[PAID] Polymarket 5m/15m Up-down Markets: Tick-level Quotes, Trades And L2 Depth (BTC/ETH/SOL/XRP), 58k Markets, ~5.7B Rows, June-August 2026

Disclosure: this is my own dataset and I sell access to it. Four days are free, linked below, so you can check the schema before deciding anything.

I run a collector against Polymarket’s public API for the 5-minute and 15-minute up/down markets on BTC, ETH, SOL and XRP. It has been running since 26 June 2026. What has accumulated so far:

• 58,600 markets (14,653 per asset — roughly 11k five-minute, 3.7k fifteen-minute)
• ~5.69 billion rows across six tables
• One Parquet file per asset, per table, per UTC day, each with a manifest carrying sha256, row count and interval bounds

Tables: book_snaps (best bid/ask), book_depth (ladder levels, ~88 per event), clob_trades and clob_trades_ws (trades with taker side, the second carries tx hash and fee for VPIN), spot_klines, book_meta.

Limitations, because they decide whether this is useful to you:

• book_snaps is top of book only. There are level and size columns but they are 0 in every row — they exist to keep the schema stable across tables.
• book_depth starts 20 July 2026. Fifteen days of the archive have it; everything before is quotes, trades and spot candles.
• The clock changes partway through. Up to 20 July the timestamp is receipt time rounded to the second, so intra-second ordering is lost and cannot be recovered. From 21 July it is exchange time in milliseconds.
• Per-day uptime runs 95.8% to 100%. Every day carries its own measured figure.

Free sample (original source): 26-29 June, four full days, no payment and no signup — https://huggingface.co/datasets/astrnvk/polymarket-orderbook-data-preview

Paid access covers 30 June onward and is priced per day ($5/day for short ranges down to $3/day past a month), with incomplete days billed proportionally lower — a day missing depth costs about half. Happy to answer schema questions here either way.