A liquidation cascade is the cleanest example of non-informational selling in crypto. When price drops fast, over-levered longs get force-closed at any price — the exchange is not selling because it learned something, it is selling because a margin engine fired. When the forced flow exhausts, price tends to snap back. The trade is the lag between the flush and the reprice.
Why the edge exists
Prediction-market and binary-option books reprice late. Right after a BTC flush, "UP this hour" is often cheaper than the spot tape justifies, because the book is still digesting the move. That gap — between forced-selling exhaustion and the book catching up — is a window you can measure, not guess.
The catch: you can only act on this if your liquidation, flow and funding values are point-in-time. A backtest that uses the revised liquidation print from five minutes later will look brilliant and lose money live. That is exactly the leak QUANT_API removes.
The signal set
Three real QUANT_API signals, polled together, define the trigger:
liquidations.liq_imbalance@15m:zscore— long-vs-short forced volume, standardized. A deep negative z means longs just got flushed.flow.ofi@5m:zscore— order-flow imbalance. The reversal confirms when aggressive flow flips from selling to buying.funding.funding_avg@1h— cross-venue funding. Negative funding means the crowd is still positioned short, leaving fuel for the squeeze up.
The rule
Buy the UP binary when liq_imbalance z < −2 (longs flushed) AND ofi z flips > +1 (flow turns) AND funding < 0 (crowd still short). One line, three real values, evaluated on every new bar.
curl 'https://api.quant-api.dev/v1/features/live?asset=BTC&features=\ liquidations.liq_imbalance@15m:zscore,\ flow.ofi@5m:zscore,\ funding.funding_avg@1h' \ -H 'Authorization: Bearer fk_live_…' # poll every minute
Prove it before you size it
Resolve the same three features at the timestamps of every past post-cascade hour with one POST /v1/features/historical call. Every value comes back exactly as it stood at that minute — leak-free by construction. Measure the hit-rate of the rule against the 50/50 prior. Clears your bar → size it. Does not → you killed a bad idea in an afternoon. Both outcomes are wins.
Want the whole thing wired step by step in a dashboard? The guided first-system playbook walks you from a fresh account to a firing rule, plan-aware.