The landscape of American college football has changed dramatically in recent years with conference realignment, increased usage of the transfer portal, and the seismic legal ruling regarding athletes' name, image, and likeness (NIL). In this work, we use a Bayesian state-space model to capture the impact of transfers and NIL on the volatility of team strengths over time. Specifically, we extend the classic autoregressive process for team strength by modeling the between-season innovation variance as a function of incoming transfers and NIL rule changes. This approach enables us to predict greater variance for team strengths in an upcoming season based on roster changes. Our results from the playoff era (2014-2025) reveal variation between schools that have embraced the volatility of transfers (e.g., Deion Sanders at Colorado) with those who have been reluctant to change (e.g., Iowa and Clemson). We demonstrate an improvement in forecasting the distribution of wins by teams entering the season with our novel approach, indicating the arrival of a new era of uncertainty in college football.

Ron Yurko is an Assistant Teaching Professor in the Department of Statistics & Data Science at Carnegie Mellon University and the Director of the Carnegie Mellon Sports Analytics Center (CMSAC). His research focuses on developing statistical methods at the interface of inference and machine learning, with applications in sports analytics and natural language processing. Under his leadership, CMSAC drives high-impact industry partnerships—including recent collaborations with the NFL Draft and partnering with the Heisman to power data insights for the Heisman Hopefuls platform. His work and commentary have been featured in major outlets such as The Athletic, ESPN, CBS, and NBC. He has industry experience in both finance and professional sports, and is a three-time degree holder from Carnegie Mellon: with a bachelors, masters, and PhD in Statistics.