Polymarket data providers, compared

Which provider covers the most Polymarket market types?

PMData, Telonex, and PolymarketData.co all cover every Polymarket market. PMData's all-market data starts on 2026-05-01; before that, it covers crypto Up/Down markets from 2026-02-01. Tardis.dev collects only markets with trading activity, not the full catalog. Beyond Polymarket, Telonex also covers Predict.fun and Binance, and Tardis.dev covers 50+ crypto exchanges.

Which provider has the most granular Polymarket order book data?

The most granular form is Polymarket's WebSocket messages stored exactly as sent: every book snapshot and price_change, unprocessed. PMData and Tardis.dev store data this way, with a microsecond receive timestamp on each message. Telonex delivers order book snapshots and quotes generated from its own local order book rather than the original messages, so the book cannot be rebuilt event by event. PolymarketData.co keeps one snapshot per minute, so order book changes between snapshots, including queue position, short-lived liquidity, and slippage, are not in the data.

Which provider has the lowest-latency Polymarket data?

Measured as receive time minus Polymarket's event timestamp, median and p99 latency are 4.6 ms and 40.1 ms for PMData, 5.7 ms and 56.1 ms for Tardis.dev (measured on its 2026-08-01 data), and 8.3 ms and 73.1 ms for Telonex. PMData collects in AWS London, colocated with Polymarket's exchange. PolymarketData.co has no per-event receive timestamp, so its latency cannot be compared.

Which provider has the most complete Polymarket data?

PMData records each market over two WebSocket connections on separate networks, so if one connection drops, the other keeps recording. Telonex also uses two connections per market. Tardis.dev uses one connection per market. PolymarketData.co does not publish how it records.

What data do I need to backtest Polymarket crypto Up/Down markets?

The order book of each market and the price it resolves on. Since 2026-08-07, Up/Down markets resolve on Chainlink TWAP. PMData collects Chainlink 30s and 60s TWAP and Data Streams prices directly from Chainlink, in the same region as its Polymarket order books, and delivers each Up/Down series as one ZIP per day. Telonex's Chainlink prices come through the Polymarket proxy. PolymarketData.co and Tardis.dev do not provide Chainlink data.

How accurate are the timestamps, and why does it matter?

Your backtest's latency, and your alignment with Binance prices or your own trading logs, are only as good as the timestamps. PMData syncs its clock to AWS PHC, which AWS documents with a clock error bound typically under 40 µs. The bound is the most the clock can be off, not the usual offset, which is far smaller. Tardis.dev writes timestamps to 100 ns but syncs its clock over Google NTP. Google publishes no error bound for standard NTP, and its own chrony examples show ±257 µs to ±396 µs. Telonex does not publish how its clock is synchronized.

Which providers keep full history?

PMData and Telonex let paid plans download their full archive: PMData from 2026-02, Telonex from 2025-10. PolymarketData.co limits history by plan: 1 month at $60/mo, 3 months at $120/mo, and full history from August 2025 at $360/mo. Tardis.dev's $1,200/mo plan gives access to the last 4 months.

Which Polymarket backtesting tool is most realistic?

Backtest realism is limited by the data: simulating fills, queue position, and slippage needs tick-level order books with receive timestamps. PMData provides pm-hftbacktest, a Polymarket fork of hftbacktest that models queue position and latency. Across 547 BTC 5-minute markets, its outcomes matched live trading in 95.32% of markets, while overstating live profit by 10.45%. Tardis.dev data works with hftbacktest and most open-source backtesters. Telonex data works with prediction-market-backtesting, built on NautilusTrader. PolymarketData.co's 1-minute snapshots cannot support fill-level backtests.

More comparisons