I wrote about prediction markets in these pages last October, when the discussion at G2E had reached a pitch I had not encountered in some years. Unusually for me, my conclusion then was measured: These platforms deserve scrutiny, but the data is still thin. The data is no longer thin and I am no longer of a mind to be measured.
The last time a financial innovation attracted this level of breathless enthusiasm, people were buying tranches of American mortgage debt and calling it a public service. Prediction markets are not quite that bad, but the gap between what these platforms claim to be and what they actually are is considerable. And it is widening by the day.
The case for prediction markets rests on the wisdom of crowds, first promulgated by Sir Francis Galton in 1907. However, I tend to believe prediction markets show the wisdom of money — more on that another day. Anyway, the wisdom of crowds is the idea that aggregating dispersed beliefs produces more accurate forecasts than any individual expert can provide. Kalshi, the dominant U.S. platform, correctly predicted a high likelihood of a Trump victory in 2024 when most polling suggested a coin flip. This is cited constantly, usually by people with a financial interest in your believing it. Well, they would, wouldn’t they?
The problem is that the product bears little resemblance to the pitch. Kalshi does not operate a peer-to-peer market in which a crowd freely discovers prices. It operates a quote-driven market in which prices are set almost entirely by professional market makers, including an army of in-house traders under the banner of Kalshi Trading. The company simultaneously owns the exchange, the clearinghouse, and one of the primary market makers operating on it.
A federal class-action lawsuit alleges that these affiliated market makers enjoy reduced fees, higher position limits, and privileged system access unavailable to ordinary users. There is a suspicion that Kalshi’s own market makers are given a privileged position. Kalshi disputes this; a company official stated that the role of market makers is to react to pricing, not to set or influence it. That is not the role of market maker I am used to.
None of this would matter quite so much if the fees were modest. They are not. Kalshi’s fee formula produces 4% of the stake at the 50-cent price point where sports markets cluster and this figure is widely quoted. But it is the best case for the participant as far as fees are concerned.
I should explain that each contract has a face value of $1. The price you pay reflects the market’s implied probability of the outcome. To put $5 on an even-money event, I would need to buy 10 contracts at 50 cents each. If I win, I collect $10, netting a $5 profit. To put $5 on a roughly 10-to-1 shot, I would need to buy approximately 55 contracts at 9 cents each. If I win, I collect $55, netting around $50. The price of the contract is, in effect, the bookmaker’s odds expressed as a probability
At 10 cents, the fee formula rounds up to 1 cent, Kalshi’s minimum fee, which is 10% of the stake. At 5 cents, it is 20%; the higher the odds, the bigger the fee percentage. The 4% figure applies to 50-cent sports contracts; the cheap longshot contracts that retail participants disproportionately favour carry fees two to four times higher. I am not aware of another product in regulated gaming that presents a 20% house advantage or more as a minor technical detail buried in a formula.
The result is that sports contracts, benefiting from the largest volumes even at the lower fee rate, generated 89% of Kalshi’s fee revenue in 2025, which came to $263.5 million for the full year. By early 2026, the annualised run rate had risen to approximately $1.5 billion, having grown sixfold in the second half of 2025 alone; in the U.S., major sports events are not evenly spread throughout the year.
At Kalshi’s current valuation of $22 billion, investors are paying roughly 14 to 15 times annualised revenue for what is, when you strip away the language of financial markets, a fee-extraction business that has secured a federal licence to present itself as something considerably grander. Fourteen times revenue is a generous multiple. For a company whose regulatory protection has yet to survive a serious legal challenge, it is either visionary or optimistic. I believe the Supreme Court will make the ultimate determination.
The academic evidence on what participants actually earn from these markets is not comfortable reading for prediction market advocates. The most rigorous study to date is “Makers and Takers: The Economics of the Kalshi Prediction Market” by Burgi, Deng, and Whelan of University College Dublin, drawing on transaction-level data across Kalshi’s entire operating history from 2021 through April 2025.
The paper finds that contracts priced below 10 cents return less than 40 cents on the dollar on average, winning at roughly a third of the rate their price implies. This is not simply because cheap contracts lose more often: A 10-cent contract is already a 9-to-1 wager and the buyer expects to lose nine times in 10 by construction. The problem is that they lose even more often than the price suggests, because the price is wrong. So much for the wisdom of crowds.
