This article is excerpted from the forthcoming book Wrong Number (Wiley, May 12, 2026)
In 1974, an economist at the University of Nevada–Reno named Bill Eadington called a conference. He invited the obvious people — mathematicians who studied gambling, statisticians who modeled probability, economists who studied risk, psychologists who treated problem gamblers, casino executives, gambling regulators. Notably, he did not invite many finance professors, who in 1974 were largely uninterested in gambling and who would in any case have constituted a different and friendlier audience. The Reno academics were the applied-probability people, and they brought the applied-probability worldview: clean models, closed-form solutions, the assumption that anything resembling a free lunch could not, by virtue of being claimed, actually exist.
Eadington also invited a category of attendee that academic conferences had never previously welcomed: the people who placed bets for a living. Financial traders, card counters, horse-race handicappers, sports bettors, race-track regulars who had figured out a market inefficiency and were quietly making their living off it. The patron saint of this group was Ed Thorp — my friend and one of the few attendees who comfortably bridged both worlds. Thorp was an MIT-trained mathematician who had proved blackjack could be beaten, demonstrated it at the tables, then walked over to the stock market and proved that could be beaten too, and quietly run one of the most successful hedge funds of the 1970s and 1980s. He was the proof of concept that the two camps existed on a continuum rather than in different universes. Most of the attendees, on both sides, treated him as a curiosity rather than as evidence.
The conference ran for decades and was, until well into the 2000s, the only place on earth where the world’s best advantage gamblers and the world’s best gambling academics could sit in the same room and argue.
They argued constantly. They argued about everything. They argued at the bar after the sessions ended. The single best moment in fifty years of running argument came when an advantage gambler, exasperated by an academic’s long debunking of a betting system that the gambler in question had used to pay his mortgage for fifteen years, asked the academic five words.
“Then why are you poor?”
Those five words are the entire history of the divide between people who live on university salaries and people who live by their bets. The standard academic rejoinder is, “Then why aren’t you infinitely rich?”
Both questions are good.
To the gambler, the academic’s objection to the betting system was disqualified by the fact that the academic had not bet on the system, against the system, or on anything else. If you really believed the system didn’t work, the gambler reasoned, you would take the other side. If you really believed it did work, you would take that side. The fact that the academic had taken neither side meant the academic didn’t actually believe his own analysis. He believed only that it was publishable. As Alex Tabarrok said much later, defending Nate Silver against the political-science establishment, betting is a tax on bullshit.
To the academic, the gambler’s claim of fifteen years of mortgage payments from a betting system was disqualified by a different objection. “Then why do you still have a mortgage?” Any claim of a small reliable edge was treated, in academic argument, as if it were a claim that money was free. If you could win 55 percent of your bets, why weren’t you betting more? If you could bet more, why weren’t you a billionaire? And since you were not a billionaire, the claim of a 55 percent win rate must be wrong.
This is how academics at the conference managed, well into the 1980s, to deny that blackjack card counting worked. The mathematics was settled — Thorp had published the proof in 1961 — and the casinos believed it sufficiently to install rules and surveillance designed to stop counters. The card counters at the conference could and did walk down the street and demonstrate the technique in real time at the tables. None of this dislodged the academic position, because the academic position was not really an empirical one. It was a logical one: if card counting worked, the casinos would not exist. The casinos existed. Therefore card counting did not work. The fact that the casinos had visibly adapted, and were now spending serious money to detect counters and bar them, was treated as a curiosity rather than as confirmation.
The honest answer is that both questions point at something real, and the reconciliation between them is the central insight of how markets actually work.
The gambler is right that the academic apparatus around probability and decision-making has spent fifty years protected from the kind of empirical accountability that every other practical field takes for granted. A statistician who has never made a clear, falsifiable, dated prediction against a benchmark and been right about it more often than chance has not, by the standards of every casino floor in the world, demonstrated that he can do statistics. He has demonstrated only that he can write papers about statistics. The peer-review system rewards him for the second skill and ignores the first. By the time he has been promoted to full professor, he has been selected almost entirely for skill at intramural academic politics and almost not at all for skill at predicting the future of anything.
The academic is right about a smaller thing, which is that the gambler’s edge does not generally scale. The advantage gambler making thirty thousand a year on a system he developed in his garage is not an undiscovered George Soros. The reason is not that his edge is illusory — although it often is, there are lots of people who mistake a lucky streak (or even an imaginary lucky streak) for a golden goose — but that the things which generate small reliable advantages in betting markets do not survive the attempt to bet ten or a hundred or a thousand times more on them. The bookmaker who lets you take a hundred dollars a game on his early Tuesday line will not let you take ten thousand. The market maker who lets you pick off his stale quotes for a hundred-share lot will not let you pick him off for ten thousand shares. Real wealth in markets requires either (a) infrastructure — a fund, a desk, a team, capital — or (b) the kind of structural advantage that almost nobody has. The lone gambler with a clever system is, almost by construction, capped at making a nice living rather than a fortune. The academic who infers from this that the edge does not exist is making a logical error. The academic who infers that the edge exists but cannot be scaled to infinity is correct, and uninteresting.
I am one of the lone gamblers, in NFL betting at any rate. In 2006, I built a simple system to predict football games against the spread. ESPN was producing a documentaryI was on the bettors' side. ESPN's producers had me down as a professor because I also publish academic papers, but I was firmly in the wise-guy camp.
My system used three binary factors of the kind professional sports bettors had been using since the 1960s: how each team had performed against the spread to date, whether each team’s most recent game had produced more turnovers than takeaways, and whether the line had moved against each team since the previous week. All inputs were easily available in public. I wrote the system down, posted it on a website, distributed a spreadsheet and an app to do the calculations, and bet a hundred dollars per game on every signal it produced at the same time every week and posted the bets immediately.
