How to Get Polymarket Historical Data: Order Books, Trades, and Resolved Markets via API

Quick answer: polyReplay is a historical data API for Polymarket's crypto prediction markets. It exposes L25 order book snapshots, trades, and resolved Up-or-Down outcomes for BTC, ETH, SOL, XRP, DOGE, BNB and HYPE, joined with Binance spot prices, via a REST API and bulk Parquet downloads.

What is polyReplay?

polyReplay is the data layer for Polymarket crypto prediction markets. We ingest every order book update and every executed trade on Polymarket's Up-or-Down crypto markets, archive them at microsecond resolution, join each market to the second-resolution Binance spot tape for the underlying asset, and serve everything back through a REST API and as bulk Parquet files downloadable from Cloudflare R2 with zero egress cost.

If you have ever tried to backtest a Polymarket strategy or run a quant study on prediction-market microstructure, you know the data is not easy to come by. The Polymarket CLOB exposes live books, but not history. polyReplay closes that gap.

What's in the dataset

| Dataset | Resolution | Coverage | | --- | --- | --- | | L25 order book snapshots | Microsecond | BTC, ETH, SOL, XRP, DOGE, BNB, HYPE | | Top-of-book quotes | Per update | All Up-or-Down crypto markets | | Executed trades | Per fill | All Up-or-Down crypto markets | | Resolved markets | One row per market | Start price, end price, winner | | Binance spot overlay | 1-second trades | BTC, ETH, SOL, XRP, DOGE, BNB, HYPE |

As of writing, we have 1.37 million markets indexed, 32,049 Parquet files archived, and 13 GB of historical data available.

Three ways to consume the data

There is no one-size-fits-all shape for market data, so polyReplay ships three access patterns:

1. REST API — single-market reads, ideal for live dashboards, alerts, and ad-hoc analysis. Start with the quickstart.
2. Bulk Parquet downloads — full historical pulls served from Cloudflare R2 with zero egress cost. Ideal for backtests, training data, and long-running research. See datasets.
3. Python SDK — a strategy decorator, a tick-by-tick backtest runner, and a paper-trading mode against live books. One strategy, three modes (backtest, paper, live). See the SDK preview.

Getting started in under five minutes

# 1. Sign up and grab an API key from polyreplay.dev/api-keys
export POLYREPLAY_API_KEY="..."

# 2. Hit a market endpoint
curl -H "Authorization: Bearer $POLYREPLAY_API_KEY" \
  https://polyreplay.dev/api/v1/markets/<market-id>

The same data is available in Parquet form for bulk downloads — see the docs for the full catalog and column reference.

Who is this for?

polyReplay is built for:

• Quant traders backtesting strategies on Polymarket Up-or-Down markets.
• Researchers studying prediction-market microstructure, price discovery, and the relationship between prediction markets and underlying spot prices.
• Data engineers who want a clean Parquet feed instead of writing their own scraper against the CLOB.
• AI agents and LLM applications that need structured, queryable prediction-market history.

"I spent two weeks writing an ingestion pipeline against the Polymarket CLOB before realising someone had already done it." — every quant who hasn't found polyReplay yet.

Why we built it

Prediction markets are one of the most interesting microstructure environments in crypto right now: binary payoffs, hard expiries, and a tight coupling to the spot tape. But the data is locked up. We built polyReplay because we wanted to trade these markets ourselves, couldn't find a clean historical archive, and figured we couldn't be the only ones.

The platform charges builder rewards on every order it routes, so the data business and the trading business align: more usage means better data means better trading.

What's next

• Sign up for an API key — free tier available.
• Read the docs — quickstart, authentication, full endpoint reference.
• Browse pricing — see plan limits and Parquet bulk-download tiers.

If you have a specific dataset or feature request, let us know.