The paper finds that Takers (bettors), predominantly retail bettors buying cheap longshot contracts, average a return of minus 31.46%. Makers (market makers, or bookmakers in my language) average minus 9.64%. Both sides lose, because Kalshi extracts fees from every transaction regardless of outcome and because cheap contracts carry fee rates from 10% to 20% of stake or more, not the 4% that is usually quoted. The real “maker”, in any meaningful sense, is the platform.
The above is a critique of product quality and structural fairness, and it is in principle addressable through regulation, disclosure, and competitive pressure. The next problem is not, because it concerns what happens to the information these markets claim to aggregate once the financial stakes become large enough for participants to consider it worth their while to manufacture outcomes rather than predict them.
In March 2026, Emanuel Fabian, a military correspondent for The Times of Israel, began receiving death threats. His offence was reporting that an Iranian ballistic missile had struck Israeli soil on 10 March. Gamblers on Polymarket had wagered more than $14 million on whether Iran would strike Israel that day and those who stood to lose flooded Mr Fabian with demands that he rewrite his story, characterising the missile as intercepted debris, so their contracts would pay out. He refused and contacted the police. Separately, a U.S. Army Special Forces soldier was charged with using classified information about the operation to capture Venezuelan President Nicolas Maduro to win $400,000 from wagers on Polymarket.
Much of the regulatory hand-wringing about insider trading on these platforms is somewhat misplaced: If large and unusual sums appear on one side of a market before a geopolitical event, the rational response for any sophisticated counterparty is to decline to take the other side. Caveat emptor.
More unsettling is the deeper logic: A market valued for its ability to aggregate information creates, at sufficient scale, a financial incentive to corrupt the very information it is aggregating. The crowd is wise until it is expensive to be honest.
The most recent illustration involved the weather contracts at Paris Charles de Gaulle airport. In April 2026, French authorities launched a criminal investigation after Meteo-France, the national meteorological service, filed a complaint reporting that a sensor at Charles de Gaulle had been tampered with. Temperature readings jumped by approximately six degrees Fahrenheit in seconds on the evenings of 6 and 15 April, on both occasions setting the daily high at precisely the moment required to trigger payouts on Polymarket weather contracts.
Total wagers across the two affected days were nearly $1.4 million, far above typical volumes. Expert analysis suggested the use of a “calibrated portable heating device”, hair dryer or cigarette lighter, perhaps. Ruben Hallali, chief executive of the weather risk company Sereno, observed: “The Charles de Gaulle incident is not an isolated curiosity. It is what happens when financial incentives meet fragile data infrastructure”. Temperature data at airports are not just numbers, they inform take-off distance calculations, climb-rate parameters, and decisions about whether to apply frost treatment to aircraft. Polymarket’s response was to switch the relevant contracts to a different nearby airport. Not to suspend trading, nor to void the payouts, but to switch airports and move on.
Those of us who have spent careers in the regulated gaming industry have been here before, though not quite at this scale and not with these geopolitical dimensions. The match-integrity infrastructure that now sits beneath regulated sports betting was built, expensively and over decades, because operators discovered that financial stakes on event outcomes create systematic incentives to corrupt those events.
Prediction markets have been granted the commercial benefits of event betting at unprecedented scale, while being exempted from the consumer-protection obligations, responsible-gambling mandates, and integrity-monitoring infrastructure that regulated operators are required to maintain. And let’s not forget taxes. The American Gaming Association is not wrong to find this asymmetry objectionable.
The question is not whether prediction markets represent a new form of price discovery. Some of them do, for some categories of events, some of the time. The question is whether a market that delivers minus 20% average returns to its participants, charges retail participants fee rates of 20%, possibly more on the contracts they favour most, sits on a $22 billion valuation predicated on regulatory protection that has not yet been seriously tested, and has demonstrably created incentives to threaten journalists and tamper with airport infrastructure warrants the characterisation of financial innovation that its advocates have been so generous in supplying.
The crowds, it seems, are wiser about some things than others.
