Platform status
Data lag: 0 h · uptime is measured by an external independent watchdog, not by us.
The external watchdog starts reporting with the tunnel — the chart fills in as soon as the first probes land.
Quality, last days
| Day | Q | Rows | Gaps | Book rebuild | Full replay | Tape vs exchange |
|---|---|---|---|---|---|---|
| 2026-08-27 | 65 (BTC 68 · ETH 65 · SOL 68) | 5049940 | 0 | failed | 2,454,452/2,454,488 check | 3/3 symbols, 0 missing |
| 2026-08-26 | 94 (BTC 94 · ETH 95 · SOL 96) | 4495685 | 0 | 12/12 ok | 2,399,147/2,399,154 ok | 3/3 symbols, 0 missing |
| 2026-08-25 | 84 (BTC 84 · ETH 86 · SOL 84) | 5399766 | 0 | 12/12 ok | 2,447,560/2,447,574 ok | 3/3 symbols, 0 missing |
| 2026-08-24 | 92 (BTC 95 · ETH 93 · SOL 92) | 5543360 | 0 | 12/12 ok | 2,438,176/2,438,188 ok | 3/3 symbols, 0 missing |
| 2026-08-23 | 68 (BTC 68 · ETH 68 · SOL 68) | 4673382 | 0 | failed | 2,380,313/2,380,323 ok | 3/3 symbols, 0 missing |
| 2026-08-22 | 87 (BTC 87 · ETH 87 · SOL 93) | 5278092 | 54 | 12/12 ok | 2,431,554/2,431,562 ok | 3/3 symbols, 0 missing |
| 2026-08-21 | 93 (BTC 93 · ETH 93 · SOL 94) | 7459728 | 82 | failed | 2,521,710/2,522,808 ok | 3/3 symbols, 0 missing |
| 2026-08-13 | 96 (BTC 96 · ETH 96 · SOL 96) | 3125052 | 26 | 12/12 ok | 2,364,902/2,365,635 ok | 3/3 symbols, 0 missing |
| 2026-08-20 | 95 (BTC 95 · ETH 95 · SOL 95) | 5847570 | 62 | 12/12 ok | 2,521,035/2,523,704 ok | 3/3 symbols, 0 missing |
| 2026-08-19 | 87 (BTC 93 · ETH 93 · SOL 87) | 5453832 | 52 | 12/12 ok | 2,435,654/2,437,364 ok | 3/3 symbols, 0 missing |
| 2026-08-18 | 97 (BTC 97 · ETH 97 · SOL 97) | 2910514 | 24 | 11/12 ok | 2,285,456/2,285,933 ok | 3/3 symbols, 0 missing |
| 2026-08-17 | 96 (BTC 96 · ETH 96 · SOL 96) | 3128708 | 28 | 12/12 ok | 2,291,697/2,292,417 ok | 3/3 symbols, 0 missing |
| 2026-08-16 | 97 (BTC 97 · ETH 97 · SOL 97) | 2402979 | 16 | 12/12 ok | 2,211,987/2,212,250 ok | 3/3 symbols, 0 missing |
| 2026-08-15 | 97 (BTC 97 · ETH 97 · SOL 97) | 2398031 | 15 | 12/12 ok | 2,199,933/2,200,066 ok | 3/3 symbols, 0 missing |
Q is the day quality index: it is computed from the share of collection time lost and from sequence gaps — measured, not declared. Rows below 70 are highlighted. Book rebuild: the order book is rebuilt from our l2 diffs and compared with the independently recorded top of book. Tape vs exchange: every trade_id of the day is compared with Binance's own public archive (data.binance.vision) — "0 missing" means the tape is identical to the exchange's record, row for row.
