High School to NCAA: Predicting Freshman Impact in College Basketball with Machine Learning

The transfer portal, elimination of scholarship limits, and policy changes to NIL (Name, Image, and Likeness), and revenue-sharing have required college coaching staffs to shift their operations akin to a professional sports team: balancing performance optimization with budget and resource constraints. Inspired by NBA-level analytics, we developed a machine learning model to predict a freshman’s contribution at the Power 5 level. (teams at Tennessee’s level). We compiled four years of game-level statistics from high school summer leagues and college basketball and calculated over 60 advanced metrics, culminating in Wins Above Replacement Player (WARP). WARP is a comprehensive measure of a player’s impact related to a “replacement-level player”. We trained a machine learning model to estimate a player’s freshman-year WARP from their high school summer statistics. These predictions were merged with recruiting rankings and information from recruiting industry leader 247Sports and deployed on a roster-management platform for coaches. This predictive model serves as a valuable tool for the University of Tennessee’s college basketball coaches to analytically evaluate high school players, make critical decisions in the roster management process, and navigate a chaotic and constantly evolving recruiting and performance environment.