Strategies here are discovered, not hand-picked. We sweep thousands of parameter combinations per strategy across 2017–2026, and a genetic search evolves the survivors further — exploring corners a fixed grid never reaches.
Each candidate is split 70/30 into an in-sample stretch it’s allowed to “see” and an out-of-sample stretch held back. It earns the ✓ Proven badge only if it clears all four gates on the unseen out-of-sample data:
100+ tradesa track record, not a small-sample fluke
OOS Sharpe > 0.5real risk-adjusted edge on unseen data
IS→OOS ratio > 0.6held up out-of-sample — it didn’t just curve-fit the past
Costs < 30% of grosssurvives real brokerage & slippage
Duplicates collapse to one canonical strategy, a bootstrap resample estimates the probability the edge is real, and the whole board is re-validated nightly. What you see is what held up — not what looked good in hindsight.