- A study of 1.72 million Polymarket accounts and $13.76 billion in volume found that just 3.14% of traders drive nearly all price accuracy.
- Only 12% of the biggest winners by raw profit were actually skilled—the rest just got lucky, and 60% of them became losers on a second test.
- Insider trades move prices 7 to 12 times more per dollar than skilled trades, raising uncomfortable questions about where information actually comes from.
The “wisdom of the crowd” is the pitch. The data says something else. A working paper from London Business School and Yale, analyzing every Polymarket trade from 2023 through 2025, concludes that roughly 3% of accounts are responsible for most price discovery—meaning they are the ones actually pushing prices toward the correct outcome. The remaining 97% provide liquidity and generate volume. In aggregate, they lose money to the informed minority.
Researchers Roberto Gómez-Cram, Yunhan Guo, Theis Ingerslev Jensen, and Howard Kung analyzed 1.72 million Polymarket accounts and $13.76 billion in trading volume. They classified “takers”—accounts that primarily submit market orders—into four groups: skilled winners (3.14%), lucky winners (29.0%), unlucky losers (61.4%), and unskilled losers (6.4%). The remaining 0.1% are market makers posting limit orders. That breakdown is less a crowd and more a casino where the house is three percent of the players.
“Skilled traders’ order imbalance is a strong predictor of both next-period price changes and final market outcomes,” the researchers write. “Lucky winners, despite their positive profits, have no significant predictive power.” The distinction matters because Polymarket and its competitors sell themselves on the idea that broad participation produces accurate forecasts. If accuracy depends on a handful of informed actors, the regulatory argument for these platforms looks very different.
How the Researchers Separated Skill From Luck
With over a million traders on Polymarket, plenty will rack up big winnings by chance alone. The authors’ solution: rerun each trader’s bets 10,000 times, keeping everything the same—same markets, same moments, same dollar amounts—but replacing the direction with a coin flip. If a trader’s actual returns consistently beat the randomized benchmark, that is skill. If not, it is luck. The method is brutal in its simplicity. Among the biggest winners by raw profit, only 12% cleared the benchmark. And roughly 60% of apparent winners became losers when their performance was tested against a separate sample of events.
Previous research aligns with these findings. A 2025 study from Technische Universität Berlin confirmed that 70% of Polymarket traders have negative profit-and-loss tallies. More recent data from researcher Andrey Sergeenkov puts the losing share at 84.1%, with just 2% having made more than $1,000 on the platform. Prediction markets are accurate. They are just not accurate because of the crowd.
The skilled minority reacts first to new information—shifting positions in response to Federal Reserve announcements, corporate earnings, and geopolitical events—while other traders show little consistent reaction. When skilled participants account for a larger share of trading, prices move closer to the correct outcome, especially in the final stretch before resolution. The crowd’s contribution is volume, not insight.
The Insider Problem Prediction Markets Cannot Solve
The same information edge that makes skilled traders valuable to price discovery becomes a different problem when that information is not public. The paper points to a concrete case: the U.S. removal of Nicolás Maduro from power in Venezuela in January. In the days and hours before the operation, three newly created Polymarket accounts piled into a contract asking whether Maduro would be removed. The market priced the odds at roughly 10%. The new accounts placed orders of tens of thousands of shares before the price moved. When the raid happened, they collectively made more than $630,000. Two stopped trading entirely soon after; the third went mostly dormant.
There is no evidence of wrongdoing by those accounts. But the pattern mirrors the case of U.S. Army soldier Gannon Van Dyke, who allegedly used classified information about the Maduro operation to make $409,881 on Polymarket—and whose every transaction was recorded on the blockchain he assumed would hide him. Both Polymarket and Kalshi have stated that trading on non-public information violates their rules. The study finds, however, that insider trades move prices 7 to 12 times more per dollar than typical skilled trades—though they are rare and concentrated in a handful of events, not the day-to-day engine of price discovery.
The CFTC is currently suing New York over its attempt to ban prediction markets, while 38 state attorneys general have backed Massachusetts’ case against Kalshi. If regulators conclude that 97% of participants are effectively subsidizing a small informed minority—some of whom may be trading on information they should not have—the argument that prediction markets serve a broad public informational benefit weakens considerably.
The working paper is available from London Business School and Yale, and was published on April 26, 2026.
