Ligengxin96/polymarket-predict-fun-tick-data

Polymarket & Predict.fun Tick Data (Sample Day)

One complete UTC day (2026-09-08) of the archive behind outcometick: the order books, trade prints and Chainlink settlement feeds of the 5-minute BTC Up/Down markets on Polymarket, and the BTC order books, settlement price feed and market list of Predict.fun. The files are the archive's own — gzip-compressed, in the archive's own directory layout — and are the same files as the free sample archives published on GitHub.

Why this dataset

These markets are decided by a Chainlink price stream compared against a strike. Most prediction-market data stops at prices or order books; this sample ships the stream the markets settle on, and each market's strike and outcome, next to the books — so you can recompute who won instead of trusting a label.

Loading

Each file type is its own config, because the schemas differ:

from datasets import load_dataset

markets = load_dataset("Ligengxin96/polymarket-predict-fun-tick-data", "polymarket_markets", split="train")
twap60 = load_dataset("Ligengxin96/polymarket-predict-fun-tick-data", "polymarket_twap60s", split="train")

Or read the files directly with pandas: pd.read_csv(path) for .csv.gz, pd.read_json(path, lines=True) for .jsonl.gz.

Files

Polymarket — polymarket-data-samples/data/

config	path	rows	what it is
polymarket_twap60s	chainlink-twap-60s/daily/prices/BTCUSD/	81,969	60-second TWAP stream. 5- and 15-minute markets currently settle on it
polymarket_twap30s	chainlink-twap-30s/daily/prices/BTCUSD/	81,946	30-second TWAP stream, the earlier settlement line for 5-minute markets; still archived
polymarket_chainlink_instant	chainlink/daily/prices/BTCUSD/	81,970	Instantaneous Chainlink BTC/USD stream
polymarket_markets	polymarket/daily/markets/BTC-5m/	292	One row per market: window, token ids, strike, outcome, the venue's market object
polymarket_book	polymarket/daily/book/BTC-5m/	141,402	Full-depth order-book snapshots, per outcome token
polymarket_price_change	polymarket/daily/price_change/BTC-5m/	3,516,537	Order-book deltas, with best bid/ask
polymarket_best_bid_ask	polymarket/daily/best_bid_ask/BTC-5m/	1,314,078	Top of book, unthrottled. Prices only, no sizes
polymarket_trades	polymarket/daily/last_trade_price/BTC-5m/	480,491	Trade prints as the venue's WebSocket broadcast them

Predict.fun — predict-fun-data-samples/data/predict-fun/

config	path	rows	what it is
predict_prices	prices/BTCUSDT/	82,342	Settlement price feed (float only; the venue publishes no fixed-point value)
predict_klines_1m	klines/BTCUSDT/1m/	1,440	1-minute OHLC derived from that feed, with a tick count. No trade volume
predict_markets	markets/	1,245	Markets for BTC, ETH and BNB at every interval the venue runs, with start and end price
predict_orderbook	orderbook/BTC-5M/	478,603	Order-book snapshots. The venue publishes no deltas and no trades

Columns

Polymarket settlement CSVs — feed_ts_ms upstream event time (UTC ms) · value price as a float · full_accuracy_value the relay's fixed-point integer string (1e18-scaled), whenever the relay publishes it · server_ts_ms upstream server time · recv_ms our receive time.

Polymarket JSONL (book, price_change, best_bid_ask, last_trade_price) — slug market · asset_id outcome token (null on price_change, which is market-level) · event_type · event_ts_ms upstream event time · recv_ms our receive time · payload the venue's message as received.

Polymarket markets — slug, asset, interval_sec, condition_id, token_ids (token_ids[0] is Up), start_sec, end_sec, resolved, outcome_prices (["1","0"] Up won, ["0","1"] Down won), strike_value (1e18-scaled), raw the venue's market object.

Predict.fun orderbook — market_id, category_slug, update_ts_ms, recv_ms, payload with asks / bids as [price, size].

Predict.fun prices — price_feed_id, symbol, provider, publish_time and server_ts (whole seconds), price, recv_ms (ms).

Predict.fun markets — category_slug, asset, interval_label, market_id, price_feed_id, price_feed_symbol, price_feed_provider, condition_id, start_sec, end_sec, start_price, end_price, status. Treat a market as settled when end_price is present; status can lag behind it.

Things worth knowing before you model anything

• Which stream settles a Polymarket market is a property of the market. Read raw.cryptoMarketConfig.twapLookbackSeconds from its markets row and pick the matching stream; do not infer it from the date — the rule has changed before.
• Timestamps are kept apart, never merged. Settlement rows carry event, upstream-server and receive time; order-book and trade rows carry event and receive time. The difference is the capture latency, measurable row by row.
• Order books are captured at a cadence. Snapshots and deltas are thinned; on this date the BTC deltas are kept at up to one per market per 20 ms. best_bid_ask is the exception — unthrottled, one row per top-of-book update, prices only.
• Trade prints are the venue's broadcast, stored unthrottled and not reconciled against on-chain fills.
• Rows are stored in arrival order, out-of-order ones included. Sort by event time if you need to.
• Gaps stay gaps. Upstream relays stall occasionally; nothing is interpolated to make a day look whole.

The full archive

This is one day. The archive runs continuously from 2026-06-08 on Polymarket (BNB, BTC, DOGE, ETH, HYPE, SOL, XRP; 5- and 15-minute markets) and from 2026-06-13 on Predict.fun (BTC, ETH, BNB), and is sold at outcometick.com by subscription or by date range. The same archive can be replayed against a strategy in a hosted sandbox; the strategy SDK is outcometick on npm and PyPI.

• Site and API docs: https://outcometick.com · https://outcometick.com/docs
• The same sample on GitHub: polymarket-data-samples · predict.fun-data-samples
• Also on Kaggle, decompressed: ligengxin96/polymarket-predict-fun-tick-data

License

CC BY 4.0. Please credit outcometick (outcometick.com).