Interview with Brad Lipton: Ordinary Prediction Market Users Do Not Know Their Opponent

Who sits on the other side of the trade

Lipton's central concern is the counterparty. In his account, ordinary users are often on the other side of wagers from professional traders who use sophisticated models, including AI-built models, and place thousands of trades a day. In his study he cites a Wall Street Journal report of one trading firm spending more than $200,000 a year on data feeds and AI tools.^[2] How large that group is, he says, cannot be established without user-level data, which Kalshi does not publish.

His study, published by the Roosevelt Institute on 7 July, relied on detailed Kalshi trade data that was publicly available at the time through Dune, an official Kalshi data partner. That level of detail now sits behind a paywall reported to cost around $40,000 a year.^[2] Daily summaries remain free.^[6] Based on more than 400 million trades, the study estimates that takers lost $583.5 million between Kalshi's launch in July 2021 and May 2026. The figure is the net result of all trades on the accepting side, which the study treats as ordinary users.^[2]
Maker and taker are Kalshi's own order-book terms, borrowed from stock-exchange practice: a maker posts an offer and waits for it to be matched, while a taker accepts an existing offer. Lipton argues that what a user sees on their phone is a list of offers, with no information about who stands behind them.

"It isn't clear that everyday participants in prediction markets know how sophisticated the person or the methods on the other side of the transaction are," he writes.

Kalshi rejected the analysis in a response published on 10 July. It argues that maker and taker describe how an order is matched, not who places it: casual app users' resting orders can be classified as maker trades, while institutional high-frequency trades can register as takers. Kalshi said its top-earning traders include a public schoolteacher from Pennsylvania, an IT professional in Minnesota and a retiree in Kansas, and pointed to figures it had already published, roughly one profitable user for every three who lose, as evidence of a fair marketplace. It also alleged the report reflected the casino industry's influence. The response did not release new trading data.^[3]
Lipton says Kalshi gave only a few examples of casual users who had done well, rather than new data to disprove the study.

Independent research points in the same direction, though it measures something different. Economists Constantin Bürgi, Wanying Deng and Karl Whelan at University College Dublin analysed more than 300,000 price observations on Kalshi contracts and found that makers earn higher returns than takers, and that cheap contracts priced below 10 cents deliver very large average losses, particularly for takers. Whelan's practical advice to traders is to be wary of low-priced "lottery ticket" contracts. The Dublin study measures returns per contract rather than dollar losses, and does not identify who the makers are.^[7] Analysts at Citizens JMP Securities reached a similar conclusion in March: because prediction markets, unlike sportsbooks, do not limit or ban winning players, retail traders are directly exposed to professional bettors and market makers, and fared worse than sportsbook customers.^[8]
Lipton frames the issue as one of disclosure. If users knew that the person on the other side was backed by a hedge fund and placing thousands of similar winning bets a day, he asks, would they still accept the offer?

Hedging tool or betting slip

Asked whether a prediction market is a form of gambling, a financial market, or something new, Lipton says it can be both. Prediction markets cover a range of activities, he says. When casual users bet on sports or the weather, the activity looks like gambling, even if sophisticated traders place some of those bets. When institutions sit on both sides of a contract, it looks more like a financial market: a tool for hedging risk. In his view, that sophisticated activity does not change the fact that people can bet on almost anything on these platforms, and much of it looks like gambling.

Lipton sees the clearest difference from traditional gambling in what can be bet on. A gambling operator needs a state's approval before offering a type of bet, which naturally limits what can be wagered: Nevada keeps a list of approved games, and New Jersey's constitution prohibits wagering on college games held in the state or involving its teams. Prediction markets work the other way round. Those registered with the Commodity Futures Trading Commission (CFTC), such as Kalshi, self-certify their contracts, arguing that they are swaps under federal law, and can list them without prior approval. The CFTC can still review and block a contract, as it did with Kalshi's contracts on control of Congress in 2023, a decision a court later overturned. Lipton questions whether the CFTC is using its oversight powers sufficiently to scrutinise the expanding range of event contracts. The federal government has also considered limited additional restrictions, such as banning bets on player injuries.^[9]

