Data
What we record, and how we prove it.
Since October 26, 2026, adding sources as we go (each start date is in the table below), our recorders have captured, update by update: Polymarket US 5-minute Bitcoin books (top five levels), every trade print, official start prices and outcomes; 15-minute and 1-hour books; SOL and XRP books; spot trades from Coinbase, Kraken and Bitstamp; OKX and Hyperliquid perpetual flow; and the CF Benchmarks BRTI once a second. Kalshi recording is paused pending the venue’s consent.
Recorded
Sources, cadence and yesterday’s numbers
Rows and coverage for Oct 29 (UTC day), from that day’s export and gap register. Coverage is the share of the day not inside a silence longer than that stream’s threshold; sparse streams have long thresholds. When a feed drops, it shows here.
| Source | What | Cadence | In archive since | Rows, Oct 29 | Coverage, Oct 29 |
|---|---|---|---|---|---|
| Polymarket US | BTC 5-minute book, top five levels | every update | Oct 26 | 666,830 | 92.1% |
| Polymarket US | BTC 5-minute trade prints | every print | Oct 26 | 736,785 | 91.9% |
| Polymarket US | BTC 15-minute and 1-hour books, depth and trades | every update | Oct 27 | 1,142,912 | 92.0% |
| Polymarket US | SOL and XRP 5-minute books | every update | Oct 27 | 799,499 | 92.6% |
| Polymarket US | Quote snapshots, start prices and outcomes (BTC, ETH, SOL, XRP) | about every 7 s; every outcome | Oct 26 | 179,583 | 100% |
| Coinbase | BTC spot trades and a level-2 summary | every trade | Oct 26 | 652,123 | 99.98% |
| Kraken, Bitstamp | BTC spot trades | every trade | Oct 26 | 276,065 | 100% |
| OKX, Hyperliquid | BTC perpetual trades | every trade | Oct 26 | 2,543,504 | 100% |
| Gemini, Bullish, Crypto.com | BTC top of book and trades | every update | Oct 27 | 1,240,309 | 100% |
| SOL, XRP spot | Reference spot for the alt rounds | every update | Oct 27 | 219,596 | 100% |
| CF Benchmarks | BRTI | once a second | Oct 28 | 80,396 | 93.1% |
| Kalshi | BTC 15-minute book and trades | paused | n/a | 0 | paused |
Proof
Every day, sealed the next morning.
- One compressed file per recorder. Each night the previous UTC day is cut into one zstd-compressed SQLite file per recorder,
same tables and columns as the live database, checked with
zstd -t. - A manifest of hashes.
MANIFEST.sha256lists the SHA-256 of every file, plus the day’s row counts. - A timestamp on Bitcoin. The manifest is stamped with OpenTimestamps, so anyone can later prove it existed on that date and has not changed since.
- Two copies, two nodes. One copy on a NAS on our first node, one on our second, each re-checked with
sha256sum -cafter copying. - A gap register. Every silence longer than a stream’s threshold is listed with its start, end and length. We publish the summary, including the bad days.
Licensing
Venue data belongs to the venues.
We do not redistribute it without a licence. Derived data we build ourselves (gap registers, feed-health logs, our own composite spot) and licensed datasets will be offered to research and commercial teams as agreements allow. Tell us what you need and we will tell you what we can ship.
For traders
Ledger: your own book, reconciled.
Your fills, fees, P&L and tax reconciliation across Polymarket US, Kalshi and Robinhood, built from your own exports and read-only keys. Early access winter 2026/27. Not tax advice.