teach themselves to be fine with risk, rather than loving the action, gets all that far.

Later, we see a claim to the contrary, from sports gambler Spanky.

Spanky claimed to me that this suited his personality. “I don’t like gambling really. None of us like real gambling. We like winning.” (3082)

Yeah, I have no doubt Spanky said that, and he might even believe it. It might be true of him in particular. In general? It’s simply not true. They like winning. They would probably not go so far as to say ‘the best thing other than gambling and winning, is gambling and losing.’ But they like gambling.

I made sense of this by reading the forums where non-degen ordinary gamblers discussed their strategies. In particular, where they discussed their bankroll management.

You’d have people who would bet 1% of their bankroll (total amount they had to gamble with) per normal game, and picked spots carefully. Then, if they had a game they truly loved, maybe they’d go up to 3%.

Which sounds all safe and responsible and reasonable, until you realize they had a bankroll of something like $50,000 and every month they had to pay their rent. So they would have to net win something like ten bets every month in order to break even. Or they wouldn’t be paying rent off the bankroll, but they’d have a tiny one, e.g. $10,000, and they’d be acting like losing it was infinitely terrible.

A lot of them of course did not have +EV, and lost. But what happened to a lot of the ones that were indeed +EV, was that they’d do this for a few months, and they’d win, but they wouldn’t end up with a much larger bankroll than they started with. Meanwhile, their life was ending one minute at a time. What was the point?

When I watched the true degens, I saw that they were (mostly) the only ones able to move boldly, for size, when opportunity knocked. They were the ones that got into position to play for size, and also the ones that took advantage.

The downside, of course, is that degens can make rather large mistakes, either in EV terms or in terms of risking a true blowup. And yep, I saw a bunch of that too.

The whole degen attitude, to me, is kind of weird?

That’s another thing you might not expect about high-stakes players. They tend to be generous with their money: tipping well, loaning to friends, offering to pick up checks, and so forth. Relative to other successful people I’ve met, they’re more aware of the ephemeral nature of money and the role that luck has played in their success. (1960)

My brain thinks this should work the other way around.

If you are drawing a steady paycheck, then you can be super generous. You get X dollars per year in income, you have Y in expenses and Z in savings, so you know what you can afford to spend to help out the less fortunate, or on a luxury.

Now suppose you are a gambler. Maybe you’re Penelope the pretty good tournament poker player. She could go an entire year, play her A-game, and lose money. Two years from now, she might be broke as she throws $10k after $10k at tournaments and fails. If she struggles, suddenly she’s going to need every penny she can get.

There was a period where I was gambling, and most of my net worth was in my formal bankroll. I was very careful with my expenses, so I could afford to take risk gambling.

When I look at poker players being ludicrously generous in these ways, without longer term financial security, it seems insane to me.

Yet it is a common pattern. Those who come into cash at random tend to be generous with it, for themselves and for others. They make it rain. Even when they know full well if asked that there will come a day they will again be broke or close to it.

Some of that is that such people often understand that they face various wealth taxes – pressure to distribute their gains over time, or inability to collect various benefits and opportunities. So instead they invest in experiences and social capital.

With poker players, I am guessing the core idea is the same. You come into cash. You know that you will probably find a way to blow it, and are now expected to be more generous in various ways, to gamble higher, and so on. So you act generously now, enjoy your winnings, and build social capital. Then you can ‘afford to’ go broke later.

People have a variety of different and bizarre relationships with the concept of luck.

Some people even think they are lucky, or others are lucky. Poker players especially seem to think this reasonably often.

Odo (From DS9, and often): “I don’t believe in luck.”

Michael McDermett (Rounders): Why do they insist on calling it luck?

Many Magic players (often): Better lucky than good.

Spider Murphy (from Cyberpunk/Netrunner): If someone has consistently good luck, it aint luck.

I don’t consider people ‘lucky’ per se outside of a particular context. But there is the theory that ‘thinking of yourself as lucky’ is good.

Richard Wiseman’s 2003 book, The Luck Factor, asserts that people who view themselves as lucky benefit from it in several ways.

Here’s a version of his claims:

Lucky people “constantly encounter chance opportunities” and try out new things. Lucky people “make good decisions without knowing why.”

They listen to their intuition.

Lucky people have positive expectations so their “dreams, ambitions and goals have an uncanny knack of coming true.”

Lucky people “have an ability to turn their bad luck into good fortune” because of their resilience.

If you remove the sheen of self-help-book claptrap, there’s a kernel of something there.

Under this theory: Lucky people don’t believe in luck, or take whatever luck gives them, or they make their own luck. They give themselves maximum opportunities to get lucky. They give themselves outs. They watch for opportunity, and they seize it.

I basically buy that, mostly, as part of a package of ways one can usefully look at things. It is a version of the Zero Effect rules about looking for things – if you go looking for something specific, your chances of finding it are very bad, because of all the things in the world, you’re only looking for one of them. But if you go looking for anything at all, your chances of finding it are very good, because of all the things in the world, you’re bound to find some of them.

The part where positive expectations means things have an uncanny knack of coming true? That part is weirder. Confidence is an obvious factor. So is being convinced to try at all, and look for a way to actually make things happen, which is remarkably rare.

A lot of this feels to me like the type of advice that is true on average, and for most people on the margin. Of course, my readers are weird, so beware.

The most important words about luck might be Spider Murphy’s. Anything that looks sufficiently like consistent luck is not luck. Almost every time I went on an extended losing streak betting on sports, I was able to later figure out why it happened, and the answer was not ‘luck.’ DPS alone I actively think was luck, because it was front-loaded tournaments.

Then there’s a series of systematic losses in college basketball that I still don’t understand. I doubt the answer is luck.

Vastly more men than women play poker. Why?

The first obvious answer is straight up misogyny and poor treatment.

So what explains the poker gender gap? The explanations I’ve come across fall into roughly five categories. 1. There’s often openly abusive and misogynistic behavior toward women, made worse by a “What happens in Vegas stays in Vegas” attitude. (2025)

“I’ve actually experienced the worst misogyny at the lowest stakes. Where people are there to have fun and they’re drinking and they feel like I’m cramping their style,” said Maria Konnikova.

However, I’ve also heard awful stories about misconduct coming from highly regarded high-stakes players—but the worst stories often come only once you’ve turned the tape recorder off. (2042)

Men struggle to make adult friendships, and poker provides a means for male social bonding that appeals to a broad cross section of men, but women aren’t always invited to the party. (2045)

Maria’s comment is interesting and the last one reinforces it. One can have a model that few women play poker, so men who prefer that choose to play poker in order to be in a male-dominated space and get a relaxed form of social interaction they can’t get elsewhere, which makes them treat women who do show up badly, perpetuating the cycle.

And I totally believe that there are plenty of awful misconduct stories out there.

Here are some other stories that make sense and probably contribute.

Men—whether through nurture, culture, or nature—tend on average to be more competitive and aggressive, essential attributes for success at poker. (2059)

Men have more financial and social capital to gamble with. (2072)

“Society doesn’t really encourage women from a young age to take risks. We’re taught growing up that we always have to be more responsible,” said Ho.

It’s an advantage to have the option of blending in, and that’s easier when you’re a white guy. (2086)

My note would be that it’s not so much that men ‘have what it takes to succeed’ in poker, in terms of competitiveness and aggression. It’s more that men more often ‘have what it takes to want to play’ and exhibit this type of competitiveness and aggression and social interaction.

All the other stuff matters a lot, and poker (like most other games) could and needs to do a much better job making women feel welcome and ensuring they have good experiences, but at the end of the day there has to be a reason everyone shows up for a zero-sum game. You have to want to do it enough to invest your time, it has to be rewarding. That’s going to skew the gender ratio.

Nate Silver spends a bunch of time on a cheating scandal at the Hustler Live game.

It was a fascinating situation, such that I’d already listened to half an hour of Doug Polk explaining the various arguments about it.

Essentially, there was this televised game. A player went all-in on a semi-bluff, and Lew made a hero call with J4 that made absolutely no sense. It was a +EV call in isolation if you knew exactly what cards her opponent had, but she couldn’t beat a huge percentage of even his bluffs. So it looks hella suspicious. But also, if you were going to cheat on a livestream, you would pick a much better spot that also was far less obvious.

What happened before is that there were several instances—both during that day’s taping, and on two previous episodes of Hustler Casino Live—where Lew would have benefited from cheating, but didn’t. Most cheaters aren’t like that. (2197)

But in nineteen hours of livestreamed hands across three sessions, there’s no hand other than J4 that looks like cheating—even after thousands of detail-obsessed poker players have scoured Lew’s footage for signs of impropriety. There’s also the fact that if Lew cheated, she picked an awfully bad spot for it. (2204)

I have a very strong intuition here as well. Cheaters do not do this. They do not have the patience or discipline. When they do, they pick better spots.

In conversations with Nate, Lew didn’t come across so well. Her surface story doesn’t add up, but that could be because she doesn’t want to admit she misread or misremembered her hand.

But my spidey sense was getting a weird vibe. At times, Lew looked out into space as though she was reading off a teleprompter. And Lew has a habit of relaying information that is either superfluous or doesn’t entirely check out. Two sources that I spoke with used the term “pathological liar” to describe Lew’s habit of tangling herself up in knots. (2123)

Ultimately, I don’t think one can know and neither does Nate.

I’m not saying this is the only theory—if I were handicapping, I’d still put the chance of cheating at 35 or 40 percent. (2222)

I would bet on the ‘no cheating’ side of that line, if I had to bet, but it seems very reasonable.

What’s remarkable is that for a long time poker players basically tolerated cheating.

Until the Poker Boom years, the prevailing attitude was often that a poker player would win by any means necessary. (2179)

Magic in its first few years was a lot like this as well. It was your job to defend yourself, if someone cheated you that was your fault. Then, in both cases, the act got cleaned up, and cheating become both rare and a good way to get exiled if people were convinced you did it.

The bigger lesson from this is actually about Adelstein, the player Lew won the hand against. There is no question Adelstein legitimately thinks that Lew cheated. But by throwing a fit over it, and insisting that she pay him back, while Adelstein won the battle – he got his refund, after also drawing live – he very much lost the war. He’d had a great game going with lots of +EV, and after this incident it is gone. Compare this to Brunson in his early days, who took being robbed in stride as part of the game.

That’s the thing about gambling. If you want to win big in cash games, you have to be good. But a lot of people are good. The rarer and more important skill is getting those who aren’t good to give you the action. Often that means ‘knowing what they are prepared to lose’ or otherwise ensuring they feel good about coming back. Other times, it means eating a loss, even if it isn’t fair.

You have to factor in everything, but only to the extent it deserves. It’s easy to either let most considerations go, or to weigh them all too close to equally, which are the human default settings.

In this case, Wesley bet huge into Dwan in a gigantic televised pot, and Dwan won by making a ‘hero call’ with a hand that would ordinarily not be good enough to catch bluffs. Dwan had enough different reasons to be suspicious he felt he had odds to call.

There were roughly twenty “data points” that factored into this decision, Dwan said. For instance, Dwan thought that Wesley—ordinarily a tight player—had come into the session with an “agenda that was a little bit less about winning money” and more about making plays that looked cool on TV.

Wesley had to be bluffing 25 percent of the time to make Dwan’s call correct; his read on Wesley’s mindset was tentative, but maybe that was enough to get him from 20 percent to 24. (Bluffing in a huge pot looks cool on TV.)

And maybe Wesley’s physical mannerisms—like how he put his chips in quickly on the river, sometimes a sign of weakness—got Dwan from 24 percent to 29.

That was enough: Dwan took his time, coolly sipping from a bottle of water, and put the chips in to win a $3.1 million pot. If this kind of thought process seems alien to you—well, sorry, but your application to the River has been declined. (4267)

This all seems obvious and natural to me, if you can actually be this precise. I am skeptical of the precision in spots like this, the idea that poker players can ‘feel’ tiny differences, but I do think the good ones (like Dwan, who is clearly very good) end up doing an excellent approximation.

Blackjack is not worth your time. Yes, you can count cards, but don’t.

The new semi-blackjack roguelike game Dungeons and Degenerate Gamblers is okay, maybe Tier 3, but for superfans only. 21 is a roughly 3-star (out of 5) movie.

If you want to do blackjack card counting anyway, Nate does have some advice.

After about an hour of practice with a computer simulation that dealt six blackjack hands at a time at a medium-fast clip, I could get the count right about 95 percent of the time. (2330)

Their single-deck blackjack game—which on a slow night they’ll let you play for as little as $15—has a house edge of just 0.18 percent, assuming you play perfect basic strategy. (2342)

Pro tip: look for games where blackjack pays out 3:2 (so you win $75 if you make blackjack on a $50 bet) rather than 6:5 (so you’d only win $60). (2347)

In Professional Blackjack, the card counter Stanford Wong estimates that his benchmark strategy, executed perfectly under relatively favorable conditions, will net you about 60 cents for every $100 you bet—meaning a player edge of just 0.6 percent. (2418)

That’s all pretty miserable, even when it goes well, and the casinos will not take kindly if they figure out you are doing it. He goes into more detail, but there are better things to do with your time and life. I don’t know basic strategy and don’t plan to learn it, let alone how to do a count.

My basic advice for a casino is that there is a poker room and a sportsbook.

There also could be restaurants or a buffet, and hotel rooms, perhaps a show or a conference or a nearby sporting event. Sure, why not. Have fun.

But that’s it. Everything else you can gamble on is a trap.

Table games are a small mistake, if you want to buy entertainment. If you want to do pure gambling at a table game like roulette or craps, go ahead I suppose, although I don’t know why you would want to, unless you are starting UPS and can’t make payroll without getting lucky, or otherwise actually benefit from making one big highly precise gamble – the fee they charge on that is not so unreasonable.

Nate’s explanation is that you play craps to be around others, bond, tell stories, ‘prove your bravery’ and such. Story value is a thing, I suppose, but I find this highly alien.

Blackjack we covered already. If you want to pure gamble and play basic strategy, I mean sure, house edge is small, go for it if that’s what you really want. I guess.

And there’s nothing wrong with using that to also get some house perks. And perhaps you can even engage in a little ‘light counting’ by playing normally, except walking away from the table if the count goes very negative, maybe go get lunch then. There’s not much they can do about that.

Slots, on the other hand? Huge mistake. Treat them like harmful illegal drugs. Avoid.

Slots, alas, are the real business of the casino.

Nevada casinos make roughly fifty bucks in profit from slots for every dollar they get out of the poker games they spread. (2451)

Nate mentions a great book, Addicted by Design, about slot machines, how they are designed and the toll they extract on people. These are the worst kinds of engineered Skinner boxes, with many slots players so far gone they do not even want to win, merely to have a ‘smooth ride down.’

And then there’s the part that casino executives don’t like to talk about. To some percentage of patrons, slots can be highly addictive—potentially three to four times faster at addicting players than card games or sports betting.

Even though the percentage of players who become problem gamblers is relatively small, these players may account for 30 to 60 percent of slot revenues because they play so often. I’m about to tell you what might be the most shocking thing that I learned in the course of writing this book.

It may seem counterintuitive at first, but it helps to explain why slot machines can trigger such compulsive behavior. Here goes: according to Schüll, many of the problem gamblers she met didn’t actually want to win. “This was the thing that I couldn’t really get for a while,” Schüll told me. “But…I kept hearing it over and over again.”

Why wouldn’t a gambler want to win? Well, when you win a slot jackpot, it’s a disruptive experience. Lights flash. Alarms ring. The other patrons ooh and ah. A smiling attendant comes by to check your ID and hand you a tax form. (2890)

Then operators realized, even if you exclude the truly addicted, almost no one playing slots cares about the payouts. They wouldn’t notice being cheated even more.

