Mean-reversionKilled

Betting the price “fades” off the day’s high or low

−1.72 bps vs +7.92

the bounce we saw vs what we’d expect, the wrong way, and far too small

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

In plain English

We bet prices would bounce off the day’s high and low; they didn’t. The setup is rare, the move was tiny, and it went the opposite way to the bet.

What was claimed

When price touches the day's high or low it bounces back, and the bounce is big enough to trade.

The bar we locked

at least 50 cases, and a bounce big enough to cover costs

What the data said

only 12 cases, and the price moved the wrong way

Why it's dead

Boundary-fade events (price touching the day's high or low, then pulling back) are structurally rare (~0.5 per symbol per month), and where they occur the price doesn't fade back, it's the wrong sign and ~50× below the magnitude needed to clear costs.

The detail

Day-30 verdict: n=12 events vs a 50-event locked floor (the minimum sample size we committed to before seeing any data); observed mean revert −1.72 bps against a +7.92 bps baseline, wrong-signed and ~50× below the 50 bps promote floor (the effect size a signal must show before we fund it). For R1 to pass, the true per-event effect would have to be 25–50× the observed estimate.

Kill date: 2026-06-12
Sample: n=12 events
Method: Pre-registered live test
Verdict: rare + wrong-signed