New Study Reveals Only 3% of Polymarket Users Are Actually Profitable — The Rest Are Funding Them

The premise that has made Polymarket famous — that the collective wisdom of thousands of traders produces more accurate predictions than any single expert — may be significantly overstated.

A new academic study from researchers at London Business School and Yale University has analyzed every transaction on Polymarket between 2023 and 2025, and the findings challenge the platform’s foundational narrative in ways that neither Polymarket nor its competitors will want to highlight.

The study, covering 1.7 million accounts and $13.76 billion in trading volume, arrives at a conclusion that is both simple and devastating: Polymarket’s predictive accuracy is not the product of crowd wisdom. It is the product of a small, skilled minority — approximately 3% of all users — who are systematically extracting money from the other 97%.

The 3% Who Run the Market

The core finding is stark. Using the complete transaction history of the platform, the researchers identified a group of traders — roughly 3.14% of all accounts — who generate the bulk of Polymarket’s price discovery. These are the users whose bets actually move markets toward accurate outcomes. Everyone else is largely along for the ride — and paying for the privilege.

As the paper states directly:

“Using the universe of transactions from a large prediction market platform, we identify these traders and show that they, around 3% of all accounts, generate the bulk of price discovery.”

This elite group, combined with market makers, captures more than 30% of all profits generated on the platform. The remaining 67% of accounts — classified by the researchers as either “unlucky” or “unskilled” — absorb the bulk of cumulative losses. They are, in effect, the liquidity that funds the skilled minority’s returns.

The study also complicates the assumption that profitability equals skill. When researchers examined the most profitable accounts on the platform, only 12% of them showed evidence of genuine strategic ability. The remaining 88% of high-earners had simply been lucky — and roughly 60% of those “lucky” accounts began losing money as soon as market conditions shifted. Skill, it turns out, is rare even among winners.

The Insider Trading Signal Hidden in the Data

The most legally significant section of the research involves what the authors describe as evidence of insider trading embedded in Polymarket’s transaction history.

Researchers identified approximately 1,950 accounts that followed a distinctive pattern: concentrated trading activity immediately before major news events, followed by complete withdrawal from the market once those events concluded. The pattern is difficult to explain through skill or luck alone — it suggests access to information that was not publicly available at the time the bets were placed.

The most concrete example cited in the study involves betting activity around the removal of former Venezuelan President Nicolas Maduro. Significant positions were taken on that outcome shortly before the United States officially announced Operation Absolute Resolve — the military operation that led to Maduro’s detention. The timing of those trades, the researchers note, is difficult to reconcile with any publicly available information at the time.

The connection to the recently charged U.S. Army Master Sergeant Gannon Van Dyke is direct. In April, the Department of Justice charged Van Dyke with using classified military intelligence to place approximately $400,000 in Polymarket bets on the Venezuela operation he was personally participating in. The academic study’s findings suggest Van Dyke may not have been operating in isolation — and that the 1,950 accounts flagged for suspicious pre-event trading deserve serious scrutiny.

What This Means for the “Wisdom of the Crowd” Argument

Polymarket and Kalshi — its closest U.S. competitor — have both built their public positioning around the argument that prediction markets outperform traditional forecasting because they aggregate the knowledge of thousands of independent participants. That argument has been used to justify regulatory acceptance, institutional investment, and the platforms’ growing role in informing everything from financial decisions to political analysis.

The London Business School and Yale research directly contradicts that framing. The accuracy of Polymarket’s predictions, the authors conclude, reflects “the wisdom of an informed minority, not the wisdom of the crowd.” The crowd is not producing the signal. The crowd is the noise — and in many cases, the funding source for those who have real information.

This distinction matters enormously for how prediction markets are regulated, marketed, and understood. If the accuracy of these platforms depends on a small number of informed — and potentially illegally informed — actors, then the case for their social utility looks fundamentally different from what their founders claim.

The Uncomfortable Conclusion

For ordinary users on Polymarket, the study’s message is blunt: if you are not in the 3%, you are almost certainly funding someone who is. The platform works as a prediction tool at the aggregate level — but the mechanism producing that accuracy is concentrated knowledge and, in some cases, information that was never supposed to be public.

The crypto and prediction market community has watched the Van Dyke arrest, the academic findings, and the ongoing Arizona criminal case against Kalshi unfold within weeks of each other. Taken together, they paint a picture of an industry at a regulatory inflection point — one where the gap between what prediction markets claim to be and what the evidence suggests they are has become impossible to ignore.

The research is publicly available. The Van Dyke case is before the courts. And the 97% are still placing bets.