← All projects
Weekly Trading Strategy
Predicts weekly sector outperformance from 11 engineered features, and honestly reports where a RandomForest fell short.
- Python
- Pandas
- NumPy
- SQLite
The problem
Sector rotation is the bet that capital moves between parts of the market in ways that are at least partly predictable a week ahead — that energy leading one week and utilities the next is signal rather than noise. The question this project asked was narrow and testable: given features available on Friday, can a model do better than a naive baseline at ranking sectors for the week to come?
Approach
Eleven engineered features per sector, a RandomForest classifier for training, and SQLite as the local store for price history and generated features.
Result
The metrics came out mediocre.