Backtesting & ML in Finance

Read these first

1. 01Information-Driven Bars

   A five-minute bar samples the same amount of clock time whether the market is dead or on fire. Information-driven bars sample the same amount of activity instead, and that single change fixes a surprising number of backtesting problems.

Then the rest

1. 02Backtest Design
2. 03Look-Ahead Bias
3. 04Survivorship Bias
4. 05Data-Snooping Bias
5. 06Overfitting
6. 07Walk-Forward Analysis
7. 08Cross-Validation for Time Series
8. 09Purged & Embargoed Cross-Validation
9. 10The Deflated Sharpe Ratio
10. 11Feature Engineering for Finance
11. 12Triple-Barrier Labeling
12. 13Tree Ensembles in Finance
13. 14Backtest Overfitting
14. 15Class Imbalance in Financial Labels
15. 16Clustering for Portfolios
16. 17Concept Drift in Financial Models
17. 18Data Leakage in Machine Learning
18. 19Dimensionality Reduction in Finance
19. 20Ensemble Methods
20. 21Feature Neutralization
21. 22Hyperparameter Tuning
22. 23Label Horizon Selection
23. 24Model Retraining Cadence
24. 25Combinatorial Purged Cross-Validation
25. 26Feature Importance with SHAP
26. 27Fractional Differentiation
27. 28Gradient Boosting in Finance
28. 29Meta-Labeling
29. 30Minimum Backtest Length
30. 31Nested Cross-Validation
31. 32Neural Networks in Finance
32. 33Reinforcement Learning for Trading
33. 34Sample Weights & Label Uniqueness
34. 35The Probabilistic Sharpe Ratio