Modeling Quarterback Dynamics

This chapter examines the intricacies of training a model to analyze targeted passing plays in football, featuring a dataset of 36,000 plays and innovative data augmentation techniques focusing on quarterback handedness. The discussion uncovers the complexities of measuring quarterback performance through deep learning, encompassing challenges in explainability and the unexpected similarities in decision-making between rookies and seasoned players. Future directions for enhancing the model include potential applications of LSTM models for deeper insights into play analysis and enhancing user interaction with the predictions.