How this works: on every tick we run Learn.autoTagSetup() against every symbol in QUOTES. That returns a base score (0–1) and a setup label. We multiply by the learned weight for that setup (from Learn.weights) to get the adjusted score. Weights start at 1.0 and drift up/down as the engine sees realized R-multiples. A setup with weight 1.3 means historical edge of +30% above baseline.