Liquidation maps show where forced selling is. We tested whether it moves price.
228 independent episodes across 28 symbols. Forced flow returned +40.07 bps at 15 minutes and beat a permutation null. A random minute in the same symbol in the same hour returned 44.52 bps.
Every liquidation product on the market shows you where forced selling sits. Coinglass calls its heatmap a probability map and attaches no probability to anything. None of them publishes whether reaching a cluster actually does anything.
We recorded 61,338 forced-order events across 757 symbols on Binance USD-M and collapsed them into 228 independent episodes at $250k or more. Episodes, not events: 90% of large liquidations fall within 60 seconds of another, so counting events as observations inflates the sample threefold.
The result
| Horizon | Mean | Hit rate | Permutation p95 | Matched control |
|---|---|---|---|---|
| 1 min | +6.27 bps | 0.519 | 7.93 — fails | 4.96 — clears |
| 5 min | +11.10 bps | 0.557 | 14.99 — fails | 25.13 — fails |
| 15 min | +40.07 bps | 0.604 | 17.06 — clears | 44.52 — fails |
At 15 minutes the effect is large: +40 bps with a 60% hit rate, comfortably beating a permutation null. That number alone would make a compelling product.
Then take any other minute from the same hour in the same symbol. 44.52 bps. More.
What it means
Liquidations happen because the market is already moving. The move is the volatility; the liquidation is a symptom of it. A liquidation map is a map of where leverage sat, not a map of what will happen next.
Limitations, stated
Three days of recording. One venue. 106 of 228 episodes had usable price data, biasing the surviving sample toward liquid symbols — which flatters a liquidity-driven effect rather than hiding one. And this is Binance: it says nothing about Hyperliquid, whose HLP backstop and far smaller book differ in ways that could cut either way. That test is running.
The failure is against a matched control, not a power threshold. More data tightens the intervals; it does not repair a comparison the effect loses on its merits.