Backtesting data
Polymarket backtesting data for 5-minute crypto markets
To backtest a Polymarket 5-minute Up or Down strategy honestly you need three things: the order book at the moment you would have traded, the trades that printed, and the official settlement of every window. Our files give you all three, aligned to the window clock.
What a realistic backtest needs
- The executable price, not the last trade. You buy at the best ask with the size that was actually there. Our book files carry best and second-best bid and ask with sizes, every change.
- Exact timing. Every row has
elapsed_msfrom the window start, so “what was the ask 150 seconds in” is one filter. - Fees. Polymarket charges takers
0.07 × p × (1 − p)per share. See fees explained. - The official result. Windows settle on 60-second Chainlink TWAPs, not on a spot price. Our window files carry Polymarket's own
open_twap60,close_twap60and result, so you grade on the real outcome. See how settlement works. - Latency. Exchange and receive timestamps let you add a realistic delay between signal and order.
Why candles and snapshots mislead
A 5-minute market can open, swing and settle between two one-minute snapshots, and the losing side's book often empties in the final minute. A backtest on sampled prices fills at prices that were never available. Tick data removes that error.
Get started
Follow the step-by-step backtesting guide (Python and DuckDB), try it on the free sample hour, then pick a market: Bitcoin, Ethereum, Solana, XRP, Dogecoin, BNB, Hyperliquid. Every file is checked daily: data quality report.