We determined that for the individual to recognize the difference between a slot machine that had a hold of five percent, and one that had a hold of eight percent, the player would have to make forty thousand handle pulls on each machine.”

Forty thousand spins is a lot. (2706)

In 1997, the year before Loveman joined Caesars, slot machines on the Las Vegas Strip had an average hold percentage of 5.67 percent. By the time he left in 2015, it had jumped to 7.77 percent, right where those Atlantic City numbers had been. (2715)

The assumption here is that someone would play slots without knowing the odds, assuming the slots were reasonable, and maybe eventually notice the odds were off. And yeah, by that standard, it is going to be a long, long time before you confidently notice the slots are too tight, especially when you factor in jackpots. If the person is even paying attention at all.

The good news is the casino isn’t allowed to rob you even more in real time. Odds changes have to go through the NGCB. But yes, that is how it works. They flat out never tell you the odds. My lord.

And unlike in blackjack, you can’t just look up the odds—they aren’t listed anywhere! In fact, the same machine can have different payouts in different parts of the casino, without any outward sign to the customer.

Besides, who says the customer even cares? If you have to ask what the payout is, well, guess what? It’s less than 100%. And, as with lotteries, the money rolls in.

On the Las Vegas Strip, slot machines represent only 29 percent of overall (gambling plus nongambling) revenues. At off-Strip casinos, they’re 53 percent. (2857)

Are there some people who get genuine entertainment out of slots? Are there retirees who have a better time with slots than they would without them? There are some, to be sure.

There are even a handful of advantage players out there on the slots, because occasionally machines will have progressive jackpots or other tricks that can make them +EV in unusual circumstances.

Ever since, I’ve always kept an eye out for advantage players. They’re not that hard to spot. Their posture is more erect. They have a purposefulness that the typical slot-playing tourist lacks. And they’re finding $20 bills that aren’t supposed to exist. Seeking Action—or Escape? It’s not literally true that every human society has had gambling. But it’s been very common, dating (2831)

The odds of this being you are not high. And it doesn’t sound like a great gig.

I’ve thus become rather radicalized on slot machines. I do think we should be allowed to have casinos, including games I would never play like roulette, craps or baccarat. Slots are different. I think they should be Considered Harmful, and banned.

If that means (as it likely does) a radically downsized casino industry? Your offer is acceptable. If it also means fewer available poker games? I’ll accept that too.

I have also always been strongly against allowing mobile casino games. Essentially forcing everyone to carry around, on their person, access to unlimited immediate gambling complete with push notifications and entrapping offers is going to lead to quite a lot of ruined lives, and it is not a reasonable position in which to put those prone to gambling problems.

Recently, based on the results of our grand experiments, I’ve also greatly soured on allowing mobile sports betting. I love sports betting when ‘used responsibly,’ but this is very clearly not that. Studies are showing dramatic declines in savings and creditworthiness, across the board, in places with mobile sports betting. I have a draft post going into the details, and they’re pretty terrible.

Meanwhile, we are not offering even responsible players a good experience, instead charging them high prices, and the ads and incentives are bleeding into all things sports in toxic fashion. With all the fees they have to pay in advertising and to credit cards and to the state, the operators feel they don’t have a choice, I get that. But c’mon.

I was watching a baseball game two days ago, and during the broadcast they noted an ‘in game parlay’ of Mets to win and Over 7.5, at odds of +105 (so bet 100 to win 105). Highlighting in-game parlays is bad enough. But the crazy part was that the score was 4-1, the Mets were at home (so likely no bottom of the ninth) and it was the middle of the seventh inning. They wanted you to go Over 2.5 runs in 2 innings, and not even win if the Red Sox rallied and won.

I happen to have a spreadsheet for things like this. I looked up the odds on the game and plugged in the situation. Even if you assume the Mets always win, the fair odds were about +237 (you bet 100 to win 237, 29.7% odds) for Over 7.5 on its own, plus Boston wins 10%+ of the time that those runs score (they win 5% of the time, and they can’t win without there being at Over 7.5 runs). Even Over 6.5, on its own, would have been a (slightly) bad bet at +105.

Highway robbery. This must end.

At first Las Vegas was ruled by the mob. You could not trust it.

Then Nevada realized that people would only flock to Vegas if it was clear they could trust it. So they cleaned up their act, for real. They passed a law making this clear.

NRS 463.0129  Public policy of state concerning gaming; license or approval revocable privilege. The Legislature hereby finds, and declares to be the public policy of this state, that: The gaming industry is vitally important to the economy of the State and the general welfare of the inhabitants.

The continued growth and success of gaming is dependent upon public confidence and trust…and that gaming is free from criminal and corruptive elements. Public confidence and trust can only be maintained by strict regulation of all persons, locations, practices, associations and activities related to the operation of licensed gaming establishments. (2533)

Nate translates that as:

1. Gambling is vital to Nevada

2. Public trust is vital to gambling

3. Getting rid of the mob is vital to public trust.

Las Vegas may have an anything-goes, libertarian, frontier spirit. But without these regulations, it would be nothing like what it is today.

To be clear, you’re very unlikely to be cheated by the house in an American casino today—they’ll take your money fair and square. But that’s because of statutes like NRS 463.0129.

In fact, it’s actually in the industry’s best interest to have stringent regulation. Why? Because of the prisoner’s dilemma. If my casino, the Silver Spike, starts removing aces from its blackjack decks—boosting profits for my shareholders but in a way that its customers find hard to detect—the optimal strategy for the Gold Nugget down the block is to reciprocate. (2545)

I wish certain other industries, in particular AI, understood this. You think you want an ‘anything goes’ world. Consider the possibility that you are very wrong, and what would happen if events cause the public to further lose trust.

So now we have casinos in Nevada we can trust, and largely also around the world. The standard was set.

The second part of the story is that at first there weren’t huge expensive ‘experience’ casinos. Then Wynn dared to build one, it worked, and that established the business model. So now everyone does these audacious Vegas projects, up to and including The Sphere, because they can (on a regulatory and financial level) and a good time is had by all, resulting in a three-level market.

Really, there isn’t just one casino business model—there are three (2574):

1. There is the high-end luxury resort business.

2. Then there is the largest segment of the market by revenue, an upper-middlebrow market, dominated by major corporations like MGM and Caesars.

3. Finally, there is the “locals market,” focused on tourists on a tight budget, retirees, and working-class and middle-class people who visit casinos mostly to do one thing: play slots. Not blackjack, and certainly not poker. Maybe a decent steak dinner purchased with casino comps. But mostly just slots.

As noted above, I think the ‘locals’ business model of slots should die in a fire.

The other two could presumably survive without the slots, and there could be a place for them.

The casino business works because, despite most of its customers leaving with less money than they came with, those customers come back.

Disneyworld is famous as a business case study because 70 percent of its first-time visitors eventually return. Well, typically about 80 percent of visitors to Las Vegas are repeat customers. (2741)

Yes, but those are not comparable statistics. So quick nerdsnipe math break.

If 70% of your first time visitors eventually return, then at most 59% of your customers can be first time visitors (if no one ever comes back a third time), but you don’t know what percentage repeat. If you end up with Power Users who come back every year it could be quite a lot.

Similarly, if 80% of your customers are repeat customers, that could be because 10% of your customers get hooked and come back every year for an average of 10 years, while the majority never return. It doesn’t sound so implausible.

My guess would be that the chance of another visit goes steadily up every time you visit Disneyworld. So let’s say that 70% of first timers return, then 80% of second-timers return, then 85% from there until someone stops. In that case, for every first time visit, you get about 6 (technically 5.993) revisits. That’s very impressive, and an 86% rate, beating Las Vegas. If it’s 70% all the way, then it’s only 77% repeat visits.

Either way, yes, Las Vegas is doing Disneyland-levels of repeat business.

The recent history of all things gaming and gambling is the history of the whale.

The old business model of selling everyone a fixed price experience? Old, also busted.

Customer value follows a power law. Your median customer is nice, but they do not have that much money that they come looking to spend. The extreme right tail of the curve, the true ‘whales’ that spend tons of money, and getting the most out of them, are where all the profit lies.

You still cater to your other customers. Without them, the whales won’t show up. You need to build a reputation. And you need the others in the background, to form a contrast and an ecosystem and allow your amenities to exist and to fill out your games and so on. What’s the point of being the big spender without lording it over the others and getting that edge? That’s a lot of what is for sale.

This, together with the expectation of baseline play for free, has ruined most of mobile gaming, and a lot of other computer gaming as well. Your experience can and will be made super toxic, exactly so that whales pay up to make it less so. You will lose, so that those who spend will win.

Anyway, welcome to the casino business. The customer is always right, and that customer is your big spender.

Some customers are spending hundreds or thousands of times more than others. “We would tell our folks that if you lose a Diamond customer, because of poor service, you have to find twenty Gold customers to replace them,” said Loveman. (2749)

The perks of having high status in a casino rewards program are almost unlimited. Officially, Caesars Seven Stars members get free or deeply discounted rooms at almost any Caesars property worldwide, thousands of dollars in dining and travel credits, and even a complimentary trip on a cruise ship.

Unofficially, their status can take them even further. VIPs often have personal hosts or concierges and can negotiate changes to game rules, substantial rebates if they lose, free food and drink from anywhere in the property, nights out at strip clubs, and even private jets. (2751)

If you’re a loser and you know it, clap your hands. Let’s negotiate. There are many casinos. There is only one you. So if you demand perks worth – and by worth we mean what they cost the casino, not what they’d cost for you to buy – half your expected losses, they should happily give them to you, if they think otherwise you would walk. Or if it means you play twice as often, or twice as high.

I remember an episode of the show Las Vegas where a big supposed whale comes to the casino, and gets the big expensive room and all the perks. But he doesn’t gamble. The host tells him his treatment comes with an expectation of play. He asks how much, puts that amount on the roulette wheel once, wins (of course, it’s fiction, and it’s enough that the casino is sweating that it might happen several more times), smiles and moves on.

Being a whale on purpose is not obviously a bad strategy, if you have the money to spend and enjoy the actual gambling parts or simply have TMM (too much money). You are buying the ultimate concierge service. The casino will go the extra mile to cater to you and your wishes and whims, and handle lots of super annoying logistics and other details for you. Claude estimated you can get maybe 20%-30% of your losses back in terms of casino costs. That doesn’t sound like much, but the markups are huge, and the connections and hookups are kind of free.

I have some experience with a whale coming in and being, from a net profits perspective, the main thing that matters for a while, being worth more than the rest of your business combined. It’s super weird, especially since you don’t know what is motivating the whale, how far you can push them, or what would cause the ride to end. Happily I was not in the perks department.

You think this is all unfair? Well, tough.

The notion of a customer who’s already paying for a luxury experience getting cut in line by someone who has even higher status strikes him the wrong way. When Wynn took his objections to Loveman, they agreed to disagree.

“I said to Gary, ‘Doesn’t that create an animosity? When these preferences are in full view? Doesn’t it make them a second-class citizen?’…And [Loveman] said, ‘On the contrary, Steve, it makes them aspirational.’

Loveman was probably right. (2767)

My hunch is this cuts both ways. It is absolutely aspirational, and it can also piss the mid-tier customer off if it becomes a practical issue. No one likes getting bumped. You are definitely lowering the quality of my product, and I will respond accordingly.

As I mentioned earlier, Trump is the clear indication that you can be in politics, and you can be doing River-related businesses like casinos, and still very clearly not have either the Village Nature or River Nature. Indeed, his abject failure to get along with either group is necessary for how the story played out.

If Trump is despised by the Village, he’s also not a member of the River. Yes, Trump might be competitive and risk-taking—and he can even have a contrarian streak, having correctly gambled in 2016 that he could repudiate John McCain, Mitt Romney, George W. Bush, and the rest of the Republican establishment and still win the party’s nomination.

But that alone doesn’t a Riverian make. He’s shown little of the capacity for abstract, analytical reasoning that distinguishes people in the River from those who undertake high-stakes but miscalculated −EV bets. (2642)

Trump at least was (he has lost some number of steps by now) very, very good at certain things. Those things proved highly valuable in 2016, especially in the primary, and previously in his television career.

They did not, however, serve him well when running a casino.

At properties like the Wynn in Las Vegas, things just tend to go smoothly most of the time. At the Trump Taj Mahal, they did not. Within a week of the opening, its slot machines mysteriously shut down. (2668)

Trump also borrowed money at extremely high interest rates to finance his casinos. Doing the math would have made it clear he was unlikely to be able to keep pace with the costs involved. And that is indeed what happened.

The other problem was that Atlantic City failed to deliver the goods.

Atlantic City also proved to be a poor bet. Its gambling revenues actually exceeded those in Las Vegas throughout most of the 1980s and early 1990s but then declined by more than half.

Why did Vegas prove to be so much more durable? Perhaps because Atlantic City bet on gambling, gambling, and more gambling rather than offering guests a well-rounded entertainment experience. (About 75 percent of the Taj’s gross revenues in 1990 were from the gaming floor.) It also has poor vibes.

Las Vegas presents you with the feeling of freedom—casinos spill into one another, the Strip is walkable, and the weather is pleasant for much of the year. AC is more of a walled city, with casinos as self-contained fortresses in a hollowed-out, high-crime city. (2673)

I’ve been to Las Vegas a number of times. It is not my kind of place, but I understand it and appreciate it. I appreciate its honesty and shamelessness, its willingness to come out and be itself and fully tacky and what it is, and say this is here for you if you want it. And to provide, in many ways, a high quality product that makes you feel like you’re enjoying aspects of the good life. The strip is fun, and yes I have jaywalked it.

The one time I went to Atlantic City, for a Magic Grand Prix, and had a chance to go out into the city a bit? Things were bleak. And by bleak I mean post-apocalyptic. It felt deserted, as dust filled the air. Finally I made my way to a casino and attempted to play a little poker. It wasn’t a good experience on any level, and I quickly left.

Most casinos are horrible places, by design.

The canonical book on casino design—Bill Friedman’s 629-page Designing Casinos to Dominate the Competition, published in 2000—laid out a series of principles that aggressively invert almost every tenet of modern architectural theory, advocating for a seemingly unpleasant environment for the patron.

“LOW CEILINGS beat HIGH CEILINGS,” reads one. “A COMPACT AND CONGESTED GAMBLING-EQUIPMENT LAYOUT beats A VACANT AND SPACIOUS FLOOR LAYOUT,” is another.

Friedman’s ideal is a mazelike layout where there are slot machines and nothing but slot machines as far as the eye can see. The luxury end of the market, thankfully, has eschewed some of these ideas. (2851)

The default way of designing a casino, described above, is deeply hostile and adversarial. Once they get you in, it aims to disorient and trap you, where you lose track of time and space and reason and end up playing terrible slot machines.

The good news is that the luxury casinos do not do this. Once a customer is in the door, you’ll make more money on a given trip with the hostile approach, but people do notice that you did that. Word gets around. You’re not going to get and keep the luxury end of the market with such tricks. They might disregard them if the whale is off in their own little world, but it’s a turn off.

Until we get to rationality and AI and existential risk, and plausibly even including those topics, sports betting is the area of the book I know best.

I’m going to mostly skip over explaining much of the lingo and terminology and basic concepts, because we live in the age of AI. If you don’t know what something means, and you don’t have Nate’s glossary to help? Ask Claude, or ChatGPT, or Gemini.

(In very brief: a line (betting odds) of +120 means risk $100 to win $120. -110 means risk $110 to win $100. If it’s -3.5 -110, that means you also have to win by at least four points, and +3.5 -110 would mean if you don’t lose by at least four points then you win. Over 7.5 means you win if 8 or more points are scored, Under 7.5 means the opposite. Sharp players are good at this and a threat to win, square players aren’t.)

