WTF Is Going On: Prediction Market Losers, Part II
Wondering who wants to make weird, long-term bets in a possibly rigged market
(You can see part I on prediction markets here. And for more on Musk, check out The Atlantic Current, where on Tuesday morning Tull and I dove deep into Musk’s history and his current role as European boogeyman. It’s honestly the best one we’ve done: both affecting and of ten quite funny.)
Last week, in discussing prediction markets, I argued that the core problem for the business model is finding and maintaining a proper pool of losers. If prediction markets are going to surface information, they need to provide financial incentives to do so. And those incentives have to be paid by someone.
‘Someone’, in this case, basically has to be losing gamblers. There are probably some weird corner cases: might a high-income individual bet on, say, a Democratic sweep in 2028, providing a hedge for possible higher income taxes? Might an ardent Trump supporter make the same bet, with winnings funding the purchase of the whiskey needed to get through the ensuing years of left-wing control (or a plane ticket to Hungary, etc.)?1
But overall, for prediction markets to run billions of dollars in volume annually across literally thousands of markets, there needs to be a large pool of losing money. And, as we wrote last week, it’s not insane to predict (pardon the pun) that such a pool will exist. There are people happy to lose money in all sorts of gambling contests, and have been for literally thousands of years. It’s perfectly possible that the 2026 midterms garner interest in the same way as scratch-off lottery tickets and the Eagles minus three and a half.
Certainly, that’s the bet that the investors in these markets are making. They have already raised billions of dollars. U.S. online gambling giants DraftKings and FanDuel both are entering the market. The rise of these markets clearly has impacted those stocks as well: investors worry that prediction markets will be able to peel off some sports-related volume (since, for reasons that are not 100% clear, the Commodity and Futures Trading Commission is not enforcing a seemingly obvious, legislatively-enacted, ban of sports “event contracts”, as they’re now).
The problem is, however, that the more closely you consider that bet, the weaker it seems. There are multiple reasons why prediction markets look much more like a fad than an innovative way to bring information to society.
The Insider Trading Problem
Insider trading in commodities markets (and, again, prediction markets are CFTC-regulated providers of event contracts, similar to futures) is illegal, but not in the way in which stock market investors are familiar. Most cases center on breach of duty to employers, rather than the disclosure of information that might actually move prices if they were widely known. This is in part because so many participants in the commodity markets are acting on their own knowledge which would move prices. For instance, a crude oil company might decide to lock in a higher future price because it knows its own production will be higher than expected, thus leading to lower prices. And that’s part of the point of futures market: to provide that certainty, even if a counterparty might be playing on a less-than-level playing field2.
In other words, insider trading in prediction markets isn’t necessarily illegal (not legal advice at all!). It is banned by terms of service for some platforms, including Polymarket. But when big trades hit the news — as in the $400K winnings on the Venezuelan invasion that we highlighted in Part I — it certainly seems as if insiders are playing. And whatever the law says, those insiders are going to be incentivized to play. (This is particularly true on Polymarket, a crypto-based platform where users can probably do a pretty good job covering their tracks.)
The problem for prediction markets is that, once again, their goals conflict. Again, the idealistic argument from prediction markets and their supporters is that they can surface information that is of value to society more broadly. And that way you do is by incentivizing people with that information to place bets with a positive expected value3. This in turn moves the market, providing the price signal that tells even neutral observers that the odds are changing.
If a White House staffer walks out of a meeting at which a Venezuelan invasion is discussed, and can get, say, 14 to 1 odds that an invasion is going to happen within a month, she might bet on that outcome, moving up the price. If multiple White House staffers do so, the price goes higher, and the signal gets stronger. Of course, it doesn’t have to be inside information: if a trader devises a better forecasting model for political races or weather or economic output, it can take positive EV bets which, again, push up the price and provide the signal that a Republican is going to win Florida’s 13th Congressional District or that average temperatures in Paris will be below average in February.
But insider trading presents a thorny problem here. From the perspective of getting information to the public, insider information obviously is a good thing. Indeed, you can see (albeit perhaps from a capitalist/libertarian perspective) an argument that financially incentivizing White House staffers to anonymously leak is good for society. Or, perhaps in a less flagrant scenario, that it’s beneficial to incentivize the indirect dissemination of even rumors, or conversations overheard at a Dupont Circle bar in D.C. Or, to use a less tangibly life-or-death example, for those with access to internal polls to bet on elections (the bets will go in the direction of those polls, meaning the signals created in the polls now can reach a much wider audience).
There are legal and moral objections to those arguments, to be sure. For instance, our hypothetical White House staffer is quite clearly breaking the law. Of course so are those who leak classified information to journalists. Doing it for money feels like a grubbier explanation (and doing it for crypto somehow even worse), though one imagines that the axe-grinding, back-stabbing, and/or general pettiness that drive many leakers is not always righteous. And from the point of prediction markets, the core argument for their existence as a public good is that they provide a channel for this high-value information to reach the public (if again, not in as direct a manner as a New York Times report that the White House reviewed plans for a strike in Venezuela).
But if insider trading is banned and/or illegal, then who, exactly, are the sources of high-value information in the types of markets (mostly political) that are of the most utility to society more broadly? There is, perhaps, a “wisdom of crowds” argument that even mostly-uninformed people betting on invasion markets or the 2026 midterms will, in sum, provide better data than a random poll or a review of op-ed pieces. That’s a much weaker argument, however, than pro-prediction market types make.
