Audit: devpyle (contributed)
What this page is for: the problems with this corpus and the evidence behind them. The files are on the download page, and what the columns mean is in the format notes.
What is wrong with this data
A handful of crypto markets, in the contributor's own format, recorded by one machine and attested by nobody here.
- It covers crypto Up-or-Down markets only. BTC, ETH, SOL, XRP, BNB, DOGE and HYPE; 5-minute, 15-minute and 1-hour markets. Every other market in these 68 hours is absent from this tier, so it does not close the gap for the venue as a whole. the contribution audit, 2026-09-16
- Its book is five levels deep, at the recorder's receipt time. clob_l2 is a 5-level snapshot, not the venue's full book, and carries no venue timestamps. poly_trades are last-trade events. the contribution audit, 2026-09-16
- Timestamps are his receipt time, in milliseconds. int64 ms from one NTP host, measured about 80 ms behind the venue, p95 230 to 800 ms. Prices are float32: compare them with a 1e-3 tolerance. the contribution audit, 2026-09-16
- Rows are not de-duplicated across his recorder restarts. De-duplicate yourself before counting events around a restart. the contribution audit, 2026-09-16
Measured against the chain
This corpus has not been graded against on-chain fills yet. That is a gap in the audit, not a verdict on the data. What the audit has covered so far.
Markets held, against markets traded
Not measured for this corpus: the audit drew no hours from it. The corpora that were measured are compared side by side, each against our own capture as the benchmark.