We start from the perspective of the gambler. How do you beat the line (i.e. win)?

You have to find good bets or promotions. Then be allowed to take advantage.

It is traditional, and Nate echoes this, for the top sports gamblers to talk about ‘be allowed to take advantage’ as the hard part, and ‘find good bets’ as the (relatively) easy part.

Beating the lines on paper is only half the battle—and it’s the easier half. “It’s not trivial to find edges,” said Ed Miller, the author of The Logic of Sports Betting. “But it’s much harder, once you’ve found edges, to find people willing to bet real money with you for an extended period.” (3071)

I think that’s an oversimplification. Both halves are very hard, and there is a tradeoff between them. A lot of the way that you get real money down, for size, over an extended period, is that you are forced to ‘move up’ to betting into bigger events and more accurate and worse odds, with various edges falling away. And you need to survive doing things like giving action to get action, or disguising your strategies, facing dramatic adverse selection problems, and so on. And your expenses go up, so you need to win more to pay them. You thus need to be extra good at finding edges, to overcome the headwinds.

Yes, ‘good enough to beat the odds betting $100 at a time’ is ‘the easy part’ in relative terms. The ability to handicap well enough to face all those headwinds, to ‘pay all the taxes’ in various senses, however, is very much no longer the easy part.

There are several core strategies people use to find good bets.

In Spanky’s world, there are two types of sports bettors, bottom-up and top-down: Bottom-up bettors seek to handicap games from the ground up “using data statistics analytics models etc.” In other words, the Moneyball approach: hope your superior modeling skills prevail against less sophisticated methods and stale conventional wisdom.

Top-down—the approach he prefers—“assumes the line is correct” and there aren’t a lot of gains to be had from modeling. But bettors can derive value via arbitrage, “information not reflected in the line such as injuries,” and through clever betting tactics. (3051)

The most successful sports bettors use a mix of both approaches. (3058)

By default, you should assume the line is correct unless you have a damn good reason to think otherwise.

It is important to be precise here on what it means to ‘assume the line is correct.’

The line here means:

1. The consensus market line, especially those that have moved recently.

2. Offered to the sharpest (best) players.

3. At the places that accept the biggest bets most freely.

4. For the lines on which the biggest wagers are permitted.

What does it mean to assume it is correct?

5. That given all the public information available…

6. …and subject to the known biases of the market…

7. …and all else being equal, mostly including ‘your attempt to handicap the game’…

8. …the true odds are within the bid-ask spread of the typical market (the vig).

9. So if you bet on either side of that, on a typical offering, it is -EV, and you lose.

Getting back to Spanky’s taxonomy, he divides strategies into top-down, where you assume the line is correct, and bottom-up where you attempt to figure out true odds.

I strongly agree that you need to combine both methods. The market is very smart, you need to know what it thinks, and ensure you get the best possible price. You also benefit from having insight into what everything means, and what edges you might be missing, and being able to spot traps and unusual situations and so on. The more (good) edges and angles you can combine, the better your chances.

Both bottom-up and top-down also have distinct subcategories. If you are going top-town, the main ones would be:

1. Compare prices at different sportsbooks.

2. Compare prices over time and take advantage of market movements.

3. Compare prices on different lines that reflect the same underlying questions.

4. Look for sources of systematic bias in the line.

5. Look for information the line isn’t taking into account.

6. Determine where the sharp action is versus the square action, side with sharp.

Everyone betting on sports needs to always be comparing prices across books. If you don’t make sure you are getting at least a decent price, you lose.

If you don’t check current prices for this, you lose. It’s like the stock market – the price now is the market price, and the price an hour ago is old news and by default means nothing. The way you measure your success in getting good prices is ‘closing line value,’ how good your best are if you assume the closing lines are accurate. While there are exceptions, players who look good on this metric are almost always good. And players who look random or bad on this metric are almost always bad.

Remember that not all sportsbooks are created equal, and their odds should not carry equal weight. The more they cater to a professional player base and allow skilled players to bet, and the higher size wagers they accept, and the better their pricing (the less vig they charge) the more respect you should give. Also give more respect to the ‘line in motion.’ Every change to the odds is for a reason, whereas a static (‘stale’) line might be laziness or not knowing how opinions are changing.

Combining that with taking advantage of market movements (‘chasing steam’) is a great way to get better prices, and also a great way to get thrown out.

You can also get closing line value through a practice called “steam chasing,” though sportsbooks really hate bettors who do it (3651).

They hate it because you are effectively taking out of their pocket. It also is a clear sign of intelligence and a pulse, which most of them hate even more. They might quickly throw you out for it.

So you have to walk a fine line. Essentially, you need to line shop, but you want to line shop among lines that are offered ‘on purpose,’ where the book doesn’t instantly feel stupid when they see the wager.

Comparing prices on different lines was one of my specialties. When we say ‘the line’ is correct, we’re mostly talking about two variables (three in soccer, where tendency to end in a draw is a free variable).

You can look at this as either:

1. You have how much one team is advantaged over the other (odds of winning: either point spread or moneyline, depending on the sport) and how high or low scoring the game will be (total).

2. You have how often the home team scores, and how often the away team scores.

If you know either set of numbers, and you know how to transpose the numbers properly to reflect sport details, then you also know the other set, and you know all sorts of other things as well.

Often, there will be odds that do not reflect the same reality as the line we are assuming is correct. The moneyline won’t reflect the spread and total, or the spread won’t reflect the moneyline. The total for one of the teams, or the odds on the first half, won’t match up correctly. Or sometimes it is something more obscure.

If you know all the math involved – and you mostly have to figure that math out yourself – then you can at minimum ‘line shop’ these different options, and find the best way to reflect ‘your opinion,’ the thing you think the market in general, or a given market, is getting wrong. Sometimes you can flat out find a line that has value even if you assume the main lines are correct, and take advantage – but also ask whether there is some reason you are given this particular opportunity, and whether your math could be wrong.

Over time, I realized that the vast majority of supposed ‘good reasons’ to deviate from the standard math matter either vanishingly little, or not at all. There are notably rare exceptions, but they are sufficiently rare that it is fine to assume they don’t matter.

Some of these errors, and some other errors, are systematic. The market gets something predictably wrong. These errors have been shrinking over time. Twenty years ago a lot of the mistakes were beyond blatant if you knew basic statistical principles. Others were not as obvious until you knew, but gaping holes once you did know. Now it’s all less obvious, and the edges are smaller.

If you have a reliable read on where the sharp money is, that’s great too. Normally on its own it is not enough, although big events like the Super Bowl or World Cup are exceptions because of the balance of power there as will be covered later. But it is highly useful to, say, avoid any actively ‘public’ sides almost no matter what. Compound your edges.

What about looking for angles? Is that a secret third thing? My take is no, also what Voulgaris describes here is barely an angle, it’s just books being awful at math:

Voulgaris has a style of betting that doesn’t fit neatly into Spanky’s top-down/bottom-up paradigm. He’s mostly looking for angles—essentially, diamonds in the rough like Castellón, highly profitable gambling opportunities that for some reason have been overlooked by the market. His most famous angle was betting totals (the combined number of points scored by both teams) in NBA games. A typical NBA total these days is 220 points.

But sportsbooks, in their never-ending quest to take as much action as possible, will also let you bet on how many points will be scored in each half. This seems straightforward if you’re a bookmaker—if the total is 220 points for the game, just divide by two and make it 110 for each half, right? Well, no. First halves are typically higher scoring by 3 points or so; the players are better rested and the defense tends to be less vigorous.

Only bookmakers did not realize this for many years, during which time Voulgaris could win prodigiously by betting overs in the first half and unders in the second. (3428)

The first half thing is a straightforward derivative play. That’s #3 on our list.

Also, can you believe the sportsbooks didn’t notice? And the other gamblers didn’t either? No one bothered to check avg(1H-Total) vs. avg(2H-Total)? It’s not as everyone assumes they have to be equal, either. In college basketball (due to end of game fouling plus overtime) the second half total is about 8 higher than the first half. So this is a hell of a thing for the world not to notice for many years.

Then again, there was a period where a game could have a total of over/under 9 runs, and you could go over/under 4.5 runs for each team at even odds. If you know even a little about distributions (e.g. that no team can score less than 0 runs) you can figure out why that is absurd.

It’s tougher out there these days. But given how absurdly not tough it used to be, there are probably still some pretty dumb things out there. It’s up to you to find them.

Then there is the tricky part where you combine that with your sports knowledge. A little sports knowledge is both highly valuable and a dangerous thing. You need to understand what things mean, and be able to recognize what is going on, without thinking you know more than you do. The worst is when you think you can handicap, unless of course you actually can pull it off.

Putting it all together and making it work requires a confluence of different skills.

The game involves three distinct skills, Miller told me, and “the number of people who check all three boxes is vanishingly small.”

1. There’s betting knowledge, “understanding markets and trading and counterparty risk,” as Miller put it.

2. There’s analytical skills—the ability to test statistical hypotheses and build models.

3. There’s sports domain knowledge—you’re not going to do a very good job of betting on a sport that you’ve never watched before.

4. There’s also a fourth area that becomes increasingly important as you scale up the size of your ambitions: networking skills. The sharpest sports bettors I’ve met aren’t necessarily super extroverted, but they aren’t lone wolves, either—they’re the kind of guys who know a lot of guys. (3402)

It is very easy to have two out of three, and to blow up due to not having the third. Which is the ‘easy’ part depends on the operation. Ultimately, the place my attempts to gamble ran into trouble was networking, which meant that I did well but wasn’t able to scale, until I did make the right connection, but that’s another story.

Can you watch sports and know ball, such that you can see something and then bet on something, and profit?

Not as often as you think, but occasionally the answer will be yes.

He noticed the famously fastidious Federer getting tipsy, something he thought was out of character and suggested a player who wasn’t in the right headspace.

This was at least potentially a useful angle—information that the general public didn’t have access to (as far as I can tell, there was never anything reported about Federer’s night out in the press).

He generally bet against Federer for the rest of the year, who indeed went into a slump, failing to make it past the quarterfinals in his next three Grand Slam events. (3453)

Tennis is a mental game, so certainly such an edge is possible. It also could have been chance, or mostly chance. Without a large sample of such claims, it is hard to say.

It does fit the pattern of the most common form of noticing something.

It is difficult to know for sure that someone is going to be great. It is much easier, and more common, to realize that Something is Wrong.

For personal reasons, I will always think of this as Limatime. Story time.

Once, there was a pitcher named Jose Lima. At first he pitched for the Pirates. He was a very good pitcher. Then he joined the Mets. At that point, he was a mediocre pitcher.

Then he lost the ability to pitch at the major league level.

He could still throw strikes. But his stuff was clearly gone. Every ball was hit hard. You can get lucky on any given at-bat anyway, but the chances of making it out of the 5th inning are… not good.

Normally, if a pitcher has a bad day or two, it shouldn’t move the odds much. This was different. So Limatime was declared. We bet very aggressively on the other team (the Marlins) and the Over. Lima got hammered. We did it again. Lima got hammered. We kept this up until the Mets figured it out.

Years later, there was another chance. Chin-Ming Wang, a Yankees pitcher who never got anyone to strike out but kept the ball on the ground, was coming back from an injury and lost the ability to keep the ball on the ground. Everyone was hitting the ball hard now. That won’t work. I got to declare Wangtime, and again collect several wins before the Yankees realized he was hurt and sent him back down.

So yes, it can be done. Perhaps it was Rogertime here. Perhaps it was not.

Even if you know what to bet on, and are willing to bet on it, you need to get down.

That starts with actually being willing to place the bet.

Most people, in most places, do not get that far.

“If your model is so smart, why don’t you bet on it?” is a common refrain in the River. Sometimes there are good reasons not to.

As I hope you’ll see from this chapter, it is far from trivial to get your money down even if you theoretically have a profitable bet. And betting sports—or almost anything else—requires a tolerance for financial swings that isn’t for everyone.

But for the most part, I endorse the sentiment. In recent years, researchers have discovered that a large share of experimental results published in academic journals—the majority of results in some fields—can’t be verified when other researchers try to duplicate them. (This is called the replication crisis.) (3186)

Nate Silver frequently will challenge people to bets, often reasonably large ones, when those people talk Obvious Nonsense. It is super refreshing.

The stuff that people are willing to put their money behind is usually going to be better. At the very least, a bet helps to align incentives.

“A bet is a tax on bullshit,” the economist Alex Tabarrok wrote in a post that defended me after I got in trouble at The New York Times for challenging the TV pundit Joe Scarborough to a bet on the outcome of the 2012 election. (3194)

For sports, people are far more often willing to bet. The question is how to get a bet down, for real size, at a reasonable price.

The best known master of the art of getting down is Billy Walters.

Billy Walters, widely regarded as the best sports bettor of all time, has long been a magnet for high-stakes company. (3464)

That company was partly people who figured out what to bet, and partly those who helped with the how. That high stakes company is liable to include one of your whales, or various other accounts, or other methods of getting a wager down.

Which is convenient when you have completely absurd edges.

It was a profitable relationship: Kent’s records showed that the Computer Group beat the spread as much as 60 percent of the time at college football, a success rate so high as to be nearly impossible today. Walters, meanwhile, played a role more analogous to that of a hyperconnected Spanky Kyrollos. His job was to get as much money down as possible for the Computer Group in as many places as possible (3481)

I totally believe that their stuff was this good. Or rather, that their stuff was only somewhat less good than the ‘standard sharp’ stuff is now, whereas the books back then were absolutely terrible. So yeah, if you pick your spots, 60% was doable.

It doesn’t mean you should do it overall. If you win at a 60% clip, then you are missing opportunities where you would otherwise win at a 55% clip, or a 52% clip. By betting more games, where your edges are smaller, you do several important things.

1. You disguise your best plays. Now, when you wager, they don’t know if you have a huge edge, if you have a small edge, or if you’re betting at ~zero EV. That means the books can’t afford to move as aggressively when you wager.

2. You give better action. When you bet big in your 60% games, you’re going to mostly move the line a lot, and the books get stuck with the bill. On your 52% games, that probably won’t happen, and the books might make a profit. That can help make up for things.

It’s important, at that level, to work with the book too, not only against the book. Yes, you’re playing against each other, but the book being able to make money dealing high limits is how you ultimately profit, and also how you get them to (at least somewhat) put up with your unprofitable business. As long as the book treats you fairly, you treat the book fairly too, to grow the pie.

Billy Walters was not down with that. Billy Walters was so not down for this that it changed how you had to handle the odds, because he also was very good at picking winners for size and ultimately moved the lines a lot. Anyone, at any time, could ‘turn into’ Billy Walters. Or then turn back into not Billy Walters, only to perhaps flip again. He was also fully capable of lying, betting the wrong side to try and provoke a move. And he seemed uninterested in reaching any sort of understanding.

How was he so good at the picking winners part?

At first he had the Computer Group, but then he kept things going from there.

What Walters preaches more than anything else—apart from the value of hard work—is the importance of seeking out consensus.

Domain knowledge? Betting knowledge? Analytical skills? He’ll take all of the above, thank you very much. Even in his late seventies, Walters and his partners were “experimenting with deep learning algorithms” and “taking a look at random forests,” he told me—some of the same machine learning techniques that are used to power AI systems like ChatGPT. (3498)

If you are good, you are 100% going to do this, what my gambling partner nicknamed the ‘Metroid Prime’ strategy of uniting all the angles. They compound and complement each other. And yes, if you are gambling in 2024 and you’re not trying to use AI and machine learning, well why the hell not?