And if insider trading is allowed in some form — or, as is more likely, isn’t able to be thwarted — then that in turn disincentivizes the gambling losers that the platforms need for the liquidity to pay off the insider traders. It doesn’t push them all away: it’s stunning how many people bet on U.S. sports only to promptly complain on X that the games are manipulated, and I fondly remember my days as a poker player when losing gamblers at the table would fervently insist that online platforms too were rigged. (It was always amusing when I asked why they thought that, and they responded, “because I played on them for years!”.)
Still, insider trading and low-information liquidity work at cross-purposes, and create a real roadblock to the platforms actually surfacing information that is truly valuable to society. The people who have that information need to be compensated on these platforms to provide it; but the gamblers who are offering the necessary compensation have to believe those people aren’t on the platform at all. It’s hard to see how both those states exist at the same time.
Who Is Betting On Politics And Warfare?
Right now, prediction market volume is coming mostly from sports: Kalshi reports the proportion is about 90 percent. This is in large part because there are 19 states that don’t allow online sports betting (most notably California and Texas), and so Polymarket and Kalshi and (online brokerage!) Robinhood provide ostensibly legal alternatives4.
Over time, that proportion needs to come down markedly. There isn’t much societal benefit to knowing the odds of the Philadelphia Eagles winning the 2027 Super Bowl; the point of prediction markets is for volume to develop in markets that have more effect on modern life (yes, I know Eagles fans would dispute that any such markets exist).
But that gets to the second problem here, which is that if these markets need losers, why are the losers betting on these markets? Getting 10 to 1 odds that the U.S. will invade Colombia this year does not seem anywhere near as broadly attractive as getting 10 to 1 odds that Oregon will win the college football championship. There no doubt are some users who’d rather bet on the former than the latter, but over 20 million people watched Monday’s college football championship game, and it’s likely that fewer of that could place Colombia on a map.
And that problem exists even for markets that are broad and large. At the narrow level — again, these platforms are adding tens of thousands of markets a month — the problem gets even more intense. Who is betting on a specific Congressional race, or the success of an odd piece of legislation, or a more arcane foreign policy question? At the risk of being repetitive, it’s information around those events, not the Super Bowl, that prediction markets are supposed to provide. And it’s hard to imagine who on Earth is possibly betting on these, particularly once the novelty of prediction markets wear off.
If those bets don’t arrive, then nearly all non-sports markets wind up with minimal volume from low-information users. And that in turn gets to the core problem: that means they have minimal volume, period. Users with the ability to capture the information needed to beat those markets (whether through research, new models, or inside information) don’t have anyone to trade with but each other, which means there usually isn’t a price at which to do a trade. More importantly, there isn’t the volume needed to incentivize research/modeling/lawbreaking, and so those steps aren’t undertaken and the information isn’t discovered or acted upon.
Theory and Practice For Prediction Markets
And this, at least at the moment, seems to be the fatal flaw for prediction markets. The idea underpinning them is attractive: by allowing traders to “put their money where their mouth is”, as Polymarket literally has phrased it, you create real-time odds of events happening. The very existence of those odds itself is a powerfully positive social externality: those odds represent a market-based, unbiased, probability of important events happening. And businesses, governments, and individuals can use those odds to improve their own decision-making, just as the liquid market for 2027 soybean prices helps farmers do the same.
But there are simply all sorts of problems with this theoretical model. There’s the “how, exactly, does this get settled?” issue which keeps popping up: for example, it appears that the possibly-insider trader who ‘won’ $400K on the Venezuelan operation didn’t win after all, because Polymarket ruled it didn’t count as an invasion. There’s the problem of interest rates: if you’re betting on the 2028 presidential election, you’re giving up ~10-12% returns relative to letting that same cash just sit in the bank5.
There’s the risk of market manipulation: as Saahil Desai recently pointed out in The Atlantic, well-timed trades could move prediction markets which in turn would capture media attention and thus potentially influence the race. (This is a bigger issue the lower volume goes: if you bet $250,000 on your candidate, or even yourself, to win a tight Congressional race, Polymarket odds for your candidate would suddenly rise sharply, an act which itself would capture attention, dampen enthusiasm for your challenger, improve fund-raising, etc. The actual return on investment there could be very high in multiple ways.)
But at the end of the day, the idea that prediction markets can matter beyond the most obvious use cases — major elections, huge foreign policy decisions, etc. — relies on there being real volume in each of these individual markets. And the only way to get real volume is essentially for gamblers to want to be on them. Given that ~90% of volume is coming from sports, data from the platforms themselves suggests that those gamblers would much rather bet on Josh Allen than on Josh Shapiro. If that doesn’t change, the theoretical case for prediction markets will run into immovable practical roadblocks.
As of this writing, Vince Martin has no positions in any companies mentioned. As written elsewhere, he may buy stock in DraftKings at some point in the near future.
Obviously, it seems far more likely that the opposite would be true: a Trump supporter would bet on the GOP candidate. This is true in sports betting as well, by all accounts, which has always struck me as bizarre: why double down on your favorite team instead of using gambling money to spice up a game in which you have no real interest?
As we noted in Part I, there are also basically zero individual investors in the futures market, and so the “level playing field” goal that drives so much of equity market regulation simply isn’t relevant.
Ie, in theory, if you made that bet in an infinite number of universes, on average your winnings would be positive.
Gamblers in those states can bet at offshore books, but they are much more difficult to transact with; most payments nowadays are done in crypto.
To be fair, you can exit that position early, but low-information gamblers probably aren’t betting on that outcome with the plan to exit in September.