The Chainlink settlement feeds and order books of Polymarket and Predict.fun crypto Up/Down markets, plus Polymarket’s trade prints — collected continuously, checksummed, and downloadable by API.
Not a reconstruction — the actual Chainlink Data Streams reports these markets settle on, tick by tick, with the relay’s full-precision fixed-point integer beside each value whenever the relay publishes one. Including the TWAP streams that became the settlement line on 2026-08-07 — 5- and 15-minute markets both settle on the 60s stream today. The 30s stream is archived through 2026-10-19 (UTC). You can recompute any outcome yourself and check it against the official one.
Who won, and the price to beat. Strike is recorded from the boundary report with row-level provenance; when we cannot establish it honestly, the field is null rather than guessed from a nearby tick. The settlement audits recompute each archived market and publish how many agree with the official outcome, how many do not, and which cannot be settled from the archive.
Polymarket settlement audit → · Predict.fun settlement audit →
Event time and our receive time on every tick — kept separate, never merged — plus the upstream server time through 2026-10-19 (UTC). You can measure the capture path instead of trusting it. Collection runs next to the venues' own infrastructure.
Frames are archived as received, out-of-order included. Every file ships its byte length and a SHA-256, so you can prove you got the bytes we published. Gaps are left as gaps: nothing is interpolated to make a day look whole.
Polymarket historical data → · Polymarket order book data →
Predict.fun historical data → · Predict.fun order book data →
Real opening rows from real files, not a mock-up. Switch tabs to see each dataset’s schema.
| feed_ts_ms | value | full_accuracy_value | server_ts_ms | recv_ms |
|---|---|---|---|---|
| 1792281600000 | 81382.94137341123 | 81382941373411230000000 | 1792281600878 | 1792281601617 |
| 1792281601000 | 81382.89254158986 | 81382892541589860000000 | 1792281601843 | 1792281602372 |
| 1792281602000 | 81383.19924613074 | 81383199246130740000000 | 1792281603126 | 1792281603799 |
| 1792281603000 | 81384.53505361412 | 81384535053614125000000 | 1792281603772 | 1792281604197 |
| 1792281604000 | 81384.983544305 | 81384983544305000000000 | 1792281605056 | 1792281605460 |
Opening rows from the real 2026-10-18 (UTC) files. Values are copied verbatim. The complete files: Polymarket sample data download — check the checksums yourself. The same sample is also on Kaggle (decompressed) and Hugging Face.
Monthly, cancel anytime. Every key reaches the last 30 archived days.
Renews monthly, cancel anytime. Card, Apple Pay.
Need a one-off historical bundle, a single asset, or an invoice? Message us— we also settle in USDT.
One API key. Files come straight from object storage, so a large pull is limited by your own bandwidth, not by us.
# what dimensions your subscription covers curl -H "Authorization: Bearer $OT_KEY" \ https://outcometick.com/v1/meta # search by asset / interval / dataset / date range curl -H "Authorization: Bearer $OT_KEY" \ https://outcometick.com/v1/files?asset=btc&interval=5m&dataset=book&from=2026-08-01&to=2026-08-12 # one-step download (302 to object storage; pandas can read it directly) curl -H "Authorization: Bearer $OT_KEY" \ https://outcometick.com/v1/dl/2026-08-12/BTCUSD-prices-2026-08-12.csv.gz
Every response carries the file’s SHA-256, so you can verify what you downloaded is exactly what we published.
full api reference →strategy SDK reference →Tick-level history for Polymarket and Predict.fun crypto Up/Down markets: the Chainlink settlement feeds the markets resolve against (a decimal value, with the relay’s full-precision fixed-point string beside it whenever the relay publishes one), each market’s strike and settled outcome, order-book snapshots and incremental updates, Polymarket trade prints, and OHLC candles from 1 second to 1 month (a tick count per candle, not trade volume).
Polymarket (BNB, BTC, DOGE, ETH, HYPE, SOL, XRP) for the 5-minute and 15-minute Up/Down markets, and Predict.fun (BTC, ETH, BNB) for the 5-minute, 15-minute, hourly and daily ones.
Continuously since 2026-06-08, growing by one archived day at a time. Each day is checked before release and normally published by 02:00 UTC the next day; a day that fails the checks is held for repair and ships by 13:00 UTC the next day.
Through a self-serve HTTP API. You buy a key, list the files for a date or a range, and download each one; every file carries its sha256 so you can verify it. No sales call and no bespoke export step.
$10 per month for one venue (more for both), each a rolling 30-day window that moves with the archive. Deeper history is a one-time purchase of any date range you pick, priced per archived day. Card works for every purchase; WeChat Pay (in CNY) and stablecoins (in USD) for one-time purchases. Prices are in USD on the English site and CNY on the Chinese one.
Yes, and that is the point of shipping the settlement feed rather than just a label. Each market carries its strike, and each settlement tick carries the relay’s fixed-point figure whenever the relay publishes one, so you can recompute every win/loss from the raw prices instead of trusting our label.
Yes, without downloading anything first. You paste a strategy in Python or Node.js and it runs in a sealed sandbox against the same tick archive we sell — settlement feed, full-depth order books and trade prints. You get back an equity curve, every fill priced against the depth that was really there, and a settlement cross-check. There is also an SDK on npm and PyPI to run it locally. Billed per market-day scanned, from $2.
Structurally, not by filtering. The replay loop pulls one row at a time from the process input, so future rows are not in the sandbox yet when your handler is called — there is nothing to peek at rather than something being hidden from you. A market’s settled outcome is stripped from the market view until the settlement hook fires, and fills are matched against the real book depth at that instant rather than assumed at the midpoint.