Polymarket historical data
Polymarket historical data for 5-minute crypto markets
Looking for Polymarket historical data to backtest the 5-minute Up or Down markets? General datasets cover thousands of markets with snapshots minutes apart. A 5-minute market can open, swing and settle between two of those snapshots. We cover just these markets, in full, tick by tick.
What we record
- Every order book change to two levels deep on both outcomes, timestamped to the millisecond by the exchange and by us.
- Every trade: price, size and aggressor side.
- Every window's result: price to beat, final Chainlink price, Up or Down, plus the market and token IDs to verify it on-chain.
Coverage
| Coin | Name | Windows per day | Order book from |
|---|---|---|---|
| BTC | Bitcoin | 288 | 16 October 2026 |
| ETH | Ethereum | 288 | 16 October 2026 |
| SOL | Solana | 288 | 16 October 2026 |
| XRP | XRP | 288 | 16 October 2026 |
| DOGE | Dogecoin | 288 | 16 October 2026 |
| BNB | BNB | 288 | 16 October 2026 |
| HYPE | Hyperliquid (HYPE) | 288 | 16 October 2026 |
Built for 5-minute windows
Every row carries window_start and elapsed_ms. Asking what the book looked like 90 seconds into every BTC window last week, and what happened next, is one query, not a day of timestamp alignment. See the backtesting guide.
Formats
Apache Parquet (zstd) and CSV, one file per coin per type per UTC day. Works with pandas, Polars, DuckDB, R and Spark.
Try it first
The free sample is a real hour of BTC data in exactly the subscriber format. No sign-up.