Are the Exchanges Healthy Right Now?

Each venue's own status endpoint, polled every five minutes and stored only when it changes. Uptime is time weighted over 30 days; incidents count moves away from ok in the last 7 days.

An endpoint that does not answer is recorded as unreachable, which is a fact about the venue's API, not necessarily about trading.

The age of the newest row in each recorded stream. This is our network's view: a stale row means the venue or our collector had a problem, and we say which honestly rather than dressing it up as an exchange SLA.

Liquidation feeds are naturally sparse in quiet hours; thresholds account for each feed's own cadence.

Every 10 minutes our collector polls free public endpoints (no keys; non-dollar quotes convert through the ECB daily fix) for a 30-coin basket: price, 24h volume, open interest and the 8h-equivalent funding rate. This table is the read side of that layer; the same numbers feed the API. The latest poll returned 67 feeds, 37 perpetual and 30 spot, on 53 exchanges.

Snapshot 2026-09-28 15:10 UTC · 53 venues

Per venue

Stablecoin pegs across spot venues

Open interest is each venue's own figure converted to USD at its last price. Kraken Futures funding is converted from an hourly absolute amount to an 8h rate; Hyperliquid and dYdX from hourly to 8h. Venues that answered with nothing in the last poll are simply absent from the table. WhiteBIT and BitMart appear per venue but stay out of every open interest total and the DEX share; we do not trust their self-reported open interest. A fraction after a total, such as (25/27), means that only 25 of the 27 venues that publish open interest reported it for that coin in this snapshot.

Hourly sum of perpetual open interest across every venue we snapshot (exchanges and DEXs; WhiteBIT and BitMart excluded), the DEX share of it, and the open-interest-weighted funding right now. One series, all venues, from our own 10-minute snapshots.

Recorded since 2026-08-21. Venues joined over time (the largest additions on 2026-09-02 and 2026-09-04), so the total steps up when a venue is added; /api/public/venues/markets lists the venues in the latest snapshot. WhiteBIT and BitMart self-report open interest we could not reconcile, so they are excluded here and in the liquidation estimate.

One-minute returns of two venues, aligned by minute and cross-correlated at lags of -3 to +3 minutes, one read per UTC day. The lead score is corr(+1) minus corr(-1): positive means the first venue's move shows up in the second one a minute later more than the reverse. Coupling is the same-minute correlation. Plain arithmetic on our own minute bars, not an information-share model.

Days with fewer than 600 aligned minutes are skipped. A score inside ±0.02 is a tie. Sub-minute leadership cannot be seen at this resolution; this table answers the minute-scale question only.

Perp DEX share of open interest

Open interest on the on-chain perp venues we snapshot (Hyperliquid, dYdX, Aster, Paradex, Lighter, GMX and Jupiter Perps) as a share of all perp open interest in the same snapshot, WhiteBIT and BitMart excluded, recorded daily since 2026-08-21. Lighter, GMX and Jupiter Perps joined the snapshot on 2026-09-02.

Stable-asset balances of the venues' liquidation insurance funds, recorded once per UTC day. A fund that grinds down during cascades is doing its job; a fund that keeps shrinking without cascades is a question worth asking.

No venue publishes historical balances. This archive exists because we record it forward, which is exactly why it starts recently and cannot be backfilled.

Every venue with a recorded dataset has its own page: feeds, markets, networks and incidents in one place.

Insurance balances come from the venues' public endpoints (Binance USDT-M insuranceBalance, Bybit v5 insurance), summing stable-asset balances only, recorded nightly at 00:20 UTC. OKX and others do not expose comparable public balances; we list what can be verified, not what cannot.

Feed status derives from our own recorded tables: real liquidation prints, the Deribit and OKX options tape, Binance open-interest snapshots and the spot minute recorders. It measures the venue AND our collector together; that compound honesty is the point, since it is exactly what downstream data quality depends on.

An insurance fund is a cushion, not a solvency proof. Balances can be topped up, split across pools or redefined by the venue. Treat trends and drawdowns as questions to investigate, not verdicts.