The range of contracts, he argues, affects even people who never place a bet. When money rides on almost any real-world event, someone has an incentive to change the outcome, making it hard for anyone, including the police, to keep track of what is going on. In one case, a staff member at the Institute for the Study of War allegedly edited its map of the war in Ukraine just before a Polymarket bet that relied on the map was settled.^[10] In traditional sport, by contrast, the eight Chicago White Sox players accused of throwing the 1919 World Series for gamblers were banned for life. Federal rules against fraud, manipulation and some forms of insider trading do apply to prediction markets, and the CFTC has fined and banned traders, among them former congressman George Santos.^[11] Lipton's question is whether those rules, and the oversight behind them, can keep pace with markets where almost any event can become a bet. As an example, he points to university admissions, where Kalshi has filed to list contracts on admission rates at selective colleges.

Disclosure is at the heart of his argument. Licensed gambling operators have to disclose information to players to stay compliant: in Nevada, for example, state rules require casinos to display payout schedules on the table or machine, or right next to it. On a sportsbook such as DraftKings, he says, it is clear that the user is betting against the operator, and the odds are shown.

Prediction markets argue that they fall under federal rules for swaps rather than state gambling law, so these requirements do not apply to them. Their platforms tell users what they will be paid if they win, but, he writes, they do not disclose how often people win or lose on such contracts, or who is on the other side.^[9]

Safer gambling: different rules for prediction markets

In Lipton's view, US problem-gambling measures are weak even for licensed gambling operators, resting largely on self-exclusion lists and helplines. Even so, Nevada requires casinos to display information about problem-gambling support and to train staff. Prediction markets, he says, operate with essentially no such rules; Kalshi's self-exclusion programme is voluntary. At the same time, he argues, the platforms pay influencers to bombard ordinary people with advertising on their phones, content that he writes may not immediately look like advertising.^[9]

A legal tide that has turned

Until recently, Lipton says, there was a lot of talk about the Supreme Court settling whether states can regulate prediction markets, with the courts divided. In his view, the picture is no longer as split, and the states are winning. The Third Circuit ruled 2–1 for Kalshi in New Jersey in April.^[13] The Ninth Circuit in Nevada in August and the Sixth Circuit in Ohio and Tennessee in September both ruled against it, 3–0.^[14][15] All three rulings concern preliminary injunctions rather than final judgments, and the courts remain divided. New Jersey has asked the Supreme Court to hear its case. The court has not yet decided whether to do so, and Kalshi's response is now due on 7 December 2026.^[16]

Views earlier in the year were different, and show how quickly the picture has moved. We spoke to Elizabeth Lan Davis, a former CFTC chief trial attorney now at the law firm Davis Wright Tremaine, in the summer, before the Ninth and Sixth Circuit rulings. Everyone was talking about prediction markets reaching the Supreme Court, she said then, but she expected a prolonged legal process rather than a quick nationwide answer. She also argued that, provided the CFTC is found to have exclusive jurisdiction, the framework then being built would not be easily undone by a future government, much as the Dodd-Frank derivatives rules were later streamlined rather than dismantled. In February, Jefferies analyst David Katz described prediction products as a difficult, narrow space with limited participation and no clear path to regulatory certainty. The scale of the market has changed since then: Kalshi recorded more than $10 billion in notional trading volume over the weekend of 24–25 October 2026, according to Aldrin Research data reported by Covers.^[17] This measures the face value of contracts traded, rather than the cash paid for them.
Asked what has surprised him most, Lipton points to how large prediction markets became, how fast, and how much money ordinary people have lost.
Whatever the courts and the CFTC decide, the question at the heart of his work remains: does the person accepting an offer on a prediction market know that it may come from a professional trader using sophisticated models, some built with AI, and placing thousands of bets a day?