Incidents
27.08 23:59 Incomplete data collection on 2026-08-27
open
Collection on 2026-08-27 was incomplete. Quality index Q = 65 (healthy is 90 and above). 0.21% of the day was lost, holes longer than 5 seconds: 3, the longest one 1m. Records in the raw stream: 5049940. Delivery latency tail was high on ETH/USDT: p99 = 2915.0 ms (penalty 29.06 of 30). Rows are complete; timestamps are exchange time, so the data itself is unaffected — only how late we received it. The order book for this day does NOT rebuild reliably: only None% of bid/ask probes matched the independently recorded top of book, and we are not hiding it. The day is left as it is — backfilling after the fact would mean the numbers on the storefront change after the sale. The day is flagged in the data map (has_gaps), and every backtest report that touches it carries a «day with collection gaps» badge.
23.08 23:59 Incomplete data collection on 2026-08-23
open
Collection on 2026-08-23 was incomplete. Quality index Q = 68 (healthy is 90 and above). 0.08% of the day was lost, holes longer than 5 seconds: 2, the longest one 0m. Records in the raw stream: 4673382. Delivery latency tail was high on BTC/USDT: p99 = 3487.0 ms (penalty 30.0 of 30). Rows are complete; timestamps are exchange time, so the data itself is unaffected — only how late we received it. The order book for this day does NOT rebuild reliably: only None% of bid/ask probes matched the independently recorded top of book, and we are not hiding it. The day is left as it is — backfilling after the fact would mean the numbers on the storefront change after the sale. The day is flagged in the data map (has_gaps), and every backtest report that touches it carries a «day with collection gaps» badge.
21.08 23:59 Incomplete data collection on 2026-08-21
open
Collection on 2026-08-21 was incomplete. Quality index Q = 93 (healthy is 90 and above). 0.02% of the day was lost, holes longer than 5 seconds: 2, the longest one 0m. Records in the raw stream: 7459728. Delivery latency tail was high on BTC/USDT: p99 = 426.0 ms (penalty 1.4 of 30). Rows are complete; timestamps are exchange time, so the data itself is unaffected — only how late we received it. The order book for this day does NOT rebuild reliably: only None% of bid/ask probes matched the independently recorded top of book, and we are not hiding it. The day is left as it is — backfilling after the fact would mean the numbers on the storefront change after the sale. The day is flagged in the data map (has_gaps), and every backtest report that touches it carries a «day with collection gaps» badge.
11.08 23:59 Incomplete data collection on 2026-08-11
open
Collection on 2026-08-11 was incomplete. Quality index Q = 67 (healthy is 90 and above). 3.00% of the day was lost, holes longer than 5 seconds: 49, the longest one 2m. Records in the raw stream: 3044981. The order book for this day rebuilds correctly — the snapshot reconstruction check passed. The day is left as it is — backfilling after the fact would mean the numbers on the storefront change after the sale. The day is flagged in the data map (has_gaps), and every backtest report that touches it carries a «day with collection gaps» badge.
10.08 23:59 Incomplete data collection on 2026-08-10
open
Collection on 2026-08-10 was incomplete. Quality index Q = 67 (healthy is 90 and above). 1.18% of the day was lost, holes longer than 5 seconds: 8, the longest one 10m. Records in the raw stream: 3251112. The order book for this day rebuilds correctly — the snapshot reconstruction check passed. The day is left as it is — backfilling after the fact would mean the numbers on the storefront change after the sale. The day is flagged in the data map (has_gaps), and every backtest report that touches it carries a «day with collection gaps» badge.
08.08 23:59 Incomplete data collection on 2026-08-08
open
Collection on 2026-08-08 was incomplete. Quality index Q = 70 (healthy is 90 and above). 38.81% of the day was lost, holes longer than 5 seconds: 6, the longest one 9h 16m. Records in the raw stream: 1621574. The order book for this day does NOT rebuild: the snapshot reconstruction check failed, and we are not hiding it. The day is left as it is — backfilling after the fact would mean the numbers on the storefront change after the sale. The day is flagged in the data map (has_gaps), and every backtest report that touches it carries a «day with collection gaps» badge.