Price Dislocation Analysis
The decision: a price on one venue gapped away from the rest of the market. Real liquidity event, deliberate pressure, or a handful of prints on a thin venue at a price no size could have transacted at?
The raw discrepancy number treats all three identically: a 2.5% deviation on a venue that traded $8,000 in a minute looks the same as one backed by $8,000,000. The fix is one step — gate on the traded volume behind the print before examining the size of the gap.
Step 1 — Get the gap, and what was behind it
Two requests: the cross-market endpoint with disc, which returns which venue printed the interval's highest and lowest price and how far each sat from the cross-market average; and per-venue ohlcv plus vwap for the same intervals, which gives the volume behind each extreme. Then discard any dislocation where the named venue's traded quote volume falls below the chosen volume floor.
A reasonable floor is the 25th percentile of per-interval quote volume across trusted venues on that pair. If it removes more than about half the dislocations, it is too high and small venues that genuinely matter are being screened out.
Two properties of discrepancy to note when reading it: both values arrive as positive numbers, measured in opposite directions from the VWAP, and at most two venues are reported per interval, so the named venue is an example of the extreme rather than necessarily the only one.
Step 2 — Classify what survived
| Pattern | Observation | Most likely |
|---|---|---|
| Real liquidity event | Gap on one venue, meaningful volume behind it, resolves in minutes, other venues move partway toward it | A large order consumed the book. Normal functioning. |
| Market-wide move | Every venue gaps the same direction in the same minute | News or a scheduled event. Not a venue finding. |
| Stale or broken feed | One venue sits away from the market and does not resolve, with little or no volume | Outage, halt, or an ingest problem. |
| A push | Gap on one venue, price returns quickly, minimal volume at the extreme, and it recurs | Someone marking a price. The one worth escalating. |
The discriminator between the last two is recurrence. A single unresolved gap is a data problem; the same venue gapping at the same time of day, repeatedly, on low volume, is a behaviour.
Check the clock first, always. Macro releases, expiries, index rebalances and funding timestamps all produce genuine simultaneous dislocations, and a dislocation that lines up with a scheduled event is the market working.
Step 3 — When it keeps happening, test the venue
If a venue keeps appearing at the extreme, the question stops being about the minute. Run Volume Integrity Assessment on it. Recurring extremes together with a large volume share are hard to explain benignly.
Stablecoin Depeg Detection
Handled by a dedicated endpoint and its own page: a fixed 1.0000 peg reference, a 1% threshold floor, and a curated dollar-pegged pair list. The short version is to gate on traded volume even harder than here, and to prefer -usd pairs over stablecoin-quoted ones. Full workflow in Stablecoin Depeg Detection.
Explanations to rule out first
A thin venue read as a market event is the most common failure by a wide margin; gate on traded volume first. Check the calendar before treating anything as venue-specific. And since venues do not print at identical instants, a dislocation that appears in exactly one interval and vanishes deserves suspicion before escalation. For volume figures intended to be relied on, use per-venue requests rather than the cross-market endpoint.