Tournament seeding and in-game score differences as predictors of winning in NCAA women’s basketball

Tournament seeds summarize expert assessments of team quality and are commonly used to predict outcomes prior to competition. This study examines how the predictive content of such pre-game information evolves once games are underway and in-game score differentials are observed. Using data from the NCAA Women’s Division I Basketball Tournaments from 2018 to 2023, logistic regression models estimate game outcomes as a function of seed differences before tipoff and of both seed differences and score differentials at successive in-game checkpoints. The results show that tournament seeds remain statistically significant predictors of winning even after conditioning on in-game scores, although their influence declines as the game progresses, while score differentials become increasingly informative. These findings indicate that in-game scoring provides an imperfect signal of underlying team strength and that expert-based pre-game rankings retain incremental predictive content even late in games.