Outsider Trading
Prediction markets don't reflect reality, they create it
Someone recently made $34,000 on a long-shot Polymarket bet that the highest temperature in Paris on April 15 would be above 22°C, by figuring out that the bet would be settled based on the readings of a single sensor near the perimeter of the Charles de Gaulle airport runway and then going there and using a hair dryer (or something similar) to heat up the air around the sensor for a few minutes—just long enough to register the elevated reading as the daily maximum (this is illegal, by the way) and win the bet. Apparently the location of the thermometer is common knowledge among the weather gambling community. Matt Levine wrote about it yesterday, framing the outcome in terms of derivative markets more broadly. When a derivative market becomes much larger than the underlying reality it represents, Levine observes, it heightens the incentive to manipulate that market: “You bet a lot of money on the derivative, and then you easily manipulate the underlying thing to make your derivative pay off.” That is, you look for big potential bets on “small realities”—like a single thermometer reading—and then “cause a small change to reality so your bet pays off.” When a fan ran onto the field during the most recent Super Bowl, there were immediate rumors that he had made a major bet on that happening. It wasn’t true (it was just a straightforward promotional stunt) but the fact that it could easily have been is another example of this type of opportunity.
Polymarket calls itself a “truth machine.” It first rose to prominence during the 2024 presidential election by appearing to be a more accurate source of truth than conventional polling; the website’s financial dashboard aesthetic evokes the stock market and thus creates an exaggerated sense of market efficiency. The utopian promise of prediction markets—and yes, it is funny to say that a gambling platform has a utopian promise—is that, as usage grows and every aspect of reality is reframed as a bettable event, we will progress toward an efficient market for the state of everything and the probability of future events, a perfect model of the world derived from the wisdom of the crowd, who are all wagering on everything all the time (this is essentially the what the stock market already accomplishes for financial reality, and what FanDuel/DraftKings metadata does for high-profile sports). Today, it might be elections and wars; tomorrow, the actions of public figures and celebrities; and eventually, granular details of the local and the personal, with an app setting an over/under line on how many minutes you’ll be late to meet me for lunch, and our other friends betting on it. Then the Polymarket truth machine will have fulfilled its destiny, its model of reality complete at the highest possible resolution.
The only problem is that, as you descend this ladder, the manipulation opportunities increase dramatically. Returning to Levine’s argument, elections, wars, and stock prices are “big realities,” virtually impossible to manipulate relative to any potential betting payout. But we have already seen Coinbase’s CEO intentionally recite a string of words on an earnings call because there was a Polymarket bet on whether or not he’d say them. Any event subject to individual control is a comparable “small reality,” all too easy to tweak. Sports betting is similar: It only works at all because professional and collegiate leagues are as regulated and rule-bound as they are, but within that universe too there are bigger realities, like which team wins a game, and smaller realities, like whether a specific baseball pitch is a ball or a strike—the latter being what was manipulated in a recent MLB scandal. “Now prediction markets exist and are a big business,” Levine writes, “so they are not only in the business of predicting reality: They’re also in the business of changing it.” I wrote last week about the ongoing divergence of physical and digital reality, suggesting that most social networks begin by straightforwardly reflecting the outside world to generate their content but grow progressively self-referential. More of what you see on Instagram now is imagery that was created for Instagram. The lament that more and more of the internet is fake is a lazy critique, typically meaning that AI slop and deepfakes are overwhelming and undermining the authentic (as if most of the human-made content is much better). A more valid version of this complaint would be that the internet has encouraged more of the world to disingenuously posture for it, offline as well as online, and that more and more of what we see is made specifically for networked circulation. The fallacy of prediction markets is the belief that distortion and data corruption only flow in one direction; that the physical world is inherently “real” and that only a digital representation of it risks being fake or inauthentic. The internet in its mature form exports plenty of artifice back the other way; a CEO reciting gibberish words on an earnings call is its own kind of slop. Polymarket’s terminal state is being a perfect model of a world it has created itself.
Become a paid subscriber to access more essays (about half of Kneeling Bus posts, and many of the best ones, are for paid subscribers only), or just because you appreciate this newsletter and want to support it.
Reads:
Robert Mariani on American Diner Gothic. “Regional accents and identities and cultures are vanishing. Liquid Internet flows into this void as a mass-vision event, the mythic power of fandom and content overflowing from and washing over millions of minds in placeless America, the new soul of a soulless world.”
The Madison Square Garden surveillance machine. “The atmosphere is so rooted in paranoia that former Knicks players warn one another about rooms being bugged, and staffers worry about being watched when they go out to local bars.”
What fkn kills me is how its so normal can't even watch a hockey game without em telling me to bet all my money