This isn't a marketing claim - it's the same scoring logic the live screener uses, walked forward bar-by-bar over ~730 days of real historical price data per market, opening a hypothetical trade every time the composite score would have cleared ±40 and resolving it against the same ATR-based stop/target the live screener would have set. No lookahead: every score at bar t only uses data available up to bar t.
One scoring approach, backtested bar-by-bar over ~730 days of real price history per market - not a snapshot claim. Every signal requires an active chart or harmonic pattern (not just trend alignment) and a composite score that clears a backtested threshold before it qualifies.
Event-driven, not napkin math - concurrent positions, risk-based sizing (€10,000 starting, 1% risk/trade), capped at 10% total risk-at-once.
chart_pattern and harmonic_pattern are 40%+ of the composite weight combined - trend/momentum/breakout alignment alone isn't enough, there has to be a real, currently-forming chart or harmonic structure (triangle, wedge, head & shoulders, ABCD, Gartley, Butterfly…) behind the signal too.
A Head & Shoulders and an Inverse Head & Shoulders aren't treated as equally reliable - each of the 20+ named patterns is scaled by its own backtested hit rate, so a historically weak pattern contributes less to the score than a strong one instead of every pattern counting the same.
The ±40 qualifying bar isn't arbitrary - the reward:risk sweep below shows score magnitude only meaningfully separates signal quality above roughly this level; below it, hit rate is flat regardless of score.