What we actually captured.
Every vendor's data has gaps. Most don't tell you where. Below is our measured record for each underlying — per-day capture, cadence percentiles, and every interruption longer than five minutes with its UTC timestamps.
It's generated from the database rather than written, so it goes stale in hours rather than quarters. If you're evaluating us against another source, ask them for the same page.
XRP · day by day
- coverage
- 99.77%
- cadence p95
- 60s
- cadence p99
- 60s
- lost
- 1.1h of 480h
Gaps over five minutes · XRP
Every interruption, investigated, with what we observed. Where the logs no longer reach the answer is an honest "not established" rather than a likely story. The full record, including failures that cost no data, is on the operations page.
| From (UTC) | To | Duration | Cause |
|---|---|---|---|
| 2026-09-05 08:59 | 09:36 | 37 min | exchange Deribit returned HTTP 503 Service Unavailable continuously for the whole window. Capture resumed within a minute of the last refusal. The L2 book daemon separately exhausted its reconnect budget and was restarted by systemd 270 times during the same period; that is our defect and is being fixed, but it did not lengthen this interruption — the exchange was refusing every request throughout, so there was nothing to capture, and chain collection resumed as soon as the 503s stopped. |
| 2026-09-01 06:01 | 06:11 | 10 min | our fault Deribit closed every one of our websocket connections at the same second, without a close frame. That alone is routine and recoverable. What turned it into a ten-minute interruption is our reconnect loop: the backoff counter resets on a successful CONNECT rather than on a connection that survives, so a socket that establishes and immediately dies restarts the delay at one second indefinitely. It never escalated. Two hundred reconnect cycles followed, each re-subscribing every channel, and Deribit answered 820 times with 429 Too Many Requests. Capture resumed when the cycle settled. This is the third rate-limit incident of the same family after 2026-07-14 and 2026-08-13; the fix made in August taught the REST retry path to honour Retry-After and did not touch the websocket reconnect path, which is where this one happened. |
| 2026-08-25 08:59 | 09:15 | 16 min | exchange Deribit returned HTTP 503 Service Unavailable. |
How this is measured
Time coverage, not a snapshot target. We report the fraction of the elapsed window during which the collector was actually capturing, derived from measured gaps. There is deliberately no "expected snapshots per day" denominator: the capture cadence changed from ~82 s to 60 s on 2026-08-12, and any fixed denominator would misreport one era or the other.
A gap is over five minutes. Comfortably above any normal spacing, so ordinary cadence is never counted as an outage. Each one is investigated and its cause published on the operations record — the exchange returning 503, our own rate-limit breach, or, where the logs no longer reach, an honest "not established".
Live capture and archive are separate. The figures above cover our own continuous capture since 2026-04-21. The historical BTC archive is a different provenance and is described on each currency's page and in the methodology.
Daily OHLC bars
Daily open/high/low/close/volume bars for
each base's own Deribit PERPETUAL instrument (e.g.
BTC-PERPETUAL).
These are PERPETUAL instrument bars, not index bars, and not options data — Deribit does not publish an index OHLC series,
and this is a different instrument from every options chain, surface,
greek and trade-tape dataset on this page. Every bar carries
instrument_name, so the series is identifiable from the
payload alone, not only from this page.
Coverage start not yet published. The backfill that would measure this dataset's earliest available day has not run. We publish a start date only once it has been measured against what actually landed — not before.
Trade tape, liquidations and blocks
Coverage start dates for BTC, empirically established by probing Deribit's own history API across calm days and every major crash — not rounded, not softened. These probes were run on BTC. Deribit did not list ETH options until March 2019, SOL options until March 2024, or XRP options until March 2024 (TRX and AVAX not until January 2026), so on those currencies no dataset starts earlier than its own tape does; each currency's coverage page carries its own dates.
| Dataset | Coverage start (BTC) |
|---|---|
| Options liquidations | November 2018 |
| Block trades | September 2019 (meaningful volume from January 2020) |
| Combo / multi-leg flags | November 2022 |
History is still landing. These datasets are being built now, currency by currency, as the historical backfill runs. Until a currency's backfill completes, any month for it may return 404 simply because the file has not been built yet — which is a different thing from the month being empty. The full history back to the dates above arrives as each currency's backfill lands.
Liquidations are event-driven, not a steady rate: most months legitimately
contain zero rows, and empty months are never written. Once a month is
built, a 404 on liquidations/{month} means exactly that — no
qualifying rows — not a gap. Both derived datasets are Pro tier, same as
the trade tape itself — no new SKU, no extra price.