About the data
Polydata is a historical archive of Polymarket prediction markets, built for research and for public transparency of prediction data and pricing. Every trade in it is read directly from public Polygon blockchain records. Each trade is linked to the market and event it belongs to, using the market description. Political and economic events are further enriched with news metadata from the GDELT Project (https://www.gdeltproject.org/): how widely an event was covered, when coverage peaked, the overall tone, and links to the leading headlines. Polydata is independent and not affiliated with Polymarket.
The archive starts on November 21, 2022 and is updated once a month, on the 10th. Each monthly release adds events that closed at least 30 days before the release date. Wallet histories and short-term BTC markets follow the same cycle. Events that are still open are excluded from the cycle. The 30-day cool-down is deliberate: outcomes settle, disputes are resolved, and missing records are recovered before the data reaches you.
Polydata is historical data for research and backtesting, not a tool for real-time trading. It is suited to research papers, essays and dissertations in econometrics, behavioral economics, sociology, social psychology and political science. Possible topics include how collective beliefs form and shift under uncertainty, how markets absorb news and rumor, and how political loyalty shows up in betting behavior. The data can also capture moral panics and herd behavior, and the boundary between informed and speculative trading. For quantitative work, the archive provides a clean test bed: you can check how well a forecasting or pricing model would have performed against real market outcomes. It does not provide live prices, order books or streaming feeds, and it never will.
We describe the data as it is. Every file comes with a README that documents each column, explains how to calculate volume and profit correctly, and notes where sources or formats change over time. Each file is licensed to a single buyer and carries a unique identifier. Full terms are set out in the Academic and Commercial licenses.