Stale market data. Counted on every venue.
How often is exchange market data stale? DEBYKO sampled every collected listing 4 times a day, 19 days to 2026-10-09: 88,980 values of ticker, mark, funding, open interest and 24 h stats on 19 venues: 93.0% current, 3.2% stale, 3.6% halted, 0.2% missing. Halted is counted apart from stale: the API answers both as stale, but a halted value is not late, the venue said it does not trade the listing.
How often exchange market data was stale.
| Layer | Samples | Current | Stale | Halted | Missing | Most common reason |
|---|---|---|---|---|---|---|
| ticker | 17,496 | 97.7% | 2.3% | 0.0% | 0.0% | horizon_passed |
| mark | 16,140 | 92.0% | 3.0% | 5.0% | 0.07% | halted |
| funding | 18,775 | 91.9% | 3.8% | 4.3% | 0.0% | halted |
| oi | 18,665 | 91.7% | 3.9% | 4.3% | 0.06% | halted |
| stats | 17,904 | 91.9% | 2.9% | 4.5% | 0.7% | halted |
Why a value was stale.
- halted: 3,232 samples, 53.3% of the stale ones. The venue said it does not trade the listing.
- horizon_passed: 1,744 samples, 28.8% of the stale ones. No new value within its horizon.
- connection_lost: 731 samples, 12.1% of the stale ones. The connection that delivered the value had broken.
- source_error: 355 samples, 5.9% of the stale ones. The venue refused the requests that would have replaced the value.
Stale share per venue and layer.
| Venue | Listings | Stale | ticker | mark | funding | oi | stats | Most common reason |
|---|---|---|---|---|---|---|---|---|
| ASTER-PERP | 26 | 4.1% | 1.1% | 4.7% | 4.7% | 4.7% | 5.4% | horizon_passed |
| AVANTIS-PERP | 45 | 2.7% | — | — | 2.7% | 2.7% | — | horizon_passed |
| BINANCE-SPOT | 12 | 0.0% | 0.0% | — | — | — | 0.0% | — |
| BINANCE-USDM | 42 | 4.3% | 0.3% | 5.3% | 5.3% | 5.3% | 5.3% | horizon_passed |
| BITGET-PERP | 25 | 2.4% | 2.3% | 2.5% | 2.3% | 2.3% | 2.3% | horizon_passed |
| BYBIT-PERP | 43 | 2.7% | 2.7% | 2.6% | 2.6% | 2.6% | 2.6% | horizon_passed |
| BYBIT-SPOT | 10 | 0.0% | 0.0% | — | — | — | 0.0% | — |
| COINBASE-INTX | 24 | 91.9% | 22.6% | 92.1% | 92.1% | 92.1% | 98.7% | halted |
| DYDX-PERP | 25 | 3.4% | 2.3% | — | 2.3% | 5.3% | 3.7% | connection_lost |
| GATE-PERP | 24 | 1.9% | 0.0% | 2.5% | 2.3% | 2.3% | 2.5% | horizon_passed |
| GMX-PERP | 36 | 2.6% | 2.6% | 2.7% | 2.6% | 2.6% | 2.6% | horizon_passed |
| HYPERLIQUID | 24 | 3.3% | 4.1% | 3.0% | 3.5% | 3.0% | 3.0% | connection_lost |
| KRAKEN-FUTURES | 98 | 3.4% | 3.4% | 3.4% | 3.4% | 3.4% | 3.4% | connection_lost |
| MEXC-PERP | 44 | 2.6% | 2.6% | 2.7% | 2.6% | 2.7% | 2.7% | horizon_passed |
| NADO-PERP | 22 | 9.2% | 2.1% | 11.1% | 10.3% | 11.1% | 11.1% | halted |
| OKX-PERP | 29 | 1.7% | 1.7% | 1.7% | 1.7% | 1.7% | 1.8% | horizon_passed |
| OKX-SPOT | 10 | 0.0% | 0.0% | — | — | — | 0.0% | — |
| SYNTHETIX-PERP | 10 | 1.5% | 2.1% | 1.3% | 1.3% | 1.3% | 1.3% | horizon_passed |
| WEEX-FUTURES | 27 | 8.8% | 3.0% | 1.9% | 18.5% | 18.5% | 2.0% | source_error |
How this was measured.
- Data window: 2026-09-20 00:00 UTC to 2026-10-09 00:00 UTC, 19 days. The job asks for 30 days; stored history of the sampled listings begins later.
- Computed at: 2026-10-09 19:44 UTC, by a job that recomputes the note on the first of every month.
Every listing DEBYKO collected at the computation, from its first registration, sampled at 03:17, 09:17, 15:17, 21:17 UTC every day. Each sample is what POST /v2/snapshots/history with at answers for the listing and layer at that instant: the newest stored value, judged by the cadence facts, connection breaks and halts recorded by then. Current means the API would have returned the number; stale that it would have returned the reason instead; missing that nothing was stored. A layer the venue does not publish, or the platform does not collect, is not counted; 5,515 samples older than what a listing's history keeps are left out.