Data-Driven Ghosting using Deep Imitation Learning

Current state-of-the-art sports metrics such as "Wins-above-Replacement" in baseball, "Expected Point Value" in basketball, and "Expected Goal Value" in soccer and hockey are now commonplace in performance analysis. These measures have enhanced our ability to compare and value performance in sport. But they are inherently limited because they are tied to a discrete outcome of a specific event. With the widespread (and growing) availability of player and ball tracking data comes the potential to quantitatively analyze and compare fine-grain movement patterns. An excellent example of this was the "ghosting" system developed by the Toronto Raptors to analyze player decision-making in STATS SportVU tracking data. Specifically, the Raptors created software to predict what a defensive player should have done instead of what they actually did. Motivated by the original "ghosting" work, we showcase an automatic "data-driven ghosting" method using advanced machine learning methodologies called "deep imitation learning", applied to a season's worth of tracking data from a recent professional league in soccer. Our ghosting method, which avoids substantial manual human annotation, results in a data-driven system that allows us to answer the question "how should this player or team have played in a given game situation compare to the league average?". In addition, by "fine-tuning" our league average model to the tracking data from a particular team, our ghosting technique can estimate how each team might have approached the situation. Our method enables counterfactual analysis of effectiveness of defensive positioning as both a measurable and viewable quantity for the first time.