A groundbreaking study by researchers from London Business School (LBS) and Yale University has overturned conventional wisdom about prediction market accuracy. The paper, released on the Social Science Research Network (SSRN) on April 20, 2026 and revised on April 25, analyzes the complete trading history of Polymarket, the world's largest prediction market by volume, and finds that the platform's celebrated accuracy is not due to a broad 'wisdom of crowds' but rather the actions of a tiny minority of skilled traders.
Massive Dataset: 98,906 Events, 172 Million Accounts
Led by Roberto Gomez-Cram, Yunhan Guo and Howard Kung of LBS, along with Theis Ingerslev Jensen of Yale, the researchers examined 98,906 events across 210,322 markets, representing more than $13.76 billion in total trading volume and approximately 1.72 million accounts. To separate skill from luck, they employed a statistical technique called 'sign-randomization testing,' which identifies traders whose returns are statistically unlikely to be random.
Only 3.14% of Accounts Are Skilled
The results are stark: merely 3.14% of all Polymarket accounts were classified as skilled winners. These traders averaged activity in 79 markets, consistently built positions aligned with final outcomes, and maintained their skill classification 44% of the time in out-of-sample tests (compared to just 10% for skilled mutual funds). The remaining 96% of accounts either broke even through luck or incurred losses.
Notably, the order flow from these skilled traders predicted both next-period price movements and final market outcomes with high statistical significance. A 1-percentage-point increase in skilled net buying correlated with an 8-basis-point rise in the probability of correctly forecasting the final result.
Explosive Growth but Persistent Concentration
Polymarket's monthly trading volume exploded from $3.3 million in December 2023 to $1.98 billion in December 2025—a roughly 600-fold increase in two years. Active accounts surged from about 1,600 to over 519,000. Yet the concentration of skill remained remarkably stable, with the same small cadre of informed traders dominating price discovery throughout the period.
Insider Trading Evidence and CFTC Action
The study also investigated potential insider trading. Researchers identified 1,950 accounts whose trading patterns—timing and conviction—strongly suggested possession of non-public information. These accounts earned an average of roughly $15,000 each and caused large price swings upon trading. A striking example: on January 3, 2026, just hours before a U.S. secret military operation related to Venezuelan President Nicolás Maduro, three accounts took positions in Maduro-linked contracts, eventually profiting over $630,000 in total.
On April 23, 2026, the U.S. Commodity Futures Trading Commission (CFTC) filed a complaint alleging that one of those accounts belonged to an active-duty U.S. soldier engaged in insider trading. Separately, the CFTC also sued the State of New York over its actions against prediction markets, escalating regulatory tensions. The study's authors, however, concluded that insider trades are too isolated to account for the broad price discovery observed across the platform.
The Masses Are Merely Supporting Cast
The paper describes the broader participant base as the 'supporting cast.' Unlucky or unskilled losers constituted 67% of all accounts and absorbed the entirety of total losses. Meanwhile, market makers and skilled buyers—together making up less than 3.5% of accounts—captured over 30% of the total profits.
Researchers conclude that prediction market accuracy reflects the behavior of a small but identifiable group of informed traders whose participation is the core mechanism of price formation. Whether these key traders will continue to participate as platform fees rise and regulatory scrutiny increases remains an open question for future research.