Then we have more of the pattern of Walters as operator, keeping everyone else in the dark, going for every edge he could, treating others as rivals and future enemies. Which, given his experiences, was not completely unjustified.

Although many of the bettors I spoke with for this book use multiple sources—for example, averaging together two or more models—Walters takes one additional step. His sources only talk to Walters, not to one another. (3507)

Billy Walters was never a ‘good enough’ type of guy. He was always a maximizing, go for the throat type of guy.

He wants all the action, and he’s good enough that he’s willing to wait for game day, and willing to bend and break the rules, for the highest limits.

“I’m not interested in betting ten, twenty, thirty, forty, fifty thousand dollars. I’m not remotely interested,” Walters told me. (3625)

The “thing that separated me from the rest of these guys is ninety-five percent of all the betting I did, I did on the day of the game,” said Walters. “Once you get to the day of the game, you can bet a lot of money. Tell me somebody who can beat this stuff consistently. I don’t want to sound braggadocious, but I’m the only guy I know.” (3628)

He’s not the only guy. Or at minimum, he didn’t used to be. I’m rather confident Billy Walters knows this. I’m not going to blow their cover here by naming them, but I can name at least three unrelated others that, at least at one time, bet for size on game day, and were clearly going to be long term winners doing that.

Billy Walters might be the only one who can also consistently find enough different accounts and beards and other ways to get down the large amounts in places that want no part of his action, where the ‘you can bet a lot of money’ means him in particular, but that’s a different story.

How does Billy Walters, or how might you, pull those tricks off to get a lot down?

This is why it’s tricky to give you ironclad advice on how to win at betting sports.

The very practices that are the most profitable—the ones that most reliably get you closing line value—are also the ones that sportsbooks are quickest to limit you for.

Bettors—particularly top-down bettors like Spanky who engage in arbitrage tactics like steam chasing—are therefore constantly engaged in acts of subterfuge.

Most of these tactics fall into one of two categories: you can make dumb bets with accounts the sportsbook thinks are sharp, or sharp bets with accounts they think are dumb. (3658)

Billy was definitely willing to do the ‘dumb bets with sharp accounts’ trick on occasion. If he felt the books were moving too far when he wagered, not letting him get the size he wanted, he’d throw in an actively wrong side, keeping them on their toes. Unfortunately for him, the threat was stronger than its execution. Once he actually did it, this often backfired, because that meant that once he touched a game you had to be paranoid for a while, so you basically didn’t let anyone wager much of anything. In the extreme this could mean lower limits all the time, for every game.

Billy’s favorite tactic seemed to be the beard, taking over a VIP’s account. Billy is the master of this, but others play the game too.

So Spanky and other bettors are constantly on the hunt for the same thing that poker players are after—whales, that is rich guys or degens with a credible history of betting big who will get a lot of rope before they’re outed as a beard. “I had an article published about me in Cigar Aficionado. That’s a very good publication that whales read,” Spanky told me.

Other bettors cultivate relationships with whales by playing poker, or just by living in the gambling high life in what I call the “splash zone.” Voulgaris once even used the boxer Floyd Mayweather as a beard, he said, “but like only for like a day, or two days, because he was so difficult to work with.” (3676)

Giving different people different limits is very much not Game Theory Optimal. There is a natural deal to be made, so it comes with the territory that the deal will get made every so often.

Information may or may not want to be free—but if a whale can bet $1 million on a game at DraftKings and a sharp bettor can only get down a few bucks, there are obvious incentives for information to overcome whatever friction there is and to flow from the sharp bettor to the whale. (3700)

That’s why the sportsbooks have a very important department or job, to watch every whale wager and ask, quite literally (and yes these exact words are used often), ‘was that Billy Walters?’

It is not easy to use a whale well. Every whale is going to have a pattern of betting. Sharp plays follow a very different pattern. Many sharp bets look hella suspicious, even in isolation. When you combine the side, price, timing, betting history and other details, even one bet is often very strong Bayesian evidence. Three is usually enough. You obviously are terrified of a false positive that will lose you an excellent customer, but you are often very quickly very confident that you are right.

Once the whale is known to be compromised, the situation is tricky. You don’t want to permanently lose the good customer, who might soon go back to their old self (see the Mayweather example above, and also these whales and sharps get into disputes all the time). Sometimes what you’re seeing is kind of a temper tantrum, someone lost a little too much and hooks up with a sharp for a bit. Or perhaps you can talk to the whale and convince them to cut it out rather than lose their status, now that you know.

You can also use the situation as a bit of a double agent scenario. Better the compromised whale account that you know about, then the compromised whale account that you don’t know about, especially if that would lose you another whale. So perhaps you don’t want to even let on that you know the whale is compromised, and you get to buy the information instead, even if you would rather not have to do that.

Being the house in sports betting is not like being the house elsewhere.

In all the other games, unless someone is cheating or at least counting cards, you cannot lose. Poker players take money from each other. Table games players are supposed to never have an edge. Slot machines provide steady income.

Sports is different.

It’s not so easy because sports betting is an adversarial game—the bettors can fight back. (3131)

Sports betting really is a skill game pitting the house against the player. The house gets to charge a fee, and can and often does book action on both sides and lock in a profit, but the players get to pick their spots. A sportsbook has to play defense in everything, everywhere, all at once.

If you make a mistake, they will pounce. The house can absolutely lose, and often does, even before expenses.

You would think there is some sophisticated AI in charge of all those prices.

Remarkably often? You would be wrong.

Behind the curtains of the SuperBook’s massive wall of sports—it has 4,250 square feet of LED screens, giving it almost as much screen space as an IMAX—was a relatively humble operation. There was a team of perhaps ten people, mostly men in their twenties and thirties, in a narrow, dark room with a phalanx of monitors, manually reviewing bets from the Westgate’s mobile app and comparing the Westgate’s lines to every other sportsbook in the world.

Three-thousand line moves per shift. Kornegay clarified for me that these are the moves his traders make by hand. Others are made algorithmically, but the most important ones aren’t. “You would be surprised how unsophisticated the software is,” he told me. (3126)

Skill issue.

Another skill issue:

While playing in a WSOP event, I noticed that one of the Nevada mobile apps was about a half minute behind FanDuel in updating its live U.S. Open odds. It was as though I could see thirty seconds in the future. I was able to get enough money down that I was guaranteed a $5,000 profit whichever golfer won. (3155)

That really, really should never happen. Thirty seconds is an eternity.

For a long time, I assumed the vast majority of moves were made automatically in various ways, at least at the smarter sportsbooks. I had a detailed mental model of what bets likely caused what moves, and how the line moves told the story of the game.

Then I got a look behind the scenes. I was shocked, utterly shocked, how much of it was fully manual and human, and how unsophisticated were the automated tools.

I was also shocked how unsophisticated was the math behind the odds. People really were just winging it remarkably often, including the live odds at some of the world’s biggest sportsbooks, Asian books dealing mostly in soccer. It was unmistakable.

Over time, those automated tools are improving.

In many cases, the math I’d done in Excel and using MySQL, using basic statistics and logic, ended up wholesale becoming the basis for the world’s derivative prices.

In various sports, with programming help, I managed to not only derive good enough odds calculations as baselines, but then take the automated tool from scratch to the point where we could offer odds on something at all with careful terrified supervision, to the point where you could do it with supervision, to the point where you could do it while supervising 20+ games at once and mostly only watching for something highly abnormal or to tune the system for the future.

Yet here is Nate Silver, in the last few years, seeing the same old big board with thousands of manual moves a day and only highly unsophisticated software, using formulas I could likely write down via blind guesses.

That doesn’t mean there is no room for the human element. You want humans around to notice edge cases, and to notice when you need to do things that are crimes against game theory if anyone saw you doing them too systematically. For example:

Murray, for instance, recalled a time when a man they’d never seen before wanted to bet big—very big—on the Golden State Warriors to win the NBA Finals. “This guy asked for as much as we would let him go on the Warriors to win the title. And we gave him a bet. I don’t recall exactly, twenty or twenty-five thousand [dollars].”

After his initial bet, the Westgate aggressively moved the line to a less favorable price—but the customer wanted to bet again.

“We just looked at each other [and said], ‘Kevin Durant’s going to the Warriors.’ ” Nothing else could explain the unknown man’s confidence.

Durant indeed signed with Golden State, and the Warriors steamrolled the league en route to the title. (3162)

My first reaction was, they let him, a person they’d never seen bet before (and presumably isn’t otherwise a whale), get down $20k on the Warriors to win the NBA Finals while Kevin Durant was unsigned?

That seems like rather a lot. It is an odd fact about sports betting that one Warriors game will get more action at lower prices than the Warriors to win the NBA Finals, but that is because we have a much better understanding of the individual games and much more liquidity and willingness to bet on an outcome that finishes within a day.

I give kudos to the casino for letting him have that $20k, and for not treating his announcement that he had info – ‘go as big as you will let me’ – as an excuse to not let him bet hardly at all. That’s good market honor and customer service.

But yeah, that’s the kind of place you need a human.

The house can lose, and a skilled player can win.

Everyone involved on the sportsbook side frequently has myopic and short term incentives. Expenses required for online sportsbooks are remarkably high. Mostly, the odds are tuned to capture maximum profits from the foolish public, rather than trying to be accurate.

In that context, the house is not interested in letting you, the sharp gambler, win.

Robins, the DraftKings CEO, said as much in 2022—“We’re trying to get smart in eliminating the sharp action or limiting it at least,” he told a group of investors. (3322)

“They give the illusion of offering many [bets]. But if you actually try to bet many of these bets, for real money, you quickly run into friction,” said Miller. (3325)

That is what happened when Nate tried betting on the NBA, without trying to disguise what he was up to. Soon he was being limited at most books, despite his bets being a size and level of sophistication the books would have zero trouble managing.

The standard term for this is ‘retail’ sportsbook, or ‘recreational’ would also do.

Miller’s book calls them “retail” sportsbooks and “market makers.” The sportsbooks you see advertised on TV are retail books.

They put a lot of money toward customer acquisition; DraftKings spent almost $1.2 billion on sales and marketing in 2022, a gargantuan sum given that they did only $2.2 billion in revenue. And they extensively profile those customers.

If they think you’re a whale, you’ll get the sorts of perks that a casino VIP does. Four seats on the glass at a New York Rangers playoff game? A $5,000 bonus bet deposited into your account because you haven’t played in a while? A case of your favorite cabernet shipped to your house? A friend of mine who’s a DraftKings VIP gets all of this and more. (3331)

A $5,000 bonus bet means a bet where you lose nothing if you lose, so it is worth roughly $2,250 even if you play it fully straight, which you’ll do if you want to remain a VIP (you can get $4k+ of value out of this by taking a long odds bet but you’d give the game away). Those are nice perks. If you can actually get exactly the perks you want, it might even be worth thinking about – especially if you can find a way to be far less dumb than you look, even if you’re not actively being otherwise smart.

When the product for DraftKings and FanDuel was mostly DFS (Daily Fantasy Sports) players were betting against each other. So the house was happy to let you win, and they advertised to imply that the winner could be you.

Today’s sportsbook commercials don’t hit the same notes. They’ll advertise that sports betting is a lot of fun or that there are a lot of different ways to bet, or show celebrities or former athletes using their product. But they’re usually careful not to imply that sports betting is a skill game or that you can be a long-term winner—because they don’t want you to be one. In the sports betting that I’ve described thus far, you are betting against the house. But in 2015, these companies were advertising a different product called daily fantasy sports (DFS). (3525)

Whereas if you are a winner today, on classic sports wagers? DraftKings cannot wait to get rid of you, to the fullest extent allowable by law.

“If a lot of that money is going out the side door to sharps who don’t even enjoy your product and are completely platform-agnostic and have no loyalty at all—then like, why wouldn’t you do something to control that?” said Jon Aguiar, an executive at DraftKings. (3540)

As a former sharp, I resent the implication that I don’t enjoy the product. Well, sure, I don’t enjoy DraftKings in particular, because your odds are atrocious and you won’t let me wager. But can’t we all have fun?

And yeah, of course I will place the wager at the sportsbook with the best odds, but that doesn’t mean no loyalty. Knowing you will be treated fairly, they will let you bet and honor your wages even when they make mistakes (unless the mistake was really obvious and egregious, but different operators interpret that principle very differently), and that you have a place you can trust to pay you if you win. Where if something happens you have a good relationship.

Again, that wasn’t DraftKings. For me that was because it didn’t exist back then, but today it’s because they could not deserve your loyalty much less – they’re kicking you out, and I’ve seen stories of them trying to weasel out of paying on big parlays.

To answer the actual question: You let them walk out with a lot of money because doing so makes you even more, over the long term. It improves your product and your information and your reputation. It’s ‘good for the game.’

If you have a completely different business model, that might not be true. In particular, if your model of your VIP is ‘this guy is going to lose X$ each month no matter what in the end,’ or ‘this guy is going to bet without even bothering to check the odds’ then nothing the sharp can do will let you make more money there.

Fully bluntly: If your model is ‘I get Dave as a customer via advertising, and then Dave has a fixed LTV (lifetime value of a customer) and I never move the line when he wages because he’s an idiot’ then you don’t want to give Billy Walters a cut.

The other answer is, because winning is aspirational. If you stop pretending anyone is allowed to win, why are we all here?

Well, one answer to Aguiar’s question is that public statements like these handcuff DraftKings’ most powerful marketing tool—its appeal to the male ego. It’s entirely plausible that DraftKings is losing out on more action from guys who mistakenly think they can win than they’re saving by not taking sharp action. “It’s designed for you to lose money,” said Kelly Stewart, a.k.a. Kelly in Vegas, one of the most prominent women in the industry and also one of the people with the cheekiest no-BS attitude. (3542)

Of course it is designed for you to lose money. Can we be less obvious about that?

Alternatively, how about more obvious?

There are two main reasons that someone might make a −EV bet: first, because they find it entertaining, or second, because they think they’re making a good bet when they aren’t. Sportsbooks like DraftKings call themselves “entertainment products,” seeking to capitalize on the first type of customer. But they aren’t doing much to encourage the second type.

In fact, if a sportsbook known to aggressively limit winning players is taking your action, you should ask yourself why—they are basically telling you they think you’re a loser. (3554)

This is the concept of ‘trying to win.’

Some gamblers, some of the time, are trying to win. They are dialed in. They have discipline and pick their spots, and don’t splash around. They work to find the right side and then to get the best price.

That doesn’t mean that the player has what it takes to be +EV. Most players who try to win don’t have it. It still gives them a chance, and it means they lose a lot less.

DraftKings and friends essentially are saying, if you show enough signs you are trying to win, they do not care if you will actually win in the long run, or even whether you are winning. They don’t like your kind, and want to show you the door.

Other players are largely there to gamble, or to have fun, whether or not they admit that. They are not giving their best work. DraftKings calls them ‘good customers.’

Almost all the legal online sportsbooks do this.

A 2022 Washington Post article matched my experience, suggesting that DraftKings, BetMGM, and PointsBet are more aggressive about limiting players, and Caesars and WynnBET are less so. (I still have a clean bill of health at Caesars, although WynnBet limited me in March 2024.) That leaves FanDuel, the largest U.S. sportsbook by market share, in an in-between category. (3561)

The book widely considered the sharpest U.S. sportsbook, Circa, essentially takes the In-N-Out Burger approach to FanDuel’s McDonald’s, offering a narrow menu but doing every item well. (3580)

I’ve been out of the game long enough I no longer know the players so well, but Circa appears to be the closest thing to an exception among the licensed books (there are also places like Pinnacle offshore). Circa sticks to the basics, does its homework, gets better information and offers sharper lines. They even offer slightly lower prices (lower juice or vig) than other books. They don’t offer big discounts, but they’re the ones offering discounts at all (everyone else does promotions and VIP clubs instead), so they don’t have to.