I ran the system for fourteen years, from 2006 through 2019. The first three years were the documentary years, and they were not good ones — cumulative profit through 2008 was about seventy dollars. ESPN never aired the documentary. I kept betting. The full track record is 675 wins against 512 losses, a 57 percent win rate against a 52.38 percent breakeven, and net profit of $11,180 on bets of about $120,000. The p-value against the null hypothesis that the system was actually breakeven or worse is 0.01 percent.
That is, by the standards of academic finance, an unusually clean piece of empirical evidence. The system was specified in advance, never modified, applied to a clearly defined universe of bets, and tested against a rigorous benchmark with the actual transaction costs of the actual market. Nobody is ever going to publish it, because nobody publishes results from sports betting. But it is more empirically substantiated than the median paper in the median economics journal.
It is also, importantly, not enough to make me rich. The hard part of sports betting is not picking winners 55 percent of the time against the spread. The hard part is finding bookmakers who will keep taking your bets after they notice you are winning. By the time you can place serious money down, you are no longer placing it against amateurs in a casino but against the professional operators who set the lines and who know exactly what you are doing. The same arithmetic applies in every other market.
And here is where the two questions — why are you poor, why aren’t you infinitely rich — reconcile into a description of how markets actually work. Almost every price in a sophisticated market sits pushed up against some arbitrage limit. Not at fair value, not in equilibrium, not where the textbook says it should be. Pushed up against a threshold beyond which someone with capital and infrastructure would mechanically bid it back.
Imagine measuring a thousand people on height alone: two of them, the tallest and shortest, are extreme. Now measure them on height and income: many more are now “extreme” in some direction — taller than anyone richer, richer than anyone taller. Add a third dimension and a fourth, and eventually everyone is on the boundary of something. Financial prices live in a space of thousands of such dimensions. Most prices are, at any given moment, pushed against the limits of many of their enforced relationships. The unbearable emptiness of high-dimensional space.
My mentor Fischer Black gave this phenomenon a name nearly forty years ago. He called it noise. The market does not converge to a clean fair-value price; it bounces around inside a cloud of arbitrage thresholds, with the boundaries set by the cost and capacity of the marginal arbitrageur. There are profits to be had inside that cloud for people who can see the boundaries clearly. There are not profits to be had for people who think the cloud is a point.
Finance professors, to their credit, eventually worked something like this out. Andrei Shleifer and Robert Vishny published “The Limits of Arbitrage” in the Journal of Finance in March 1997, twenty-three years after the first Reno conference. The paper argued, correctly, that arbitrage in real markets is conducted by specialists with limited capital and short performance horizons, that idiosyncratic risk constrains how aggressively they can lean against mispricing, and that prices can therefore stay wrong for longer than the arbitrageurs can stay solvent. This is most of the right idea. It arrived a generation late, and in academic dress it has been used mostly as an after-the-fact explanation for why a documented anomaly failed to disappear, rather than as a working framework for finding new ones. The advantage gamblers had been operating on the same basic principle since at least the 1960s, with the difference that they used it to make money rather than to excuse their failure to.
The finance professors who took the limits seriously, rather than using them rhetorically, have generally done well out of the insight. Thorp managed Princeton-Newport for nearly twenty years and personally accumulated a fortune. My one-time boss Cliff Asness, a Chicago PhD whose thesis was on the momentum anomaly, founded AQR Capital Management on the explicit premise that the anomalies the literature kept rediscovering were real and exploitable. Jim Simons abandoned an academic career in pure mathematics to build Renaissance Technologies. The pattern is the same in every case: the academic who actually believes his own model goes and trades on it, and the academics who do not believe their own models stay in the literature, where the question “then why aren’t you infinitely rich” cannot be asked of them, because they have never claimed to be even finitely rich.
The applied-probability academics who dominated the early Reno conferences kept treating the market cloud as a point. Either prices were efficient or they were predictably wrong, and either way you didn’t need to put money on the line to know which. Behavioral finance and the limits-to-arbitrage literature have softened that position considerably over the past three decades, mostly by importing into the journals what working traders already knew. But the dominant academic posture toward small reliable edges is still suspicion rather than curiosity. The default move when an advantage gambler or a working trader claims a 55 percent win rate is still to ask why he isn’t a billionaire, on the implicit theory that the only edges worth believing in are the ones that scale to infinity. The advantage gamblers and working traders, for their part, have spent fifty years answering the question, and the answer is always the same: the structural resistance of every other participant to letting them scale.
Why are you poor? Why aren’t you infinitely rich? The two questions sound symmetric, and at the Reno conferences they functioned that way — each side throwing the same accusation back at the other. They are not actually symmetric. The gambler’s question to the academic is the harder one, because the academic has no good answer except a confession that he does not actually believe his own analysis. The academic’s question to the gambler has a perfectly good answer, which is that the gambler’s edge is real but not scalable, and that being right about a small thing is not the same as being positioned to monetize being right at scale. Once you understand that asymmetry, you understand most of what is actually going on in markets.
The 1974 conference is gone now. Eadington died in 2013. The advantage gamblers have mostly retired or moved into finance, where the same dynamics play out at larger scale. The applied-probability academics have moved on to behavioral finance and machine learning and other corners of the literature. The finance professors, the ones I deliberately didn’t mention much above, are running quant funds. Thorp is ninety-three and still publishing. The question “then why are you poor” is still good. Anyone proposing to manage your money should be able to answer it. Anyone proposing to debunk someone else who is managing money should be able to answer it the other way around.
