All about momentum: Investigating high-pressure situations in the NBA through scoring probability

One of the defining characteristics of real basketball stars, and even great role players, is how well they perform under immense mental pressure. This paper presents a method to identify high-pressure situations during a basketball game through shooting success. Our analysis incorporates a novel feature set that emphasizes both player- and team-level momentum, including scoring streaks. Additionally, we redefine player roles using a clustering-based approach with deep learning techniques, allowing for a more nuanced evaluation of performance under pressure. Using six seasons of NBA data, we find that shotmaking is mainly impacted by the so-called momentum , i.e., when a team outscores their opponent significantly over a short period of time.