FanDuel says it strikes a compromise.

Farren had said something that foreshadowed it when I’d asked him how FanDuel treats winning players. “You’ve had a bet to win a reasonable night, you know, a couple of grand or whatever,” he continued. “We get information, we straighten our prices. There’s something there—there’s a place for us all to live.” Essentially, FanDuel was offering sharp bettors a compromise: you can bet up to the amount that your information is worth. (3619)

Certainly that is way better than throwing those players out entirely. A sharp bet at a fraction of your usual limits is a great deal, especially when you have lots of customers.

Indeed, the question is not why FanDuel does it, it’s why the others are so foolish as to not do it, and also why the sharp player would sell their information cheap, although one good answer is ‘they don’t have a big enough bankroll to mind.’

DraftKings kicks out winners now, because you are betting directly against the house, and they don’t want a bunch of sharps walking out the door with a bunch of money.

Which raises the question: Why were they so stupid as to allow this in Daily Fantasy Sports, when the sharps were a much, much bigger problem?

Allow me to explain.

Billy Walters wagering on sports, even if you give him unlimited rebets at the standard maximum, can at worst force you to move his side to a slightly minus-EV price, or be more expensive than other options. At that point, he stops wagering.

If that is sufficient to get sufficient action on the other side, and you charge -110, you are going to make money, ultimately approaching 5%. If you can use this to book only action on the other side past that point, you can do better.

If you can quote one price to sharp betters like Billy (the ‘real price’), and another different price to smaller foolish square betters, or get them to bet parlays, you can do better still.

And you do all this while giving your square gamblers a better price on the square side. It makes you a better place to play.

DFS is not like that. In DFS, yes, today you make 10% no matter what, by design. So the instinct is to say, if you want to turn pro, great, more volume means more profit.

Except where does that money come from? The other players.

And what happens to the winners? They scale.

Phil Ivey, like Mike McDermott, knows he can beat the 10/20 at the Chesterfield. He would crush that game. But he also would never dream of playing in it, because it would be boring as hell, and even playing perfectly his hourly rate would suck.

A great poker player might play 2, 6 or in some cases 20 tables at once online, at some cost to their quality of play. But they can’t scale without limit or cost. There is a reason why bots have to banned online – otherwise the (mostly GTO?) bots would multiply without limit until the games were useless for anyone.

Whereas in DFS, you can enter as many times as you can click the buttons. The skilled labor is in figuring out the best lineups to play, and which events to enter. Then there are ways to automate the clicking of the buttons, either via humans or programs. Selecting the right events can mostly also be automated.

With money turning over every day, the winners quickly end up with large bankrolls. They then multiply with a GTO (game theory optimal) mix of lineups, until there are so many of them that the chance of facing each other makes this not so profitable.

So what happens then? We can simulate this as a pool of players, Alice (and those like Alice) and Bob (with those like Bob). Alice plays perfect lineups. Bob plays something worse. Alice is an advantage player, so let’s say she keeps entering until she is only winning 2% on average. Let’s say that when Alice plays another Alice or Bob plays another Bob they each win half the time, and when Alice plays a Bob she wins a 10% profit on average.

What happens to Bob, given the game costs $11 to enter and pays out $21 if you win?

Alice is a small majority of the entries, so half the time Bob loses 5%, and half the time Bob loses 20% (so that Alice can make her 10%). So now Bob is only getting back 88.5 cents per dollar wagered.

That’s a level of tight slots anyone can feel. And Bob keeps seeing the same handful of accounts as opponents, over and over, with the same optimized-style lineups. It’s frustrating, it’s unwinnable, and it’s boring. Even if Bob wants to get stronger and become an Alice, even if he works hard and does pretty well, the skill hill is huge.

Ideally Bob would be able to solve this with cheaper events. Sure, the $100 event is hard, that’s fair game, but the $10 game is almost the same after a while, as is the $1 game. You can’t win anywhere.

That’s not a good product.

Indeed, that’s part of why I only played bigger tournaments. In a big tournament you have to use mixed strategies, because your goal is to stand out, and to do that you need correlations (which others were painfully bad at) and to not pick the same players everyone else was using. Whereas in 1-on-1s? That’s pure maximizing the EV of the players, so you basically just have to be perfect at that to have any chance.

I think this is ultimately what happened to DFS, eventually resulting in a doom loop.

The standard theory is that the house will look to balance action, taking equal amounts on each side, to lock in a profit and guard against mistakes.

That theory is mostly wrong.

Not worried about balancing money—happy to take more action on one side, especially if it thinks the other side is sharp In principle might want to balance money on different sides of the line, though this is hard in practice. (3348)

If you are only a small part of the market, and your customers are not especially smart, then if you copy the odds listed elsewhere, or even if you lean a bit, you’re going to get unbalanced action, but the unbalancing will be if anything worse than random. Why would you want to balance out that action? Well, you might otherwise be playing higher than you want to risk, but most of the time that won’t be true.

There are times when you want to play it safe. But mostly you don’t need to. The average retail bookmarker, even on normal lines, makes more money per bet than if its customers bet at random. That’s core to the business model.

If you have a bunch of dumb customers, you want to take advantage of them. If you have a bunch of not so dumb customers, you want to use that information, to take advantage of the situation anyway.

Occasionally, one of these ‘recreational’ sportsbooks will decide to grow a spine. They might even do so in a way that isn’t smart, like having most of your action on the Super Bowl on the Eagles, and then moving the line the other direction anyway?

In fact, FanDuel was knowingly inviting action on the Eagles. Were they doing this to balance out their books? Nope. Just the opposite. Most of the public money was already on Philly. (3600)

If “all the star players scored [in a] high scoring game” and the Eagles won, FanDuel could potentially have been on the line for $400 to $500 million, Farren said. It was “not enterprise risk, not like, you know, we-can’t-pay-the-bills sort of risk” but “we kind of had to make sure we had the money there to pay for it.” (3609)

The book said they kept the price where they think is fair. But why would you do that, if the money is already so unbalanced? You can do it to stop people from betting the other way, but you’re already limiting the people who would do that, they’ve maxed out. Most public money is not that price sensitive. Do you need to be taking a shot here, even if it doesn’t rise to the level of enterprise risk? Wouldn’t you rather make a smaller bet at a much better price?

There are a few other houses that serve the role of ‘market makers,’ trying to figure out the odds on their own, including posting the first odds available.

But the retail books aren’t particularly good at bookmaking. That’s where the market makers come in. (3356)

How do they make up the line? They… make up the line.

How’d BOSS come up with Raptors −3.5? Some nerd in a boiler room made up a number. (3373)

It is remarkably easy to get a ‘reasonable’ number for an opener. Most gamblers develop a pretty good intuition for how odds work, and there are various power rankings and past lines to draw from. NBA is more dangerous than most other leagues because of injuries, but in most places you know what you’re getting and there was a time I could do them in 5 seconds in my head and it was fine. Indeed, I’d see NFL lines opened for 5k a pop after 30 seconds of two people sharing instincts, when those involved did not feel like waiting for the traditional openers, and it basically went fine.

Which is, indeed, exactly what Nate sees.

“We’re usually first to market on NFL lines,” Kornegay told me of how the Westgate sets its opening lines. “That process is very unsophisticated. It’s like, yeah, like, what do you make it? Five, three and a half, four—okay, give us four. And then sometimes we’ll have a little debate on it.” Basically, that nerd at BOSS was inviting me into the conversation: “Hmm—Raptors-Pellies, let’s go three and a half?”

“Oh no, the Raptors have looked pretty good since that trade, gotta be at least four and a half, five.”

“Okay, you seem to have a lot of conviction—let’s go with four and a half.”

“Yeah, I could push you on five versus four and a half, but that’s close enough.”

I’m getting paid to partake in this conversation—paid in expected value. But I’m not getting paid that well. (3375)

Those lines are therefore pretty easy to beat. All you have to do is be better than them, in the subset of places you decide to wager, and you get to pick your spots. For baseball, when I last ran my numbers a few years ago, my program even without any human adjustments was better than the opening lines so long as you had an eye for obvious situations where there was a big change the program was missing, and then you get to make human adjustments and also pick your spots.

But as Nate notes, the price of that is acceptable.

“The object of any bookmaker is to get to the closing line as early as possible, as fast as possible, for as cheap as possible,” Spanky told me. (3387)

Yes and no.

Suppose I had a vision of the future, and I knew the closing line, where the odds would finally end up (but not who wins from there). I would certainly want to move in the direction of the closing line. And if I actually knew 100% what it was, and I moved there too early, worse things have happened. But I can do better. If the line on the Raptors is -3.5 and I think it’s going to close at -5.5, I won’t move to -5.5 right away. I’ll move to -4 or -4.5, or (if my book allows it) maybe I’ll move to offer +3.5 +105 or +3.5 +110 (or +111, for a full arbitrage) on the other side. There’s no need to give anything away. Also you could be wrong.

If you ‘have a hunch,’ even a very strong one, your goal is to stay ahead of the move so you get action on the side you want action, at the best possible price. You don’t want to spook the entire market if you can help it.

What’s hard to beat is the market. Because at some point, Spanky is going to bet the game. Rufus Peabody is going to bet the game. Some fucking hedge fund in Dublin is going to bet the game. And if the bookmaking process is working properly, I have to beat those guys. (3396)

Again: Well, maybe.

You kind of have to beat those guys, in the sense that they have already impacted the line by then (or decided not to do that), but you also kind of don’t. You only have to beat the market price, which is catering to them and also a bunch of idiots, and you again get to pick your battles.

As with all trades, if I start betting where there is value, I am going to stop before I eat up all the value. The sportsbook might respond by moving farther than that anyway, but I’m not going to take what I think is a minimal-or-zero-EV bet. That’s especially true if it would get the sportsbook in more trouble on the game, putting them in a deeper hole, and drawing more attention to me and my future bets.

Or: If I often bet three times, then they’re going to end up charging more for my second bet, than if I usually stop at two. If every time they move 10 cents (e.g. from -110 to -120) I still bet again, they’ll always move at least that far. And so on.

There are exceptions, when someone is being a true degen on top of being good. And of course some gamblers are degens and also were never good to begin with. But in many cases, there will still be some give remaining, if you are selective, especially if you are attacking from a distinct angle. If you’re using the exact same system as Rufus and Spany, and you let them bet first, then you have a bigger problem.

Which brings us to ‘touts.’ Touts offer to sell you picks so they you can make money too. If you are ever considering using a tout, you can refer to this handy guide, created by Spraguer and posted to Twitter by (as it happens, I didn’t realize until I looked to put it into this review) none other than… Rufus Peabody:

Also see the rest of the linked thread.

Nate Silver tried his hand at sports betting.

He started off on a hot streak. But that was fueled not only by luck, but in large part by unsustainable tactics. Nate took advantage of various opportunities that made it easy to identify him as a sharp, so he got kicked out of a lot of places. You can only pull that trick once.

But on the regular, game-by-game bets that I devoted so much time to? Well, let’s be precise. I bet a total of $1,809,006. And I finished the year ahead by a whopping $5,242—for a paltry ROI of 0.3 percent. (3720)

I was surprised at how little it took for sportsbooks to limit me. We’ve covered this, but I wish I’d been more aware at the outset of how quickly you can get limited. I wasn’t really making much effort to cover my tracks. (3724)

I was surprised, in many cases, back in the day, how long it took many places to limit me, and how much I was able to get away with before that happened. Like Nate, I didn’t try to disguise what I was doing early on, and I’m not convinced that was a mistake on principle. Better to get what you can while the getting is good, because you don’t know what kind of attention they are paying and what will get you ‘caught.’

I was surprised at how often my early bets moved the lines. (3729)

Yep, early bets by default move lines, even if you don’t know who is doing it.

I was surprised by how fast-paced sports betting can be. Just like in poker, many sports-betting decisions involve incomplete information. You’ll see a line that looks favorable, and you’d like to do some due diligence on it—maybe there’s an injury that you weren’t aware of? But a good line may go away after even five or ten seconds. And if the line is still available, it may not be such a great bet after all. (3772)

Sports betting is like poker or other trading, in that it is long periods of not much happening interspersed with high stakes and high pressure moments of sheer terror. You need to be ready to pounce when opportunity knocks, in terms of opening lines, especially new live or halftime odds, or reacting to certain kinds of news. If you don’t, the adverse selection gets strong. There are also lots of bets you place after a long period of contemplation, and lots of time spent doing your homework.

I was surprised by the streakiness—even though I shouldn’t have been. (3738)

Yep, again, that’s largely how randomness works. But also there are often other things going on as well. I had some big runs both positive and negative along the way, and frequently there was a good reason. The biggest sustained systematic losses turned out not to be accidents.

I hadn’t realized how much injuries—and other situations where you can get a leg up through inside information—can dominate other concerns. (3777)

That’s an NBA thing. The NBA is the league of injuries and who is going to show up today and try to win versus who is resting or coasting or hurt. Other sports, including NCAA Basketball, are not like that, although football has some of it. You can mostly ignore injury news in for example NCAAB or MLB, adjust only for the big news you’d hear about anyway, and be fine.

Sports betting took more mental bandwidth than I expected, even when I wasn’t “on the clock.” (3786)

Real. Also a skill issue. The book lists a lot of necessary skills, and I’d add to that: Learning how to not watch the scoreboard or worry about or even check on the outcome, and set things aside. I was not always able to pull this off, but I was pretty good, and I got better over time.

Sports betting took more emotional bandwidth than I expected. I don’t think I went on tilt or became addicted to sports betting at any point—but sometimes it’s hard to know. (3793)

I most definitely did go on tilt at various times. The good news was I (mostly?) responded to that by taking a break, rather than chasing losses, doing heat checks or getting sloppy. Poker puts you in situations where you are on tilt, but you don’t want to pass up opportunity. Sports certainly gives you opportunities you don’t want to miss, like NFL sundays, but it’s not like when you have that amazing poker table.

Part of the problem with sports betting is that casinos think of it as an amenity.

“We watched sports wagering for many years. We had it to ourselves. We consider it in many cases an amenity.” An “amenity” means something casinos provide for their patrons to meet customers’ expectations. Not necessarily a loss leader, but also not a profit center. (3302)

But sportsbooks and racebooks make up only about 2 percent of gaming revenues at Las Vegas Strip casinos and 1 percent of overall revenues. They can be a bigger part of the business at off-Strip properties like Westgate and Circa, where sportsbooks help to draw in huge crowds on football weekends. (3308)

Casinos in the state of Colorado lost nearly $11 million in NBA betting to their patrons in June 2023. Why? It’s because the Denver Nuggets won the NBA championship that month. (3316)

Certainly it makes sense, as an amenity, to offer sports for bigger sizes, better prices and more variety than would maximize sportsbook profits. Perhaps you wouldn’t even waste the space otherwise. And I’m grateful that they’re at least offering it at all.

However, you know the way to offer the best sportsbook? Actually offer a high quality sportsbook, with high limits and good prices and good service and so on. Attract more customers and a better crowd. Make it a plus amenity, not a box you check.

Also, if you do it right, you can win a lot of money that way.

Several times since I left that world, I have toyed with the idea of creating my own sportsbook, or trying to get hired to run one of the major ones.

In particular, I wanted to build the heir to the old school form of Pinnacle Sports, except with full American licenses and ideally allied with a major Vegas casino group.

Instead of being scared of winners and our own shadow, as almost every book is now, we would once again welcome the professionals as allies. We’d even make lucrative deals with some of the winners, to ensure they played it straight and we got to use their information.

