Predict.fun historical data,
behind the same API.
Predict.fun sits behind the same API and the same key as Polymarket. Its raw files keep this venue’s own field names; the backtest runner is what maps them into the shared event model.
What is in the archive
The two venues cover different assets, so nothing here carries over from the Polymarket page: call /v1/meta?venue=predict with your key for the assets, intervals and date range you actually have.- orderbook
- Order-book snapshots (Predict.fun)
- markets
- Per-market metadata, strike and settlement outcome
- prices
- Settlement feed, tick by tick (instantaneous Chainlink stream)
- klines
- OHLC candles derived from the settlement feed, with a tick count — no trade volume; the settlement feed is a price feed
A real Predict.fun day, opening rows
The first rows of real files from the Predict.fun sample day — this venue’s own field names, in the same files a key downloads.| price_feed_id | symbol | provider | publish_time | server_ts | price | recv_ms |
|---|---|---|---|---|---|---|
| 1 | BTCUSDT | CHAINLINK | 1792281600 | 1792281601 | 81491.035 | 1792281601151 |
| 1 | BTCUSDT | CHAINLINK | 1792281601 | 1792281601 | 81491.035 | 1792281602045 |
| 1 | BTCUSDT | CHAINLINK | 1792281602 | 1792281603 | 81491.035 | 1792281603368 |
| 1 | BTCUSDT | CHAINLINK | 1792281603 | 1792281603 | 81491.035 | 1792281603915 |
| 1 | BTCUSDT | CHAINLINK | 1792281604 | 1792281605 | 81491.035 | 1792281605261 |
Opening rows from the real 2026-10-18 (UTC) files. Values are copied verbatim. The complete files: Predict.fun sample data download. The same sample is also on Kaggle (decompressed) and Hugging Face.
What is different about this venue
One book, and its own vocabulary
Predict.fun quotes a single normalised book; the other side is its mirror at 1−px. We archive it as it comes rather than fabricating a second ladder that never existed upstream — so the files you download carry this venue’s field names (market_id, category_slug, its own timestamp column), not Polymarket’s.
The backtest runner is where the two venues become one event model, so a strategy does not care which it is replaying. If you are parsing the raw files yourself, read each venue’s schema — they are documented separately because they are different.
Cross-venue on one clock
Both venues are captured with an upstream timestamp and our receive time kept apart, and archived on the same UTC day boundary, so the two line up without a clock-alignment step first. Collection runs next to each venue’s own infrastructure.
What does not line up is where they start: each venue has its own first complete archive day, and they are not the same date. Coverage is reported per venue for that reason, never merged.
How to pull it
Add venue=predict to the same endpoints. Nothing else about the call changes.# list this venue’s files, newest day your key can reach curl -H "Authorization: Bearer $OT_KEY" \ "https://outcometick.com/v1/files?venue=predict&asset=btc" # per-market metadata, strike and outcome curl -H "Authorization: Bearer $OT_KEY" \ "https://outcometick.com/v1/files?venue=predict&dataset=markets" # download one — filter to the dataset you are naming the file after URL=$(curl -s -H "Authorization: Bearer $OT_KEY" \ "https://outcometick.com/v1/files?venue=predict&dataset=orderbook&asset=btc" \ | jq -er '.files[0].url') curl -L -H "Authorization: Bearer $OT_KEY" "$URL" -o orderbook.jsonl.gz
This venue has its own free sample. The Predict.fun sample archive is one real day of the same files, so you can read the schema and write your parser before deciding whether to buy anything.
Questions about Predict.fun data
What Predict.fun historical data is there?
Order-book snapshots, per-market metadata with strike and settled outcome, the settlement price feed, and OHLC candles derived from it — for Predict.fun’s crypto Up/Down markets.
Are the files in the same format as Polymarket’s?
Same API, same key, same filters and download route — but not the same files. Predict.fun’s raw files keep its own field names and its book levels are [price, size] pairs; in its price feed, publish_time and server_ts are in seconds while recv_ms is in milliseconds. The schemas page documents the two venues separately.
Can I buy Predict.fun data on its own?
Yes. A date-range purchase can cover Predict.fun alone or both venues, priced per archived day; current prices are on the pricing table.
Is there a free sample?
Yes, its own: one real Predict.fun day of the same files, linked above, so you can write your parser before buying anything.
Backtest against it too
The runner takes venue=predict and replays it in the same sealed sandbox, through the same event model — settling on this venue’s own price feed rather than Polymarket’s.