DetectorKilled

Ride liquidation cascades as they accelerate

−1.14 bps

average return after a cascade, every group of cases lost money

bps = basis points = a hundredth of a percent (100 bps = 1%).

In plain English

Liquidation cascades bounce back rather than keep running, so riding them as they speed up loses money.

What was claimed

The bigger a liquidation cascade gets, the harder it keeps running, so ride the acceleration.

The bar we locked

bigger cascades should keep running harder

What the data said

they lose money, and the pattern doesn’t hold at any size

Why it's dead

Cascades mean-revert, they don't continue. Same finding as COUNTER-001 and the HLP cascade detector, approached from a different angle.

The detail

n=15,992 cascades. D10 (the top tenth, the densest cascades) mean at 5m: −1.14 bps. Monotonicity check (if the theory held, each step up in cascade size should step up the return): 0/3 adjacent decile pairs showed the expected continuation gradient. Below any cost model.

Kill date: 2026-03-26
Sample: n=15,992
Method: Pre-registered live test
Verdict: same result, different angle

What it would have done to your money

You put in$10,000

You run it for100 trades

You would have

$9,887

−$113

lost (1%)

You started with$10,000

This strategy left you$9,887

Instead of not trading$10,000

Locked to this strategy's real measured result of -0.0114% per trade, compounded. You choose the amount and the time. We don't choose the return, the strategy already did.