Applying N-of-1 GAMMs to Quantify Pitch Count Associations with Fastball Metrics and Biomechanics: A Comparative Modeling Study

Abstract

Workload monitoring remains a persistent challenge in baseball, particularly in collegiate pitchers exposed to repeated throwing demands across games, practices, bullpens, and travel. Although pitch count is one of the most widely used workload indicators in applied baseball settings, it does not fully capture the broader context in which performance and biomechanical responses occur, and traditional group-based approaches may obscure meaningful heterogeneity in athlete-specific workload-response patterns. Therefore, the overall objective of this dissertation was to apply individualized N-of-1 generalized additive mixed models to investigate rolling pitch count associations in collegiate baseball pitchers across performance, biomechanics, comparative modeling, and applied monitoring contexts. Across the 2023, 2024, and 2025 seasons, individualized fastball models demonstrated that rolling pitch count associations could be characterized at the pitcher-season level, with velocity providing the strongest and most practically interpretable signal, while spin rate and vertical break functioned as complementary outcomes. Across the 2024 and 2025 seasons, individualized biomechanics models further showed that workload-related response could also be expressed mechanically, although model performance varied across biomechanical outcomes and pitcher-seasons. Comparative analyses demonstrated that individualized N-of-1 models outperformed group-based models across both performance and biomechanics domains, supporting pitcher-specific modeling when the goal is to characterize within-athlete workload-response patterns. Finally, individualized velocity models were translated into an appearance-level CUSUM monitoring framework, showing that pitcher-specific expected-versus-observed surveillance could identify heterogeneous in-season deviation patterns. Collectively, these findings support individualized modeling as a more informative framework for evaluating workload-response patterns in collegiate baseball pitchers than reliance on population-average approaches alone.This research is comprised of four parts: • The first part aims to quantify within-pitcher associations between rolling pitch count and fastball performance metrics, including velocity, spin rate, and vertical break, in collegiate baseball pitchers. • The second part aims to quantify within-pitcher associations between rolling pitch count and key pitching biomechanics in collegiate baseball pitchers. • The third part aims to compare individualized N-of-1 models with traditional population-level mixed-effects models in explanatory performance and residual independence. • The fourth part aims to apply cumulative-sum (CUSUM) monitoring procedures to individualized model outputs to detect early, systematic declines in performance across a season.