wavelab
wavelab counts Elliott waves on live BTCUSDT and, for every count, marks the price at which that count would be wrong. It is a research engine, not a signal service — and the honest result of nine years of it is the reason Assay exists.
What nine years said
Every count the engine produced was replayed causally over the whole history, with the entry, the stop and the target it would have drawn at the time. 1,449 trades resolved. Then each signal was paired with a twin entered at a random nearby instant, with the same geometry — which is what separates "does the wave count contribute anything?" from "did BTC go up?".
| Measure | Elliott | Random twin |
|---|---|---|
| Expectancy | +0.007R | +0.015R |
| Longs | +0.144R | +0.165R |
| Shorts | −0.107R | −0.109R |
| Fibonacci fit vs outcome | rho +0.0038, p 0.886 | |
The random twin reproduces the asymmetry exactly. Over a period in which BTC rose 1,748%, buying worked and selling did not — and it made no difference whether the entry was chosen by Elliott or by a die. The bootstrap interval on the expectancy is [−0.076R, +0.090R]: it contains zero with room on both sides.
Which is how Assay happened
A result like that is only worth having if you can tell it from a flattering one. The machinery that produced it — the causal replay, the purged walk-forward, the random controls, the paired null arm — turned out to be the useful part, so it was pointed at any strategy instead of just this one. That is Assay: five tests that almost nothing passes.