Bills March Madness Prediction Model

What can I say? I'm a big fan of March Madness and math. This application is a demo that predicts NCAA men's basketball tournament outcomes using real team stats and historical results, then details what's behind it — ratings, metrics, statistics, predictions and how well past forecasts actually did against real results. I've tried to document how I built a real, data-driven prediction model and GUI end-to-end: a full pipeline from raw stats to a prediction fans like me can leverage when the brackets open in March. I've leaned heavily on transparency — you can view exactly how confident the model was and how accurate it actually was, a bit underwhelming! In the "offseason", I expect to tweak the model to see if I can realize better results next March.

Need more info? This is an auditable NCAA Division I men's basketball prediction system: with opponent-adjusted possession/efficiency ratings, expected scoring margin, calibrated win probability, and a full bracket forecast (deterministic and Monte Carlo simulation), rebuilt each season from versioned model fits rather than a single fixed snapshot. For those of you intrigued by that last sentence, take a look at the the Data Dictionary that outlines the metrics and modeling techniques used by this application.

This prediction model is a personal project built for demonstration and entertainment purposes only. Sports are unpredictable. Accuracy is not guaranteed. In addition, this is not financial or betting advice. I do not condone or encourage gambling nor can I be liable for any financial losses or decisions made using this data.

For more general information about this site, and the AI-aided demo development behind it, see the About This Site page.