Prediction markets like Polymarket may not be living up to their “wisdom of crowds” billing — instead, a tiny minority of skilled traders appears to be pulling the levers of price discovery, a new working paper finds.
Researchers Roberto Gómez‑Cram, Yunhan Guo, Theis Ingerslev Jensen and Howard Kung (London Business School and Yale) analyzed every Polymarket trade from 2023–2025 — 1.72 million accounts and $13.76 billion in volume — and reached a clear conclusion: roughly 3% of traders are responsible for most of the markets’ movement toward the correct outcomes. The other 97% mainly supply liquidity and volume, but on aggregate they lose money to that informed minority.
Distinguishing skill from luck was central to the study. With millions of participants, some users will show big profits purely by chance. To separate genuine edge from randomness, the authors simulated each trader’s activity 10,000 times, keeping the timing and sizes of bets identical but randomizing buy/sell direction (a coin‑flip benchmark). Traders whose real returns consistently beat that randomized baseline were judged skilled.
The results were striking. Among the raw biggest winners, only 12% consistently outperformed the coin‑flip benchmark; many apparent top performers turned out to be “lucky winners.” When tested on a separate set of events, about 60% of those lucky winners became losers. In short: most headline profits are explained by chance, but a small repeat group does have a measurable forecasting edge.
Those skilled traders matter for market quality. When they make up a larger share of activity, contract prices move closer to the true outcome, particularly in the final hours before resolution. They are also the first to react to new public information — shifting positions around Fed announcements or earnings releases — while most other traders show little consistent, timely response.
The study also underscores a tougher problem: what happens when the edge comes from nonpublic information. Exchanges such as Polymarket and Kalshi prohibit trading on insider knowledge, but the paper documents suspicious-looking episodes consistent with that risk. As one example, in the days and hours before a U.S. operation to remove Nicolás Maduro was priced into a Polymarket contract, three newly created accounts placed unusually large bets on a market that initially priced Maduro’s removal at roughly 10%. Those accounts collectively pocketed more than $630,000 when the event resolved; two went dormant shortly after and the third largely stopped trading. The paper notes there’s no evidence of wrongdoing on those accounts, but the pattern illustrates how private information could move prices dramatically. When insider‑style trades do occur, they shift prices far more aggressively per dollar — roughly seven to 12 times more than typical skilled trades — though such episodes are rare and concentrated in a few events.
Policy and product implications are significant. The findings challenge the narrative that prediction markets succeed because a dispersed crowd aggregates knowledge. Instead, they appear to work largely due to a small, repeat set of informed traders. That raises questions for market designers, exchanges and regulators: how to detect trading on nonpublic information, how to protect less sophisticated users who fund informed profits, and whether markets need structural changes to preserve credibility.
Bottom line: prediction markets aren’t failing — they can be accurate — but that accuracy seems to rest on a narrow base of consistent, well-informed players rather than a broad, omniscient crowd.
Read more AI-generated news on: undefined/news