Rather than limiting everyone we didn’t know was a sucker, we would push limits for almost everyone, as often as possible, even in spots where we might be underdogs for a while, until we built up a marketplace. If you were a whale? Modulo some safeguards, the sky’s the limit.

Rather than charging -110 or worse across the board, we would charge low, low prices across the board. At maximum we’d charge -105 for the major sports, and push from there.

We’d have a page on your website and app where you’d select what channel you were watching, and every time there were gambling odds on the screen, you’d see what the odds on TV were, next to our lower – and often opinionated – prices.

Behind the curtain, we’d do extensive automated and manual analysis on everything, and git gud. We’d have our own models on every level. We’d usually know what the best gamblers in the world were going to do before they did it, and often barely adjust our odds when they did it.

Alas, that dream never happened, beyond consulting for a few projects, where I do think I did excellent and valuable work.

While there is a hugely profitable end state that depends on doing things like this, that requires getting there, and that means bringing in customers and doing all sorts of legal and marketing and other work, work I am not good at nor do I want to do. And until it scaled big enough, the plan wouldn’t work.

Once it did scale big enough, I would expect it to snowball, as word got out and your product kept improving. But you have to get there.

Also I learned that there were too many entrants seeking not enough customers and licenses, and paying too many government fees and taxes and marketing costs. It was not a good business, in general, to try and enter.

I could have tried a lot harder. I could have pitched the whole thing to billionaires and worked through gamblers I know. Or I could have aggressively pitched various casinos, online and offline, to sell the project. Ultimately, after everything I’d been through, I must not have wanted it enough, and I had alternative opportunities I was happy to take. So I swore off letting myself think about these questions all day.

Now in 2024, with the rise of generative AI, the opportunity cost of trying, for me in particular, seems very high. It would take quite a lot to be worth shifting my attention. We’d need the right budget and team, and I’d need a lot of the upside. And I’m still not confident it would be worth taking my eye off the generative AI ball, if it meant I had to choose.

Still, how cool would it be? For everyone.

Imagine a graph that I call the “U.” Plot the popularity of the sport with the American sports-betting public on the x-axis, and how profitable it is to bet on the sport on the y-axis; it forms a U-shaped pattern. For extremely obscure sports—Russian ping-pong became a fad at one point during the pandemic—it isn’t really worth the sportsbook’s time to price them to a high degree of precision.

On the other end of the U are extremely popular events: for instance, the NCAA tourney, big MMA fights, or the World Cup. For events like these, “the amount of money bet by the public ends up dwarfing the amount bet by professionals,” and the market doesn’t necessarily clear to an efficient price. (3233)

I can very much verify that the U-curve is a big deal.

The odds on low-level sports are quite inaccurate. If you are any good at all, you should be able to win.

For example, at one point I took my baseball handicapping program, designed for MLB, and used it with essentially no changes on Japanese Baseball, which at least back then was on the left side of the U-curve. Did I know who a single player in the league was? No. That did not matter, because the lines were so bad.

You can also get a similar effect by betting early. If you bet NCAA Football games on Monday morning, you’ll get lower limits, but much softer lines. On game day, things are much harder.

That’s what I mean by finding good bets not being so obviously the ‘easy part.’ Yes, you can find good NCAAF bets on Monday morning and that’s relatively easy. But because you need to get down, those lines are only sort of real to you. Want to go big? You’ll need to be able to beat the lines on Saturday morning at 11am eastern.

Whereas yes, if the event has enough public money, that square public money overwhelms the market. The sharp gamblers will scale up their action somewhat for The Big Game, but at some point risk and bankroll management get in the way.

The best is to take an event like the Super Bowl, and take advantage of all the props and weird offerings, which the public loves and no one at the sportsbook knows how to price because they are not even in the habit of offering them.

Every year in late January or early February, Peabody—along with his betting partners and backpacks full of more than $100,000 in cash—makes a pilgrimage to the Westgate, which is the first sportsbook to post Super Bowl prop bets.

When the SuperBook first publishes its prop lines about ten days before the game, bettors queue up to place them under a strict set of rules. No more than two bets at a time—it’s sort of the sports-betting equivalent of Noah’s Ark—of no more than $2,000 each.

Besides, when an originator like Peabody bets, the SuperBook at least gets valuable information—one of the best bettors in the world has tipped his hand and the Westgate can use that to adjust its lines. So Super Bowl props are a win-win—Peabody gets a +EV bet, and the Westgate gets to learn what he thought when it has ten more days to take millions in public money.

The public is losing out, but at least they’re having fun. If every day were the Super Bowl, sports betting would be a booming industry—but of course it isn’t. (3254)

This is exactly how Westgate and Peabody should work together on every line, on every game, all the time. Except they don’t, because Westgate isn’t interested.

I loved betting the opening lines on things. As in, I’d be at the computer for an hour in the afternoon, hitting refresh on tomorrow’s MLB lines, hoping they would show up, my spreadsheet of odds ready to go. If they disagreed by more than a few percent, I’d hammer it. Sometimes twice. It is so much easier.

Eventually you run into the limits being too low. Until then, since most gamblers don’t actually try to bet that much on most games, almost all sports gamblers, who are looking to win, should be looking to bet as early as possible.

Even if you do want more than that, so what? Bet now, maybe also bet later. Do you think your bet now is going to impact the future path of the game’s prices that much? You think you are that important? If you are right, it moves a little faster. If you are wrong, it likely snaps back quickly as someone takes the other side.

Of course, if you are someone who tends to bet a ton of money, who the books respect, then showing your hand like this actually does matter. So it is deeply foolish. Everyone else? Get there firstest with the mostest. Or wait for another limit jump, then pounce on that.

There is potentially also the worry that if you are a little too quick, you are giving away that you are looking to pounce. That you are ‘trying to win.’ But from what I could tell back in the day, no one much minds, or often even notices. Indeed, if you move right away after things are available, then that is highly convenient. There’s probably someone paying close attention, who was just now thinking about those odds. You’ve saved them the need to context shift. How thoughtful of you.

Those are the gamblers. The casinos, the sports betters, the poker players.

Those were good times. It is a world I dearly miss. Perhaps again, one day.

Next time, it’ll be about The Business. Silicon Valley, venture capital and crypto.

The most likely person to write On the Edge was Nate Silver.

Grok thinks the next most likely was Michael Lewis, followed by a number of other writers of popular books regarding people thinking different.

I see why Grok would say that, but it is wrong.

The next most likely person was Zvi Mowshowitz.

I haven’t written a book for this type of audience, a kind of smarter business-book, but that seems eminently within my potential range.

On the Edge is a book about those living On The Edge, the collection of people who take risk and think probabilistically and about expected value. It centrally covers poker, sports betting, casinos, Silicon Valley, venture capital, Sam Bankman-Fried, effective altruism, AI and existential risk.

Collectively, Nate Silver calls this cultural orientation The River.

It is contrasted with The Village, which comprises roughly the mainstream mostly left-of-center institutions, individuals and groups that claim that they are The Experts and the Very Serious People.

If you are thinking about Secret Third Thing, that Village plus River very much does not equal America, I was thinking a lot about that too. Hold that thought.

The book is a collection of different topics. So this review is that, as well.

The central theme here will be Yes, And.

Nate Silver wrote On the Edge for people who are not in The River. I suspect his main target was The Village, but there are also all the people who are neither. He knows a lot more than he is saying here, but it is a popular book, and popular books have to start at the beginning. There was a lot to cover.

This joy of this review? I don’t have to do any of that. This is aimed at those who read things I write. Which means most of you already know a lot, often most, of what is in at least large portions of On the Edge.

So this is a chance to go deeper, be more detailed and opinionated, with a different world model in many ways, and expertise in different spots along the River.

As with my other book reviews, quotes by default are from the book, and the numbers in parenthesis are book locations on Kindle. I sometimes insert additional paragraph breaks, and I fix capitalization after truncating quotes while being careful to preserve original intent. Also notice that I change around the order when it improves flow.

This review is in four parts, which I plan to post throughout the week: The Fundamentals (this post), The Gamblers, The Business and The Future.

On the Edge is a series of stories tied to together by The River and risk taking.

I see this as a book in four parts, which I’ve rearranged a bit for my review.

This post will cover The Fundamentals: The introduction, the overall concepts of The River and the Village, and various universal questions about risk taking. That includes the book’s introduction, and universal discussions pulled from later on.

Part 1 of the book, which I will cover in the second post The Gamblers, is about the world of gambling. You have poker, sports betting and casinos.

This was my favorite part of the book.

I have some experience with these topics, and got what I would call Reverse Gell-Mann Amnesia.

In normal Gell-Mann Amnesia, you notice the newspaper gets wrong the things you know the most about. Then you go on to not assume that the newspaper is equally inaccurate on other topics.

In reverse Gell-Mann Amnesia, you notice that the book is getting the details right. Not even once, while covering these topics, did I think to myself, ‘oh, that’s wrong, Nate got fooled or confused here.’

Are there places where I would have emphasized different points, taken a different perspective or added more information, or even disagreed? Oh, sure. But I’ve read a lot of this style of book, and Nate Silver definitely Gets It. This is as good as this format allows.

Drayton’s review found Nate making a number of ‘elementary’ mistakes in other areas. And yes, once you get out of Nate’s wheelhouse, there are some errors. But they’re not central or importantly conceptual, as far as I could tell.

Part 2 of the book, which I will cover in what I’ll call The Business, was about gamblers who play for higher stakes over longer periods, betting on real world things. As in, we talk about stock traders, Silicon Valley and venture capital. I think that the spirit is largely on point, but that on many details Nate Silver here buys a bit too much into the insider story that gets pitched to him and other outsiders, in ways that assume the virtues of good gamblers and The River are present to a greater extent than they are. They’re there, but not as much as we’d like.

Part 3 of the book, which I will also include in The Business, was about crypto and especially Sam Bankman-Fried. This part was a let down, and I’m mostly going to skip over it. On basic crypto my readers know all this already.

After Going Infinite and my review of that, it did not feel like there was much to add on SBF. I worry this tied presented SBF as more central and important to The River and especially to EA and similar areas than he actually was. I do get why Nate Silver felt he had to cover this, and why he had to run with it once he had it. We’ll hit some highlights that are relatively unique, but mostly gloss over it.

Part 4 of the book, which I will cover in The Future, was about AI and existential risk, including rationalists and EAs. He picks excellent sources: Sam Altman, Roon, Ajeya, Scott Alexander, Oliver Habryka, Eliezer Yudkowsky. He also talked to me, although I did not end up being quoted.

The parts that discuss the history of OpenAI reflect Silver essentially buying Altman’s party line in ways I found disappointing. I will do my best to point to my corrections of the record and distinctions in perspective.

The parts that talk about AI technically will be nothing new to blog readers here. I don’t think he got anything wrong, but we will mostly skip this.

Then there is the discussion of AI existential risk, and the role of the EA and rationalist communities. While I was disappointed, especially after the excellent start, I totally see how Nate got where he ended up on all this. It was a real outsider attempt to look at the situation, and here I can bring superior knowledge and arguments to bear.

Nate’s overall view, that existential risk is obviously real and important if you think AI is going to keep advancing, but that we cannot at this time afford to simply choose not to proceed, seems incomplete but eminently reasonable. Long book is long, and necessary background information is a big problem, but the discussion of existential risk arguments felt extremely abrupt and cut off, in ways the rest of the book did not. In contrast, he spends a bunch of time arguing we should worry about technological stagnation if we do not proceed with AI.

One thing about the final section I loved was the Technological Richter Scale. This was very good Rhetorical Innovation, asking people to place AI on a logarithmic scale of impact compared to other technologies. This reveals better than other methods that many, perhaps most, disagreements about AI existential risk are actually disagreements about AI capabilities – those not worried about AI largely do not believe AI will be ‘all that.’ I covered the scale in its own post, so it will have its own reference point.

What is The River?

Every book like this needs a fake framework, a new set of categories to tie together chapters about various topics, that then they see everywhere.

What is The River?

The River is a sprawling ecosystem of like-minded people that includes everyone from low-stakes poker pros just trying to grind out a living to crypto kings and venture-capital billionaires. It is a way of thinking and a mode of life. People don’t know very much about the River, but they should.

Most Riverians aren’t rich and powerful. But rich and powerful people are disproportionately likely to be Riverians compared to the rest of the population. (73)

The River Nature is a way of thinking and a mode of life.

If you have that nature by default, you are a Riverian, a citizen of the The River.

If your group or activity rewards and celebrates that nature, then it is along The River.

In this mode, you think in terms of probabilities and expected value (EV). You seek the most accurate possible model of the world, including what actions are how likely to lead to what results. You the riverian look to make the best decisions possible. You are not afraid of risk, but seek to take only the good risks that are +EV.

You can then be somewhat risk averse when making decisions, risk neutral or even risk loving. All riverians have weaknesses, ways they systematically mess up. The key is, you accept that risk is part of life, and you look to make the most of it, including understanding that sometimes the greatest risk is not taking one.

Riverians are the Advantage Players of life.

They want life to be about everyone making good decisions. A True Riverian learns to inherently love a correct play and hate a mistake, in all contexts, from all sides that are not their active opponents. They want those good decisions and valuable actions to be rewarded, the bad decisions and destructive actions punished. They want that to be what matters, not who you know or who you are or how you play some political game.

Riverians hate being told what to do if they don’t think it will help them win. They despise when others boss them around and tell them to do dumb things, or are told to copy what others around them do without justification.

The River is where people focus on being right, taking chances and doing what works, and not letting anyone tell them different.

As you would expect, The River and its inhabitants often looks stupid. Things are reliably blowing up in various faces and others, especially The Village, are often quick to highlight such failures.

Given everything that took place while I was writing this book—poker cheating scandals; Elon Musk’s transformation from rocket-launching renegade into X edgelord; the spectacular self-induced implosion of Sam Bankman-Fried—you’d think the River had a rough few years. But guess what: the River is winning. (77)

Few would accuse Elon Musk or SBF of being the innocent victims of bad luck regarding recent events in their lives. Mistakes, as they say, were made. Massive, historical mistakes. But the alternative, the world and culture and nature where such mistakes are not happening, where people don’t take successful risks and then get into position to take even bigger ones, is worse, not better.

As a fellow inhabitant of The River, this one rings far more true and important than most. I am definitely convinced the River is real, and that it definitely includes most of the groups listed in the book, although we’ll see there is one case I think is less clear.

One very clear truth is that playing poker or betting on sports is Not So Different from investing in tech startups or investing (sometimes ‘investing’) in crypto tokens.

The River isn’t all fun and games. The activities that everyone agrees are capital-G Gambling—like blackjack and slots and horse racing and lotteries and poker and sports betting—are really just the tip of the iceberg. They are fundamentally not that different from trading stock options or crypto tokens, or investing in new tech startups. (94)

If you had to divide that collection into two groups, you could divide it into the casino gambling on one side and the ‘investments’ on the other, and that would be valid.

Almost as valid would be to move poker and sports betting and other skill games into the ‘investment’ category, and leave slot machines and craps and other non-skill games in the other (with the exception of a small number of Advantage Players, which the book explores). In practice, a zero-day stock option is closer to a sports bet than it is to buying the stock and holding it for a month. More on that throughout.

There is quite a lot of that actual straight up gambling.

Literal gambling is booming. In 2022, Americans lost around $60 billion betting at licensed casinos and online gambling operations—a record even after accounting for inflation. They also lost an estimated $40 billion in unlicensed, gray-market, or black-market gambling—and about $30 billion in state lotteries. To be clear, that’s the amount they lost, not the amount they wagered, which was roughly ten times as much. (188)

A total of $130 billion means the average adult lost on the order of $500 gambling, but of course that is wildly unevenly distributed. Most lost nothing or very little. A few lost a lot.

