Step 3 – Strategy validation, regime detection and UI improvements
Main changes:
- Implemented multi-horizon trend regime detection in src/core/market_regime.py
- EMA(20, 50, 100, 200) trend score model
- 5 regime states: bull_strong, bull_moderate, sideways, bear_moderate, bear_strong
- asymmetric persistence filter (bull 5 bars, sideways 3 bars, bear 2 bars)
- window regime classification using average trend score
- Extended walk-forward outputs
- window_regimes now include:
- regime_detail
- bull_strong_pct
- bull_moderate_pct
- sideways_detail_pct
- bear_moderate_pct
- bear_strong_pct
- avg_score
- UI improvements in Step 3
- regime analysis cards redesigned to show 5 regimes
- visual "M layout": bull regimes on top, sideways + bear regimes below
- table updated to display detailed regime percentages
- equity chart background colored by regime (colorblind-friendly palette)
- trade density chart improved with aligned Y-axis zero levels
- UX improvements
- automatic scroll to charts when selecting a strategy
- better regime badges and labeling
- colorblind-friendly visualization
Result:
Step 3 now provides full strategy inspection including:
- OOS performance
- regime behaviour
- regime distribution per WF window
- visual regime overlays
Next step:
Implement strategy promotion / selection layer before Step 4 (parameter refinement).
dam
a42255d58c