Although basketball is a dualistic sport, with all players competing on bothoffense and defense, almost all of the sport's conventional metrics aredesigned to summarize offensive play. As a result, player valuations arelargely based on offensive performances and to a much lesser degree ondefensive ones. Steals, blocks and defensive rebounds provide only a limitedsummary of defensive effectiveness, yet they persist because they summarizesalient events that are easy to observe. Due to the inefficacy of traditionaldefensive statistics, the state of the art in defensive analytics remainsqualitative, based on expert intuition and analysis that can be prone to humanbiases and imprecision. Fortunately, emerging optical player tracking systemshave the potential to enable a richer quantitative characterization ofbasketball performance, particularly defensive performance. Unfortunately, dueto computational and methodological complexities, that potential remains unmet.This paper attempts to fill this void, combining spatial and spatio-temporalprocesses, matrix factorization techniques and hierarchical regression modelswith player tracking data to advance the state of defensive analytics in theNBA. Our approach detects, characterizes and quantifies multiple aspects ofdefensive play in basketball, supporting some common understandings ofdefensive effectiveness, challenging others and opening up many new insightsinto the defensive elements of basketball.
展开▼