The multiplier depends on the game. A 10x multiplier for casinos and grey market gambling seems reasonable.

For state lotteries, the multiplier is… less.

On average, the government keeps about 35 cents of every dollar you spend on a lottery ticket, and some states keep 80 percent or more. Lottery tickets are purchased disproportionately by the poor. (2871)

As the game Illuminati describes the state lottery, it’s a tax on stupidity, and the money rolls in. That is unfair. Slightly. Only slightly. It is absurd how terrible the official lotteries are.

There is nothing like being where you belong to remind you who you are, as Nate experienced after going to his first real poker tournament after Covid.

The other big realization I had on that flight home from Florida was that this world of poker players and poker-playing types—this world of calculated risk-taking—was the world where I fit in. (206)

And yet, the people in the River are my tribe—and I wouldn’t have it any other way. Why did my conversations flow so naturally with people in the River, even when they were on subjects I was still learning more about? (378)

Why indeed? Why does he think he fits in so well?

First, there’s what I call the “cognitive cluster.” Quite literally: How do people in the River think about the world? It begins with abstract and analytical reasoning. (387)

The natural companion to analytic thinking is abstract thinking—that is, trying to derive general rules or principles from the things you observe in the world. Another way to describe this is “model building.” (391)

Then there’s the “personality cluster.” These traits are more self-explanatory. People in the River are trying to beat the market. (422)

Relatedly, people in the River are often intensely competitive. (430)

Finally, I put risk tolerance in this cluster because—whether they’re degens or nits in other parts of their lives—being willing to break from the herd and go against the consensus is certainly not the safest professional path. (436)

Nate’s history is that he was a poker player, happily minding his own business, then Congress sneaked a provision into a bill that killed American online poker and took away his job.

There was one silver lining: the UIGEA piqued my interest in politics. The bill had been tucked into an unrelated piece of homeland security legislation and passed during the last session before Congress recessed for the midterms. It was a shifty workaround, and having essentially lost my job, I wanted the people responsible for it to lose their jobs, too. (235)

I have a deeply similar story, with sports betting and the Safe Port Act. Congress tucks a provision into a different law and suddenly online sports betting transforms and being a sports better in America became vastly more difficult. Both of us got fired.

Nate went into politics. I chose a different angle of response. We both did well in our new modeling work, and then both got frustrated over time.

In Nate’s case, the problem was that election forecasts and regular people don’t mix.

But here’s the thing about having tens of millions of people viewing your forecast: a lot of them aren’t going to get it. (246)

Expected value is such a foundational concept in the River’s way of thinking that 2016 served as a litmus test for who in my life was a member of the tribe and who wasn’t. At the same moment a certain type of person was liable to get very mad at me, others were thrilled that they’d been able to use FiveThirtyEight’s forecast to make a winning bet. (267)

But permit me this one-time informal use of “rational”: people are really fucking irrational about elections (308)

Likewise, the tendency in the media is to contextualize ideas—The New York Times is no longer just the facts, but a “juicy collection of great narratives,” as Ben Smith described it. (420)

In my case, those I worked with declared (and this is a direct quote) ‘the age of heroes is over,’ cut back on risk and investment accordingly, and I wept that this meant there were no more worlds to conquer. So I left for others.

We are both classic cases of learning probability for gambling reasons, then eventually applying it to places that matter. It is most definitely The Way.

Blaise Pascal and Pierre de Fermat developed probability theory in response to a friend’s inquiry about the best strategy in a dice game. (367)

He notices, but he can’t stop himself.

I feel like it’s my sacred duty to call out someone who’s wrong on the internet. (6417)

I indeed see him doing this a lot, especially on Twitter. Nate Silver, with notably rare exceptions, you do not have to do this. Let it go.

There’s also another community that competes with the River for power and influence. I call it the Village. (440)

The Village are the Respectable Authority Figures. The Very Serious People.

It consists of people who work in government, in much of the media, and in parts of academia (although perhaps excluding some of the more quantitative academic fields such as economics). It has distinctly left-of-center politics associated with the Democratic Party. (442)

My title for this section is not entirely fair to The Village. It is also not as unfair as it sounds. Members of The Village are usually above average in intelligence and skill and productivity. The vast majority of people are not in either Village or River.

But yeah, in the ways Village and River differ strongly? I mostly stand by it. The failure to use the River Nature, the contrast in modes of cognition, is stupefying.

In some contexts, those not in The Village proper will attempt to play the role of The Village in a given context. In doing so, they take on The Village Nature, to attempt to operate from that same aura of authority and expertise. And yes, it reliably makes them act and talk stupider.

What makes people far stupider than that is being trapped in the Hegelian dialectic, and in particular the one of party politics.

Indeed, Riverians inherently distrust political parties, particularly in a two-party system like the United States where they are “big tent” coalitions that couple together positions on dozens of largely unrelated issues. Riverians think that partisan position taking often serves as a shortcut for the more nuanced and rigorous analysis that public intellectuals ought to engage in. (466)

That is Nate Silver bending over backwards to be polite. Ask most members of The River, whether they back one party, the other or neither, and they will say something similar that is… less polite.

The Village also believes that Riverians are naïve about how politics works and about what is happening in the United States. Most pointedly, it sees Donald Trump and the Republican Party as having characteristics of a fascist movement and argues that it is time for moral clarity and unity against these forces. (510)

The Village thinks that if you do not give up your epistemics to support their side of the Hegelian dialectic, then you lack moral clarity and are naive. No, that claim did not start with Donald Trump. Nor is it confined to The River.

Riverians are fierce advocates for free speech, not just as a constitutional right but as a cultural norm. (488)

To the extent they express an opinion on the issue, whether or not they belong to The River, every single person whose opinion I respect is a strong advocate for free speech.

I remember when I thought The Village believed in free speech too. No longer. Some members of The Village do. Overall it does not.

That is a huge problem. The Village that exists today is very different from The Village that my parents thought they were members of back in the day. I would still ultimately have the River Nature, but I miss the old Village.

I buy that there exist The River and The Village.

What about everyone and everything else?

This becomes most obvious when the book or author discusses Donald Trump.

Obviously Donald Trump is not of The Village.

The temptation is to place him in The River, but that is also obviously wrong.

Donald Trump may have an appetite for some forms of risk and even for casinos, but he does not have The River Nature. He does not think in probabilities and expected values. He might want to run a casino, but that is because he was in the real estate business, not because he has any affinity for River-style gamblers.

If you look at Donald Trump’s supporters, it becomes even clearer. These people hate The Village, but most also view The River as alien. They don’t think in probability any more than Villagers do. When either group goes to a casino, almost none of them are looking for advantage bets. They, like most people and perhaps more so, are deeply suspicious of markets, and those who speak in numbers and abstractions.

The Village might be in somewhat of a cold war with The River, but the River is not its natural enemy or mirror. Something else is that.

So what do we call this third group? Not ‘everyone not in the Village or River’ and not ‘the other political party’ but rather: The natural enemies of The Village?

I asked for the LLM consensus is in, and there is a clear winner that I agree is indeed this group’s True Name in this schema, that works on many levels: The Wilderness.

In the extended metaphor, it used to be that Village and River were natural allies. Now that this is not the case. The Village presents the world as a Hegelian dialectic between it and the Wilderness, treating every other group including the River as irrelevant or some side show.

Their constant message to the River is: You don’t fing matter. Their other message is that they do not tolerate neutrality. When the Village turns on you and yours for not falling in line – and the River Nature as a matter of principle does not bow down, which is a big hint as to who they centrally are – but especially when the Village turns directly on you, you feel cast out and targeted.

Thus, increasingly, some members of The River, and others who The Village casts out over some Shibboleth, end up in The Wilderness. This is a deeply tragic process, as they abandon The River Nature and embrace The Wilderness Nature, inevitably embracing an entire basket of positions, usually well past the point of sanity.

See: Elon Musk.

Does that still leave a Secret Fourth Thing?

One can of course keep going.

Most people, even if they ‘put their trust in’ the Village, Wilderness or River, or even The Vortex. They do not at core have any of these natures. They are good people trying to go about their business in peace. One can call this The People.

(Note: I tried to make this fit the Magic color wheel in a fun way, but it didn’t work.)

An alternative telling here in another good book review suggests The Fort as the right-wing mirror image of The Village. The Fort is where Ted Cruz and Samuel Alito hang out. It’s important to note that The Wilderness and The Fort are not the same place. And we both agree that The People are a distinct other thing.

The actual section is actually “Why the Valley Hates the Village (4756)” but this is one case where I think one can push the book thesis further. Yes, centrally Silicon Valley, but the entire River hates the Village, mostly for the same underlying reasons.

Those reasons, centrally, are in my own words something like this:

The Village in many ways does mean well.

But it fundamentally views the world as a morality play and Hegelian dialectic. The us against the them. ‘Good causes’ to advance against enemies like ignorance and greed. Those ‘good causes’ and their Shibboleths are often chosen based on what feels good to endorse on a surface level, rather than what would actually do good. Often they get hijacked by various social dynamics spun out of control, often caused by people who do not mean well. They rarely ask deeply about the effectiveness of their proposals.

Not only do they think this is going on, they think it is the only important thing going on. And they think primarily on Simulacra level 3, in terms of alliances and implications and how things sound about saying what type of person you are and are allied with, rather than about what actually causes what.

When they lie or break the rules or the norms of common decency it is for a good cause. When others do it they cry bloody murder. And they justify all that by pointing at The Wilderness and saying, have you seen The Other Guys? As if there were only two options.

They have never understood economics and incentives and value creation, or trade-offs, treating you as a bad person if you point out or care about such considerations. They treat your success and your wealth and legacy as fundamentally not yours, and think they have a right to take it from you, and that not doing so would be unfair. They have great respect for certain particular local details that made their Shibboleth and Good Cause lists, while completely ignoring vital others and Seeing Like a State.

Indeed, if you have other priorities than theirs, if you break even one of their Shibboleths or triggers, or fail to sufficiently support the wrong one at the wrong time, they often cast you out into The Wilderness, doing their best to have people place you there regardless of whether that makes any sense. And the resulting costs, in the form of the inability to Do Things of various sorts, is growing over time, to the point where our civilization is in rather deep trouble.

To be fair, trust in Big Tech has also dropped sharply in polls. But Silicon Valley is not particularly dependent on public confidence so long as it continues to recruit talent and people continue to buy its products. (5268)

Big Tech now sees the specter of actively dangerous enemy action. So does what Marc Andreessen calls ‘Little Tech.’ So does the rest of The River.

The River, in the past, mostly put up with The Village. The Village has a lot of practical advantages, did broadly mean well, and there were common enemies who were seen as clearly worse. And importantly, The Village was mostly leaving The River alone in kind, or at least giving it some space in which to safely operate, and the paralysis effects were nowhere near as bad.

Starting around 2016, a lot of that changed, and other parts reached tipping points. Also they blew their remaining credibility and legitimacy in increasingly stupid ways, cumulating in various events around the time of Covid. The Village’s case for why The River should accept its authoriah is down to a mix of ‘we have the legible expert labels and are Very Serious People’ and ‘you should see The Other Guy (e.g. The Wilderness’). Both are increasingly failing to hold water.

The first because lol. The second because at some point that’s a risk people in The River will be willing to take. Also decision theory says you can’t let them play you like that indefinitely. We can’t let the Hegelian dialectic win.

There was a deal, well beyond how media coverage works. The Village broke the pact.

This deal’s getting worse and worse all the time. The rent got too damn high.

Everyone got really mad at each other about 2016. (4762)

That was a big breaking point. Meta is awful, Facebook is awful, Mark Zuckerberg is awful, and also their AI positions might well doom us all. But the story that a few Facebook ads were why Trump beat Hillary Clinton in 2016 was always absurd.

And yet, that became The Narrative, in many circles. That had big effects.

One that Nate doesn’t focus on is Gell-Mann Amnesia. If The Experts are so convinced of something like this, that is clear Obvious Nonsense, what else that they tell you is Obvious Nonsense?

Another is that Big Tech, and tech in general, got a clear lesson in the Copenhagen Interpretation of Ethics. If tech is seen interacting with something, they were informed, then tech will be blamed for the result. Which, given how much tech interacts with everything, is a real problem.

Patrick McKenzie has noted that one big result of this was that when we got a Covid vaccine, we had a big problem. Rather than reach out for Google’s help, the government tried to muddle through without it. Google and Apple and Amazon did not dare step forward and offer help in getting shots into arms, for fear of blowback. Even if they got it right, they feared they would be showing up the government.

That is how a handful of volunteers on a Discord server, VaccniateCA, ended up becoming our source of information on the vaccine.

The Village is about group allegiance, while Silicon Valley is individualistic. (4782)

The Village really is about group allegiance, and also increasingly (at least until about 2020) group identity. Which groups and ideas you are for, which ones you are against. They are centrally Simulacra Level 3, although they also spend time at 1, 2 and 4.

The River is individualistic, as Nate says, but even more it is about facts and outcomes and ground truth: Simulacra Level 1. Most River factions are very Level 1 focused.

Silicon Valley is the major part of the River that compromises on this the most. They talk about the importance of Level 1, ‘build something people want.’ Yet they are very willing to care deeply about The Vibes, and often primarily operate more on Level 4 (and at times the Level 4 message is exactly that you too should be doing this), as well as some of Levels 2 and 3. A certain amount of lying, or at least selective presentation, is expected, as is loyalty to the group and its concepts. It’s complicated.

Meritocracy is another big deal for The River. Back in the day they could tell the story that The Village was that way too, but that story got harder to tell over time.

Especially in explicit politics, it was clear that merit was going unrewarded.

But campaigns are not always very meritocratic. “It’s very rare to actually be able to assess whether someone did a good job,” Shor said. Campaigns have hundreds of staffers and ultimately only one real test—election night—of how well they did, which is often determined by circumstances outside the campaign’s control. Relationships matter more than merit. So people get ahead by going with the program. (4790)

Extreme merit still gets rewarded, if you have a generational political talent. In the absence of that, the variance overwhelms skill, so politics rules politics.

There are turf wars—and philosophical ones—between Silicon Valley and Washington over regulation. (4801)

This is natural and expected. To some extent everyone is fine with it. The problem is that there are signs this may be taken way too far, in ways that kill or severely damage the Valley’s business model. See the proposals for unrealized taxes on capital gains, for various crazy things that almost get done with social media, and now various claims about AI.

Silicon Valley is skeptical of the “trust the experts” mantra that the Village prizes. (4832)

Skeptical is a nice word for how The River views this by default, in worlds where the experts are plausibly trustworthy. Then various things happened, and kept happening. Increasingly, ‘trust the experts’ became an argument from authority, and from status within the Village, and its mask as a ‘scientific consensus’ resulting from truth seeking became that much harder to maintain.

Tech leaders are in an ideological clash with their employees and blame the Village for it. (4855)

Yes. Yes, they do. It is not a reasonable way to look at the situation, which involved things such as:

In Silicon Valley, you’re supposed to feel like you have permission to express unpopular and possibly quite wrong or even stupid ideas. So Damore’s firing represented a shift. (4878)

The Village created a social climate, especially among tech employees, where Google felt forced to fire Damore and engage in other similar actions, as did many other tech companies. Based on what I have heard, in many places including Google things got very out of hand. This did not sit well.

But what about Silicon Valley’s elites—the top one hundred VCs, CEOs, and founders? There’s no comprehensive catalog of their political views, so I’ll just give you my impressions as a reporter who’s had a variety of conversations with them.

