2df1ea92e1
BREAKING CHANGES: - Removed orphaned dqn.rs monolithic trainer (4,975 lines) - Removed orphaned dqn_ensemble.rs module (816 lines) - Removed orphaned tft.rs and tft_complete_int8_integration_test.rs - TFT trainer split into modular directory structure DQN Module Refactoring: - Split trainers/dqn.rs into modular structure (config.rs, statistics.rs, trainer.rs) - Fixed hyperopt 39D search space (continuous params only) - Boolean flags (use_dueling, use_double_dqn, use_per, use_noisy_nets) are now FIXED architectural decisions - use_distributional defaults to false (Candle BUG #36 - scatter_add gradient issues) Clean Module Structure: - ml/src/trainers/dqn/ directory with proper mod.rs exports - ml/src/trainers/tft/ directory with config.rs, types.rs, model.rs, trainer.rs, tests.rs - All P0 features validated: TD-error clamping, batch diversity, LR scheduler, priority staleness Documentation: - Added comprehensive docs in docs/codebase-cleanup/ - ADR-001 for DQN refactoring decisions - Rainbow DQN component matrix and quick reference guides Build Status: Compiles with zero errors 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
1.8 KiB
1.8 KiB
DQN Backtesting Report
Model: DQN-Weak-Example Baseline: DQN-Trial35-Baseline Generated: 2025-11-04 07:56:33 UTC
Performance Summary
| Metric | Value | Target | Status |
|---|---|---|---|
| Total Return | -3.20% | >0% | ❌ |
| Sharpe Ratio | 0.60 | >1.5 | ❌ |
| Max Drawdown | 35.40% | <20% | ❌ |
| Win Rate | 38.0% | >50% | ❌ |
| Alpha vs B&H | -2.10% | >0% | ❌ |
Comparison to Baseline
| Metric | Baseline | New Model | Change | Direction |
|---|---|---|---|---|
| Returns | 12.10% | -3.20% | -15.30% | ↘️ |
| Sharpe | 1.80 | 0.60 | -1.20 | ↘️ |
| Drawdown | 18.30% | 35.40% | +17.10% | ↘️ |
| Win Rate | 52.0% | 38.0% | -14.0% | ↘️ |
| Alpha | 1.50% | -2.10% | -3.60% | ↘️ |
Trade Statistics
| Metric | Value |
|---|---|
| Total Trades | 110 |
| Avg Trade Return | -0.029% |
| Win Rate | 38.00% |
| Trades vs Baseline | -28 |
Deployment Recommendation
Status: ❌ REJECT - Not Production Ready
Model only passes 0/5 production criteria. Performance is insufficient for production deployment.
Critical Issues:
- ❌ Negative total return (-3.20%)
- ❌ Low Sharpe ratio (0.60 < 1.5)
- ❌ Excessive drawdown (35.40% > 20%)
- ❌ Poor win rate (38.0% < 50%)
- ❌ Negative alpha (-2.10%)
Action: Do not deploy. Retrain model with improved hyperparameters or different architecture.
Production Criteria Checklist
- Criteria Passed: 0/5
- Total Return: ❌ FAIL (-3.20% > 0%)
- Sharpe Ratio: ❌ FAIL (0.60 > 1.5)
- Max Drawdown: ❌ FAIL (35.40% < 20%)
- Win Rate: ❌ FAIL (38.0% > 50%)
- Alpha vs B&H: ❌ FAIL (-2.10% > 0%)
Report generated automatically by Foxhunt ML Evaluation Framework