Headline: Study — It’s not the crowd but a tiny informed minority that makes prediction markets accurate A new working paper from researchers at London Business School and Yale upends a core selling point of prediction markets: that broad crowdsourcing of information produces accurate prices. Instead, the authors find that a very small subset of traders does nearly all the price-discovery work on Polymarket — and the rest of the crowd mostly provides liquidity and loses to them. What the researchers did - Dataset: every Polymarket trade from 2023–2025 — 1.72 million accounts and $13.76 billion in volume. - Method to separate skill from luck: for each trader the team re-ran their bets 10,000 times with everything identical except buy/sell direction chosen by coin flip. That produced a “no-edge” benchmark distribution of profits; traders who consistently outperformed that benchmark were deemed skilled rather than lucky. Key findings - Just 3% of traders drive the bulk of price discovery: they move prices toward the correct outcome and react first to new information (e.g., Fed announcements, earnings). - The remaining 97% mainly provide liquidity and, on average, lose money to the informed minority. - Of the largest raw-money winners, only about 12% beat the coin-flip benchmark — meaning most big winners look like luck once you control for chance. - Roughly 60% of “lucky winners” don’t hold up out-of-sample and become losers when tested on separate events. - Skilled traders’ activity measurably improves market accuracy, especially in the final stretch before resolution. - Insider-style trades, when they occur, move prices far more aggressively per dollar (roughly 7–12x the impact of a typical skilled trade), but these are rare and concentrated in a few events rather than a constant feature of markets. A concrete, worrying episode The paper connects these general patterns to a specific suspicious spike of activity: in January, ahead of a U.S. operation that removed Nicolás Maduro from power in Venezuela, three newly created Polymarket accounts placed large bets on Maduro’s removal when the market implied roughly 10% odds. Those accounts collectively pocketed more than $630,000 after the raid; two stopped trading and the third went largely dormant afterward. The authors note there’s no direct evidence of wrongdoing for those accounts, but the episode illustrates how nonpublic information could be exploited and disproportionately move markets. Policy and platform implications - Platforms like Polymarket and Kalshi ban trading on nonpublic information, but the study shows a tangible mechanism through which insider knowledge — or a highly informed few — can skew prices and profits. - For market designers and regulators this raises questions about surveillance, account onboarding, and whether current rules are sufficient to prevent rare but impactful insider-style trades. Bottom line Prediction markets aren’t failing — they still price events well — but their accuracy appears to come less from the wisdom of crowds and more from a small, repeatably skilled minority (plus occasional insider-style bets). The crowd supplies volume and liquidity, but in aggregate it’s often on the losing side of trades against that informed few. Read more AI-generated news on: undefined/news