On any given day, the market is either drowning out stock-picking signals or letting them through. Traditional metrics ignore this. We built a system that doesn't.
When the market moves sharply, correlations spike and every stock becomes a beta instrument. Measuring alpha on those days is like grading a pilot's skill during a hurricane. You're measuring the storm, not the pilot.
Every conventional metric blends signal days with noise days into a single number. This system stops doing that.
Each trading day is classified using a threshold derived from market structure, not optimized on any portfolio's returns. The cutoff separates days where idiosyncratic returns can surface from days where beta dominates everything.
Instead of "did this manager beat the market," ask: "on the days when stock-picking could register in returns, did this manager pick stocks well?" Harder to answer. More honest.
Separate genuine skill from regime noise before committing capital.
Show skill with precision, on exactly the days it can be measured.
A falsifiable framework. The validation design tests for false positives before testing for true ones.
Regime classification, market environment, alpha decomposition, and Storm Center. Running now on FlightDeck with real data.
Open FlightDeck