A new study challenges the foundational claim of prediction markets—that broad participation harnesses collective wisdom to forecast outcomes accurately. Researchers found that just 3% of traders account for most price discovery on Polymarket, meaning a tiny group of informed participants moves prices toward correct outcomes while the vast majority loses money.

Market Context

Prediction markets like Polymarket and Kalshi have grown into multi-billion dollar platforms where traders wager on everything from election results to corporate earnings to geopolitical events. The industry has long argued that these markets work because aggregating many participants' beliefs produces accurate forecasts—a concept rooted in the 'wisdom of crowds' theory. But this new research suggests accuracy depends not on crowd participation but on a small, skilled minority.

Analysis

The study, a working paper from Roberto Gómez-Cram and Yun Guo of London Business School along with Theis Ingerslev Jensen and Howard Kung of Yale, analyzed every Polymarket trade from 2023 to 2025—spanning 1.72 million accounts and $13.76 billion in trading volume. The researchers directly tested whether broad participation drives market accuracy or whether a concentrated group of informed traders does the heavy lifting.

The findings support the latter view. These 3% of traders consistently predict outcomes correctly and shift prices in the right direction. The remaining 97% provide liquidity and generate volume but, in aggregate, sit on the losing side of trades against this informed minority. Their losses directly fund the skilled traders' profits.

Distinguishing skill from luck presented a methodological challenge. With over a million traders, many apparent winners could simply be statistical outliers. To filter noise from genuine edge, the authors reran each trader's bets 10,000 times, swapping every position's direction via coin flip while keeping markets, timing, and dollar amounts constant. This created a benchmark for random performance. Traders whose actual results consistently beat the coin flip demonstrated skill; those who did not were classified as lucky rather than skilled.

The numbers are stark: among the biggest winners by raw profit, only 12% beat the benchmark. Many apparent winners didn't maintain their status—roughly 60% of "lucky winners" turned into losers when checked against a separate sample of events.

Skilled traders prove most valuable in moments that matter. Their activity improves market accuracy most significantly in the final stretch before resolution, and they react first when new information hits—whether Federal Reserve announcements or corporate earnings releases—while other traders show little consistent reaction.

The same edge making skilled traders valuable to price discovery raises uncomfortable questions about information integrity. Both Polymarket and Kalshi prohibit trading on non-public information, but the researchers documented a concrete case: the U.S. removal of Nicolás Maduro from power in Venezuela in January 2026. In the days before the operation, three newly created Polymarket accounts placed unusually large bets—including orders of tens of thousands of shares—on a contract asking whether Maduro would be removed when the market priced odds at roughly 10%. After the raid occurred, these accounts collectively made more than $630,000. Two stopped trading entirely shortly after, and the third went mostly dormant. The researchers noted no evidence of wrongdoing on these accounts.

Insider trades, when they occur, move prices seven to 12 times more aggressively per dollar than typical skilled trades—but remain rare and concentrated in specific events rather than representing day-to-day market mechanics.

Key Numbers

- 1.72 million: Polymarket accounts analyzed in the study

- $13.76 billion: total trading volume examined from 2023-2025

- 3%: proportion of traders driving price discovery

- 12%: biggest winners by raw profit who beat the coin-flip benchmark

- 60%: share of 'lucky winners' who became losers when checked against separate event samples

- $630,000+: collective profits made by three accounts before Maduro's removal on Polymarket contracts priced at ~10% odds

What to Watch

Regulatory scrutiny of prediction markets may intensify following these findings. Polymarket and Kalshi have faced questions about their ability to detect non-public information trading despite explicit prohibitions. Traders should monitor whether platforms implement new surveillance mechanisms or disclosure requirements. For investors using prediction market prices as sentiment indicators, the study suggests they are largely tracking a skilled minority rather than true crowd wisdom—especially valuable context when assessing political betting markets ahead of elections or geopolitical contracts tied to sensitive events.

The researchers' methodology—using resampling techniques to separate skill from luck—may become a standard framework for evaluating trader edge across prediction and sports markets. Market operators could face pressure to disclose similar analyses publicly.