← Projects

NBA Game Predictor

Sports ML

Moneyline picks hit 66.3% across 1,064 settled games, and Grade-A over/under picks returned +16.5% ROI on 95 bets. The pipeline runs itself daily on GitHub Actions.

Screenshot of NBA Game Predictor: Interface illustration with real model output: one night’s slate with predicted scores, win probabilities and grades.
Interface illustration with real model output: one night’s slate with predicted scores, win probabilities and grades.

Highlights

The full story

A fully automated NBA pipeline that runs daily on GitHub Actions. Pre-game team stats and live sportsbook odds go in; two XGBoost regressors predict home and away points, which become moneyline, over/under and spread picks graded against final scores. Over 1,064 settled games: 66.3% moneyline accuracy, and Grade-A over/under picks returned +16.5% ROI across 95 bets.

Built with

XGBoost · Sportradar API · PostgreSQL · GitHub Actions