Explainable Defense Coverage Classification in NFL Games using Deep Neural Networks

Research Paper will be posted in the coming weeks. Check back soon!
Download the
Full Paper Here

Huan Song, Mohamad Al Jazaery, Haibo Ding, Lin Lee Cheong, Kyeong Hoon (Jonathan) Jung, Mike Band


Coverage scheme is at the core of understanding and analyzing any football defensive strategies. However, manual identification of the coverages is time-consuming and laborious. This paper presents a deep neural network model that classifies the coverages automatically and accurately. In addition, we tackle the opaqueness of the deep learning model through comprehensive model explanations using play embedding analysis and gradient-based approaches. The model explanations provide confidence that the model aligns with human experts' understanding, help speed up visual review processes, and bring additional insights about defense coverage schemes.