ML² is an advanced regressor with one core tenet: to predict NBA money lines with machine learning. Metrics used as reliable indicators of performance include previous general player performances and against the selected opponent, variance in previous performances, and general statistics like points per game. Our predictive model has approximately achieved an impressive 60% accuracy on players’ “unders” with 58% of predictions being within 3 points of the actual performance; a mean absolute error of 4.07 on validation data.

Our model uses a tree-based regressor built and tuned using XGBoost to predict an NBA player’s points in their upcoming games. Data was extracted using nba_api and aggregated using pandas. The front end currently exhibited is written using React.js, HTML and CSS whilst our model is baked into an API using Django and hosted using AWS EC2 and Route 53, Gunicorn, NGINX .