I’ve been building a Polymarket dataset for the cases where a candle or a periodic snapshot isn’t enough. If you’re testing short-window signals or execution, the useful question is often what was actually resting on the bid and ask when the decision was made.<\/p>\n
The paid catalog has event-level order-book updates, top-of-book quotes, up to 25 levels per side, and a trade tape where available. The rows carry Polymarket’s source timestamp when present, our receive timestamp, market\/token join keys and gap-audit information. Coverage is per market and day, so check the catalog for the particular series you need. The earliest buyable order-book day is 2026-02-21; feed trade tape starts 2026-04-13. Raw inbound websocket messages are available only from our own capture beginning 2026-05-11.<\/p>\n
There is a no-sign-up sample from the BTC Up or Down 4h series on 2026-08-19: a full UTC day across 13 markets\/26 tokens, with L1, L2 and trades in Parquet or CSV, plus one hour of raw websocket messages. The page shows the schema and download links: https:\/\/tickfoundry.com\/samples<\/a><\/p>\n If you’re working on a backtest, I’d be interested in what matters most to you: quote lifetime, depth at a specific timestamp, or checking whether a data gap overlaps your signal window?<\/p>\n Disclosure: I run TickFoundry, which sells the historical dataset. The linked sample is free to download.<\/p>\n<\/div>\n submitted by \/u\/madrading <\/a>
[link]<\/a><\/span> [comments]<\/a><\/span><\/p>
