3% of Polymarket Users Capture Most Trading Profits

A working paper analyzing every Polymarket transaction from 2023 through 2025 found that 3.14% of accounts qualify as “skilled winners.” The researchers used data on 1.72 million accounts, 210,322 markets and roughly $13.76 billion in trading volume.

The authors define “skilled winners” as accounts whose order flow consistently predicts both short-term price moves and final event outcomes. The paper reports that skilled winners together with market makers make up under 3.5% of accounts but capture more than 30% of trading gains. The study classifies 67% of accounts as unlucky or unskilled losers and reports that group accounts for the platform’s aggregate losses.

To separate skill from luck, the researchers developed a sign-randomization test that re-runs each trader’s history 10,000 times while randomly flipping buy and sell directions. By that benchmark, raw profit and loss overlaps with the skilled group for only about 12% of top earners. The paper finds roughly 60% of accounts labeled as “lucky winners” in one sample reverted to losses when tested on a held-out set of events.

The study reports persistent skill: 44% of accounts classified as skilled in a training sample remained skilled in a held-out sample. The authors ran a parallel test on active mutual funds and found about 10% persistence in that group.

The paper flags potential insider activity. It identifies 1,950 accounts that opened shortly before a single event and went dormant after it resolved; those accounts moved prices roughly seven to 12 times more per dollar traded than skilled traders. The authors say those accounts are concentrated in isolated events and do not account for the platform’s overall accuracy. The paper includes a case study of three accounts opened between Dec. 27 and Jan. 3 that collectively cleared more than $630,000 betting on the ouster of Nicolás Maduro before a U.S. military operation was disclosed. That episode corresponds with a Commodity Futures Trading Commission complaint filed this month alleging insider trading by a U.S. Army master sergeant.

The paper notes Polymarket has been reported to be in talks to raise about $400 million at a $15 billion valuation. It also notes lawmakers in Washington, New York and California have introduced bills or executive orders aimed at limiting insider participation in prediction markets.

The paper quotes Kalshi CEO Tarek Mansour describing prediction markets as harnessing “the power of the wisdom of the crowds.” It also quotes Polymarket CEO Shayne Coplan calling Polymarket “the most accurate thing we have as mankind right now, until someone else creates some sort of a super crystal ball.”

The paper was written by Roberto Gómez-Cram, Yunhan Guo and Howard Kung of London Business School and Theis Ingerslev Jensen of Yale and was revised April 25. The authors state their methods and dataset make it possible to separate skillful traders from luck-driven winners and to quantify how trading gains and losses are distributed across participant types.