It’s worth keeping in mind that rich people are usually conservative. If Silicon Valley’s elites were voting purely with their pocketbooks, they’d vote Republican for lower taxes and fewer regulations, especially with Khan heading the FTC. (4862)

The elites for a while faced sufficient pressure from the Village, and especially from their employees, they did not dare move against the Village and felt like they were being forced to abandon many core River values and to support democrats despite the financial problems with that and the constant attempts by the Village to attack the SV elites, lower their status and potentially confiscate their wealth and break up their businesses and plans. Recently that danger and fear of ideological backlash has subsided, everyone is sufficiently fed up and worried about actual legal consequences, and feels like the masks are already off, and so there is a lot more open hostility.

Another classic clash point was Thiel getting his revenge on Denton for outing Thiel.

I asked Thiel about this passage. Hadn’t he been a hypocrite to focus on destroying a rival, Denton, half a continent away in New York in a completely unrelated business, all for the sin of outing a gay man who lived in the world’s gayest city, San Francisco?

Thiel quickly conceded the point. “In any intensely competitive context, it is almost impossible to simply focus on a transcendent object and not spend a lot of time on the personalities of one’s rivals.” (4900)

I think it is highly rational and reasonable decision theory to retaliate for that, even if you think everything turned out fine for Thiel after being outed. If someone hits you like that, they are doing it in part because they don’t think you’ll hit back, and others are watching and asking the same question. It is +EV to make everyone think twice, and especially to in advance be the type of person who would do that, especially in a way others can notice. Which Thiel very much was, but Denton didn’t pay enough attention, or didn’t care, so he found out.

One thing that did not ring true for me at all was this:

I asked Swisher why tech leaders like Thiel and Musk are so obsessed with their media coverage. She didn’t need much time to consider her answer. “It’s because they’re narcissists. They’re all malignant narcissists,” she said. (4908)

I’ve long had to mute Kara Swisher. She is constantly name calling, launching unfounded personal attacks and carrying various people’s water, and my emotional state reliably got worse every time I saw anything she wrote without providing me with useful information. She is not a good information source, and this type of comment is exactly why.

Yes, they care about their image, but their image is vital to their business and life strategy. I never got this sense from Thiel at all, and I don’t have any info you don’t have about Musk but certainly he cares far more than most people about big abstract important things, even when I think he’s playing badly.

A toy risk question that came up later, to get us started.

SBF then made an analogy that you’d think would trigger my sympathies—but actually raised my alarm. “If you’re making a decision, such as there’s no way that it goes really badly, then I sort of feel like—you know, zero is not the correct number of times to miss a flight. If you never miss a flight, you’re spending too much time in airports.”

I used to think about air travel like this. I even went through a phase where I took a perverse joy in trying to arrive as close to the departure time as possible and still make the flight.

Now that I’m more mature and have a credit card that gives me access to the Delta Sky Club, I don’t cut it quite so close. But the reason you should be willing to risk missing a flight is because the consequences are usually quite tolerable. (6033)

A fun tip I recently realized is that if you never miss a flight, that generally means you are spending too little time at airports.

Explanation: If you actual never miss a flight that means you don’t fly enough, since no amount of sane buffer gets your risk anywhere near zero, airlines be tripping.

On the practical question, the key indeed is that there is limited upside and limited downside. This can and should be a calculated risk. Missing your flight is sometimes super expensive, sometimes trivially cheap – there are times you’ll take a $500 buyout to skip the flight happily (although note to airlines: You’ll get a much better price off me if you ask me before I head to the airport!), others where even for thousands the answer is a very clear no.

And there are times when leaving an extra two hours at the airport is a trivial cost, you weren’t doing anything vital and have plenty of good podcasts and books. Other times, every minute before you leave is valuable. And also there are considerations like lounge access.

I don’t have lounge access, but I’ve found that spending time at airports is mostly remarkably pleasant. You have power, you have a phone and if you want a laptop and tablet, there’s shops and people watching, it’s kind of a mini-vacation before the flight if you have a good attitude. If you have an even better attitude, so is the flight.

You would not, however, say ‘if you are still alive you are not doing enough skydiving.’

(Or that you are not building sufficiently advanced AIs, but I digress.)

If your decision is very close, why not randomize? Nate does this often, to his partner’s frequent dismay.

We’re indifferent between the Italian place and the Indian place, there’s no reason to waste our time agonizing over the decision. (1049)

I agree I should do this more. Instead, I try to use semi-random determinations, with the same ‘if I made a big mistake I will notice and fix it, and if I made a small mistake who cares’ rule attached.

However I also do frequently spend more time on close decisions. I think this can be good praxis. It is wasteful in the moment, but going into detail on close decisions is a great way to learn how to make better decisions. So in any decision where it would be great to improve your algorithm, if it is very close, you might want to overthink things for that reason.

At other times, yeah, it doesn’t matter. Flip that coin.

The thesis is that wherever you find highly successful risk takers, you see the same patterns, the same River Nature.

[Those are] hardly the only people who undertake risks. So I want to introduce you to five exceptional people who take physical risks: an astronaut, an athlete, an explorer, a lieutenant general, and an inventor. (3925)

The thesis: If you want to succeed at risk taking, you need to be Crazy Prepared. You need to take calculated risks with your eyes open. You need the drive to succeed enough to justify the risks. You need all the good things.

Even if they aren’t quantitative per se, they are highly rigorous thinkers, meticulous when it comes to their chosen pursuit. One thing’s for sure: our physical risk-takers are definitely not part of the Village. (3934)

Successful risk-takers are cool under pressure. They don’t try to be heroes, but they can execute when the chips are down. (3988)

Successful risk-takers have courage. They’re insanely competitive and their attitude is: bring it on. (4013)

Successful risk-takers have strategic empathy. They put themselves in their opponent’s shoes. (4035)

Successful risk-takers are process oriented, not results oriented. They play the long game. (4067)

Successful risk-takers take shots. They are explicitly aware of the risks they’re taking—and they’re comfortable with failure. (4098)

Successful risk-takers take a raise-or-fold attitude toward life. They abhor mediocrity and they know when to quit. (4127)

Successful risk-takers are prepared. They make good intuitive decisions because they’re well trained—not because they “wing it.” (4171)

Successful risk-takers have selectively high attention to detail. They understand that attention is a scarce resource and think carefully about how to allocate it. (4193)

Successful risk-takers are adaptable. They are good generalists, taking advantage of new opportunities and responding to new threats. (4230)

Successful risk-takers are good estimators. They are Bayesians, comfortable quantifying their intuitions and working with incomplete information. (4252)

Successful risk-takers try to stand out, not fit in. They have independence of mind and purpose. (4284)

Successful risk-takers are conscientiously contrarian. They have theories about why and when the conventional wisdom is wrong. (4306)

Successful risk-takers are not driven by money. They live on the edge because it’s their way of life. (4350)

Is it true that (most) successful risk-takers are not driven by money? Is that in any way in opposition to the edge being a way of life? In the end, most people are in key senses not driven by money. The money is either the means to an end, or it is ‘the score.’ But money is still highly motivational, as that means to an end or as that score. Was SBF motivated by money, or not? What’s the difference?

Of all these, I’d say the most underappreciated is being process oriented. Missing that will get you absolutely killed.

Whereas, as we see when we get to Silicon Valley, being unaware of the risks or odds can kind of work out in some situations. So can, as we see with the astronaut, not being contrarian.

So for example on a spacecraft:

On the Blue Origin orbital spacecraft. Vescovo, who is also a former commander in the U.S. Navy Reserve, told me that the mentality required in exploration, the military, and investing is more similar than you might think. “It’s risk assessment and taking calculated risks,” he said. “And then trying to adapt to circumstances. I mean, you can’t be human and not engage in some degree of risk-taking on a day-to-day basis, I’m just taking it to a different level.” (3983)

Or for a pilot, fictional bad example edition:

What most annoyed Vescovo about Top Gun: Maverick was Tom Cruise’s insistence that you should just trust your gut and improvise your way out of a hairy situation. “The best military operations are the ones that are very boring, where things go exactly according to plan. No one’s ever put in any danger,” he said. “You want to minimize the risks. And so yeah, Top Gun, it looked great on film. But that is not how you would try and take out that target.” (4172)

I haven’t seen Top Gun: Maverick, but you don’t have to in order to understand.

So is there never a place for trusting your gut? That’s not quite what Vescovo is saying. Rather, it’s that the more you train, the better your instincts will be. “When something is for real and is an emergency, how many times have you heard people say, ‘Oh, you know, the training kicked in.’ ” Training, ironically, is often the best preparation to handle the situations that you don’t train for. (4180)

The problem with Top Gun: Maverick wasn’t with Maverick—his instincts probably were pretty good—but that he was imploring other pilots to trust their gut and be heroes when they didn’t have the same experience base. (4191)

Yep. The way you trust your gut and improvise well is to be Crazy Prepared. Then you have a gut worth trusting. You can’t give that advice to people before they’re ready, and people often do it and cause a lot of damage or failure.

“The best players will study solvers so much that their fundamentals become automatic,” he wrote—the complex System 2 solutions that computers come up with make their way into your instinctual System 1 with enough practice. That frees up mental bandwidth for when you do face a hairy situation, or a great opportunity. (4187)

That’s how the true masters do it. The less you have to think about the basics, the more they’re automatic, the more you can improve other stuff.

Exactly in the Magic competitions where I was Crazy Prepared I was able to figure out things under the lights I’d never considered in practice – for example in my victory in Tokyo, I’d not once named Nightscape Familiar on a Meddling Mage in all of practice, or named Green for Voice of All without a Crimson Acolyte against Red/Green, or even considered cutting Fact or Fiction while sideboarding. Winning required, as it played out, that I figure out to do all three.

Who would take the risks without the pressure?

Karikó found it more in the United States than in Communist-era Hungary. “If I would stay in Hungary,” she told me,[*6] “can you imagine I would go and sleep in the office?” In the United States, she found “the pressure is on in different things, so that is why it’s great.” (4029)

For me the answer is that you can sort of backdoor into situations where the risks are so overwhelmingly +EV that you have no choice but to take them, and you have the security of knowing you have a safety net if you need one – I was never worried I would end up on the street or anything, I could always get a ‘real job’ (as I did at Jane Street), learn to code (well enough to earn money doing it) or even play poker.

In so many ways, our civilization handled Covid rather badly. Nate identifies correctly one of the two core mistakes, which was that it was a raise-or-fold situation, and we called, trying to muddle through without a plan.

I’d argue, for instance, that the world might have been better off if it treated the COVID-19 pandemic as a raise-or-fold situation. (4148)

The few countries like New Zealand and Sweden that pursued more coherent strategies—essentially, New Zealand raised and Sweden folded—did better than the many that muddled through with a compromise approach. (4150)

This is essentially correct for the pre-vaccine period. Raising and folding were both reasonable options. The strategy we chose was neither, and we executed it badly, aside from (by our standards) Operation Warp Speed.

The other core mistake was botched execution at every step. That’s another story.

What I think of as ‘contrarian’ Nate calls ‘independent’ here?

There is an oft-neglected distinction between independence and contrarianism. If I pick vanilla and you pick chocolate because you like chocolate better, you’re being independent. If you pick chocolate because I picked vanilla, you’re being contrarian.

Most people are pretty damned conformist—humans are social animals—and Riverians are sometimes accused of being contrarian when they’re just being independent. If I do the conventional thing 99 percent of the time and you do it 85 percent of the time, you’ll seem rebellious by comparison, but you’re still mostly going with the flow. (4307)

This is in opposition to Nate saying ‘successful risk-takers are conscientiously contrarian.’ Successful risk-takers are being, by this terminology, independent.

On the (OF COURSE!) contrary, I think of this very differently. Being ‘a contrarian’ means exactly what Nate calls independent here: Being willing to believe what seems true, and do what you prefer, and say it out loud, exactly because you think it is better.

Most people are, most of the time, rather unwilling to do this. Even when they disagree or do something different, they are still following the script.

Yes, they will sometimes pick chocolate over vanilla despite you picking vanilla. At other times, they will pick chocolate over vanilla in part because you picked vanilla… because it is standard to not order the same thing at the same time, so also picking literal vanilla would actually in many cases be contrary and independent. But they’ll still look for chocolate, not the weird sounding flavor they actually like and want, if they can make it work – which is similar to how those pushed out of The Village end up in The Wilderness (or Vortex), rather than in The River or doing some secret third thing (I think on reflection the secret thing is kind of Always Third, even if it’s not?).

Someone who actually does unconventional things 15% of the time is a world-class rogue actor. You may think that someone (let’s say Elon Musk) is doing tons of unconventional stuff and is totally out in space, but when you add it all up he’s still on script something like 99% of the time. The difference is 99% versus 99.9%, or 99.99%.

Which is smart. The script isn’t stupid, you shouldn’t go around breaking it to break it, and if you broke it 15% of the time you would, as they say, Find Out. When we say 85%, we mean that 15% of the time Elon is doing some aspect of the thing differently.

After all, when you order the Cinnamon Toast ice cream, it’s still ice cream, and you are probably still putting it in a cup or cone and eating it, and so on.

What Nate Silver is calling ‘contrarian’ here is what I’d call ‘oppositional.’ It is the thing where Your Political Party says ‘I think apple pie is good’ and Their Political Party says ‘well then I suppose apple pie is bad.’

I’m going to finish the introduction with Nate’s discussion of prediction markets.

Nate Silver, now advisor to Polymarket, is definitely a prediction market fan.

He still has reservations, because he has seen their work.

My views are mostly sympathetic, but not without some reservations. That’s in part because of some scar tissue from too many arguments I’ve had on the internet about the accuracy of prediction markets versus FiveThirtyEight forecasts. The FiveThirtyEight forecasts have routinely been better—I know that’s what you were expecting me to say, but it’s true—something that’s not supposed to happen if the markets are efficient.

Then again, maybe this doesn’t tell us that much. Elections are quite literally the Super Bowl of prediction markets—there’s so much dumb money out there (lots of people who have very strong opinions about politics) that there isn’t necessarily enough smart money to offset it. (6736)

It makes sense that presidential elections are a place where prediction markets will be great for generating liquidity, and great for measuring how much various changes impact probabilities, but exhibit a strong bias, and sometimes be pretty far off.

We certainly saw that in 2020, and the markets in 2008 and 2012 also had major issues.

My guess is that Polymarket is biased in favor of Trump, because those trading in a crypto prediction market are going to have that bias, and because a lot of traders realize that if Harris wins they have a good chance of being able to buy Harris at 90%, or at least 95%, similar to what happened in 2020. That should in turn bias the odds now.

There should still be plenty of reasons to keep this in check, so the market won’t be too far off. And changes over time should mostly reflect ground truth. PredictIt has Harris up 55-45 while I write this while Polymarket is 50-50, and Metaculus also has Harris at 55-45. That’s a substantial difference, but it sounds bigger than it is. Right now (the evening of 9/11/24) I think that PredictIt is right, but who wins won’t settle that question.

One underrecognized pattern is that odds often do tend to stick at exactly even far more often than they should. So often they are selling dollars for fifty cents. There are a lot of people who are willing to bet at 50% odds, but no higher, and often one of them is willing to go big, so things get stuck there.

But the bigger concern I have, ironically enough, is that prediction markets may become less reliable if people trust them too much. (6746)

The danger is that they are trusted too much relative to their liquidity. In absolute terms, it would be fine to go all the way to Robin Hanson’s Futarchy, where prediction markets determine government decisions. But to do that, you need there to be enough liquidity for when people have biases, lose their minds or attempt manipulations.

We’ll wrap up there for today, and tomorrow resume with the wide world of gambling.