Study: Only 3% of Polymarket Traders Are Skilled; Minority Captures 30%+ of All Gains

Gate News message, April 27 — A new academic paper analyzing all Polymarket transactions from 2023 through 2025 concludes that the platform’s accuracy reflects “the wisdom of an informed minority, not the wisdom of the crowd.” The working paper, revised April 25, was authored by Roberto Gómez-Cram, Yunhan Guo, and Howard Kung of London Business School and Theis Ingerslev Jensen of Yale. The study covers 1.72 million accounts, 210,322 markets, and approximately $13.76 billion in trading volume.

Just 3.14% of accounts qualify as “skilled winners,” whose order flow consistently predicts both short-term price moves and final outcomes. These skilled traders and market makers together capture more than 30% of all gains while comprising under 3.5% of all accounts. Using a sign-randomization test that re-ran each trader’s history 10,000 times with buy/sell directions flipped randomly, the authors found that raw profit-and-loss is a poor proxy for skill: only 12% of top earners overlap with the skilled group, and roughly 60% of “lucky winners” reverted to losses when tested on a separate sample. Skill proved unusually persistent, with 44% of accounts classified as skilled in a training sample remaining skilled in a held-out sample, compared with approximately 10% for active mutual funds. The 67% of accounts classified as unlucky or unskilled losers absorbed the platform’s entire aggregate loss pool.

The authors flagged 1,950 accounts suspected of insider activity that opened shortly before a single event and went dormant after resolution. These accounts moved prices roughly 7 to 12 times more per dollar than skilled traders. A case study identified three accounts opened between Dec. 27 and Jan. 3 that collectively cleared more than $630,000 betting on Maduro’s ouster before the U.S. military operation was disclosed. This aligns with the Commodity Futures Trading Commission’s first-ever insider trading complaint involving event contracts, filed against U.S. Army Master Sgt. Gannon Ken Van Dyke for trading on classified information.

The findings emerge at a sensitive moment for prediction markets. Polymarket is reportedly in talks to raise $400 million at a $15 billion valuation, while lawmakers in Washington, New York, and California have introduced bills and executive orders targeting insider participation. The paper directly challenges industry marketing, quoting Polymarket CEO Shayne Coplan’s claim that prediction markets aggregate information more effectively than experts, calling Polymarket “the most accurate thing we have as mankind right now.”