← Projects

NASCAR Race Predictor

Sports ML

Road courses, short tracks and superspeedways race differently, so each track type gets its own four-model ensemble, which outputs calibrated win and top-N probabilities for every driver.

Screenshot of NASCAR Race Predictor: Interface illustration with sample probabilities: win and top-N chances for every driver from a per-track-type ensemble.
Interface illustration with sample probabilities: win and top-N chances for every driver from a per-track-type ensemble.

Highlights

The full story

Calibrated win and top-N probabilities for every driver in a race. Road courses, short tracks and superspeedways race differently, so each track type gets its own ensemble of XGBoost, LightGBM, random forest and gradient boosting, trained on 7,566 historical race records and compared live against odds from 9 sportsbooks.

Built with

XGBoost · LightGBM · scikit-learn · Sportradar API