Kalshi ML Testing
Predicting market outcomes with supervised learning, calibrated probabilities, and rule-based trading.
The project
- Built supervised learning models to predict binary outcomes in prediction markets using historical price, volume, and temporal features
- Applied isotonic regression for probability calibration, improving alignment between predicted probabilities and observed outcome frequencies for trading applications
- Designed and backtested a rule-based trading strategy using probability edges, achieving consistent positive expected value
Figures
