A new academic working paper has upended one of prediction markets’ most repeated claims: that accuracy emerges from the collective wisdom of large crowds. The supporting evidence appears in the filing.
Researchers from London Business School and Yale analyzed every Polymarket trade placed between 2023 and 2025 and found that just 3% of traders are responsible for most price discovery on the platform.
The study, authored by Roberto Gómez-Cram, Yunhan Guo, Theis Ingerslev Jensen and Howard Kung, examined 1.72 million accounts and $13.76 billion in trading volume.
Their conclusion is direct: a small cohort of informed traders moves prices toward correct outcomes, while the remaining 97% provide liquidity and volume but consistently sit on the losing side of those trades.
Skill Versus Luck in a Million-Account Marketplace
Separating genuine skill from statistical luck across such a large user base required a rigorous method. The researchers reran each trader’s historical bets 10,000 times, holding every variable constant except direction.
The same markets, the same timing, the same dollar amounts were used, but a coin flip determined whether the simulated bet went long or short.
That simulation produced a baseline showing what profits would look like with zero real informational edge. Traders whose actual returns consistently beat the coin-flip benchmark were classified as skilled.
Those whose returns matched or trailed it were classified as lucky. Among the biggest winners ranked by raw profit, only 12% cleared that bar.
The durability of apparent skill also proved fragile under scrutiny. Roughly 60% of traders initially labeled lucky winners turned into net losers when the researchers tested their records against a separate sample of events.
That regression to the mean weakens the case that broad participation, rather than a concentrated informed minority, is what keeps prediction market prices accurate.
The Green Beret Case as a Data Point
The study’s framing draws an explicit line to a recent scandal. A U.S.
Army Green Beret was arrested after allegedly betting on the outcome of a classified American military raid, with his Polymarket positions moving in the right direction before public confirmation. The authors treat that case not as an isolated incident but as an extreme version of the dynamic their data describes.
The incident involved a bet tied to the potential removal of Nicolás Maduro from power, a politically sensitive event with clear national security dimensions. The researchers argue the soldier’s alleged behavior illustrates how informed insiders, not the crowd, are the actual price-setting agents on platforms like Polymarket.
For prediction market advocates, the findings create an uncomfortable tension. The markets do appear to generate accurate prices, but the mechanism behind that accuracy is concentrated information held by a small elite rather than distributed crowd intelligence.
The 97% majority effectively subsidize the profits of the 3% who actually know more.
Whether that model undermines the legitimacy of prediction markets as a forecasting tool or simply reframes how they work remains an open question. What the data makes harder to argue is that participation volume alone drives accuracy.
Not Financial Advice: This article is for informational purposes only. Crypto investments are highly volatile. Always do your own research.
