Machine learning · Research
Machine Learning Forecasting Research
Rigorous ML models for financial markets, validated the way production demands
— Model- - Random
Fold 1
Fold 2
Fold 3
Fold 4
The challenge
Most trading backtests look great and fail live because of look-ahead bias, unrealistic costs and cherry-picked settings.
The solution
Point-in-time universes, decisions only on closed bars with execution at the next open, realistic fees and funding, quarterly walk-forward retraining and null tests with random portfolios to prove that results aren't luck.
What it does
- Feature engineering and ablation studies
- LightGBM cross-sectional ranking models
- Walk-forward validation across multiple years
- Realistic transaction costs and funding
- Null tests against random portfolios
- Interactive HTML research reports
Next project
Lush Rituals
Have a project in mind?
Tell me what you need. I reply within one business day with ideas and a ballpark quote.