ORACLE OF BTC SHOWS ITS WORK

Research write-ups

Every experiment here was pre-registered before its data was touched. Nulls get published with the same prominence as survivors — see how this works.

2026-07-11

We tested 49 machine-learning models on Bitcoin. Zero survived.

A full ML pipeline — walk-forward, cost-adjusted, multiplicity-controlled across 532,780 hypotheses — produced no deployable model. Publishing the null is the point.

Read