Jerusalem, 21 October, 2026 (TPS-IL) — Researchers from Tel Aviv University analyzed 1,014,122 forecasts across four major prediction platforms: Polymarket, Kalshi, Manifold, and Metaculus. The study, published in the Proceedings of the National Academy of Sciences (PNAS), examined predictions across 12 domains over a decade, including elections, geopolitics, security, and technology.
Overall, forecast probabilities closely matched actual outcomes. Events assigned a 70% probability generally occurred at roughly that rate. However, accuracy varied by subject. Markets systematically overestimated the likelihood of events involving military conflict, geopolitics and technology, while underestimating election-related outcomes across portions of the probability range.
Led by Professor Michael Gilead of Tel Aviv University’s School of Psychological Sciences, the research also found that forecasts became more informative as deadlines approached. Strong calibration was already evident after just 25% of a forecast’s allotted period had elapsed.
