model lane · 2025
Football Performance and Admissions
A four-person team project linking NCAA football records with College Scorecard admissions data to test the "Flutie effect," using XGBoost and SHAP.
- data (covered)
- model (covered)
- eval (covered)
- serve (not covered)
- monitor (not covered)
1st place out of 4 teams in the 2025 BYU Statistics Department Case Study Competition.
stackPythonXGBoostSHAPscikit-learnpandas
linksAs part of a four-person team, I built pipelines (Python, pandas, scikit-learn, XGBoost) linking NCAA football records with College Scorecard admissions data to test the “Flutie effect,” which is the idea that a winning football season brings in more applications.
We implemented regression models, clustering, and engineered features like lagged win rates and tournament runs to evaluate how athletic success relates to application volumes. We also used SHAP values to show which factors were driving the predictions.
We delivered interactive visualizations and a project site, and presented our findings to faculty judges.