A Multiresolution Stochastic Process Model for Predicting Basketball Possession Outcomes

Abstract

Basketball games evolve continuously in space and time as players constantly interact with their teammates, the opposing team, and the ball. However, current analyses of basketball outcomes rely on discretized summaries of the game that reduce such interactions to tallies of points, assists, and similar events. In this paper, we propose a framework for using optical player tracking data to estimate, in real time, the expected number of points obtained by the end of a possession. This quantity, called expected possession value (EPV), derives from a stochastic process model for the evolution of a basketball possession. We model this process at multiple levels of resolution, differentiating between continuous, infinitesimal movements of players, and discrete events such as shot attempts and turnovers. Transition kernels are estimated using hierarchical spatiotemporal models that share information across players while remaining computationally tractable on very large data sets. In addition to estimating EPV, these models reveal novel insights on players' decision-making tendencies as a function of their spatial strategy. A data sample and R code for further exploration of our model/results are available in the repository https://github.com/dcervone/EPVDemo.

The expected value of possession in professional rugby league match-play

“…While the expected value of possession has been investigated in team sports such as American football (Carter & Machol, 1971;Romer, 2006), Australian football (O'Shaughnessy, 2006) and basketball (Cervone et al, 2014), there is currently little understanding of the situational value of possession in rugby league. This type of information has important applications for informing playing strategies and directing training programmes, assessing in-game decisionmaking and for rating individual and team performances.…”

“…This may be due in part to difficulties in capturing specific technical and tactical actions and attributing these often joint actions to an individual player in a complex team sport (Gerrard, 2007). In addition, analysis of complex chains of possession typically requires sophisticated analytic techniques, some of which may be difficult to interpret (Cervone, D'Amour, Bornn, & Goldsberry, 2014;Shafizadeh, Sproule, & Gray, 2013). Nonetheless, in recent years a number of commercial statistics services have emerged and now provide detailed data on player actions and movements for many sports including rugby league.…”

Modeling offensive player movement in professional basketball

“…A simpler point of view on the same problem is that one needs to generate a new sensible spatial coordinate given the most recent spatial coordinates we observed. Such a model for player movement is explained in the aforementioned paper on estimating expected possession value (Cervone et al, 2016). The paper's summarized research output is a framework that uses the SportVU data to perform estimation of the expected number of points obtained by the end of a possession, by way of a stochastic process model that models the evolution of a basketball possession.…”

“…One way to visualize the empirical η s is to show the average acceleration vector at each cell of the court -see Figure 4 in Cervone et al (2016) for the type of image that we seek to produce. The idea is: for each cell v = 1, .…”

“…Are we able to simulate player movement? Such questions also have downstream implications; for example, recent models for estimating instantaneous possession value rely on an underlying player movement model (Cervone et al, 2016).…”