Historical prediction market data,
row by row.
Polybook is a historical dataset of prediction-market order books: one-second best bid and ask, resting size and settlement for BTC, ETH, SOL and XRP 5-minute and 15-minute up/down markets. Markets disappear from the venue the moment they settle — this is the record that survives.
Coverage at a glance
| History starts | March 2026 — continuous since |
|---|---|
| Markets | BTC · ETH · SOL · XRP up/down |
| Timeframes | 5-minute · 15-minute (8 streams) |
| Resolution | 1 second per order-book snapshot |
| Rows | ≈158 million across all streams |
| Formats | REST API · CSV (gzip) · Parquet (month-partitioned) |
| Bulk delivery | Full snapshot download, 7-day window |
What each row contains
One consistent schema across every stream, so a model written for one market works for all eight.
| Timestamp | ISO-8601 and Unix, UTC |
|---|---|
| Best bid / ask | Both outcomes, as probabilities in [0,1] |
| Resting size | Shares available at the best bid and ask |
| Freshness | Explicit staleness flag — no silent gaps |
| Market phase | Pre-open, trading, closing, settled |
| Identity | Market id, slug, symbol, timeframe |
| Settlement | Outcome recorded for settled markets |
What people build with it
Backtesting
Replay order-book states second by second to test execution logic, fill assumptions and short-horizon signals on markets that no longer exist on the venue.
Market microstructure research
Study how probability quotes form, how depth thins into expiry, and how settlement resolves — with the metadata needed to make results reproducible.
Model features
Derive features such as spread, imbalance and depth decay from historical book states instead of guessing them from settlement prices alone.
Ways to get the data
REST API
Query windows and rows over HTTP with an API key. Starter covers recent history on one stream; Pro covers everything across all eight.
CSV & Parquet
Per-window gzipped CSV for quick inspection, month-partitioned Parquet for heavy analysis — both the same schema as the API.
Full archive
One-off purchase of the complete historical snapshot as a bulk download, with a seven-day download window and no API limits.
Questions
How far back does the historical data go?
Recording began in March 2026 and has run continuously since. The archive holds every market recorded to date — roughly 158 million order-book rows across eight streams.
Which markets and timeframes are covered?
BTC, ETH, SOL and XRP up/down markets on 5-minute and 15-minute horizons — eight streams. Each market is captured from before it opens until it settles.
What exactly is recorded per row?
Best bid and ask for both outcomes, resting size at each side, a freshness flag, market phase, market id and slug, and the settlement outcome once resolved.
Can I get the full history as a bulk download?
Yes — the Archive product is a one-off purchase of the entire snapshot as month-partitioned Parquet plus raw per-window CSV and market metadata, downloadable for seven days after purchase.
Do I need an API key?
For API access, yes: Starter and Pro include a REST API key with per-minute rate limits. Bulk Parquet downloads do not use the API.
Is there a free sample?
Yes — the most recent settled market window as CSV plus market metadata, no account required.
Start with a sample, then decide
The free sample is the real thing — same schema, same delivery, one settled window.
Polybook is an independent data provider and is not affiliated with Polymarket. Market names are referenced only to describe the data recorded.