Evimaria Terzi and PhD candidate Iasonas Nikolaou prepare to present new paper “Understanding Team Collapse via Probabilistic Graphical Models” at ACM SIGKDD 2025
Professor Evimaria Terzi of Boston University, PhD candidate Iasonas Nikolaou, and University of Pittsburgh’s Konstantinos Pelechrinis have developed a groundbreaking computational model that delves into the phenomenon of “team collapse” in professional basketball. Their recent paper, titled Understanding Team Collapse via Probabilistic Graphical Models, has been accepted for presentation at the ACM SIGKDD 2025 conference in Toronto this August.
Traditional analyses of team performance often focus on individual statistics and game outcomes. However, this study introduces a novel approach by employing probabilistic graphical models to examine how individual actions, such as a missed shot or a turnover, can influence the behavior and decisions of teammates, potentially leading to a collective decline in performance. The model captures the intricate dynamics of team interactions, providing a computational perspective on how momentum shifts and breakdowns in coordination can dramatically alter the outcome of a game.
A key insight from the research is the identification of “pillar” players—individuals whose performance and behavior are crucial in maintaining team stability and momentum. These players act as anchors, and their actions can significantly impact the overall team’s dynamics. Understanding the roles of these pivotal players can provide insights into strategies for enhancing team resilience and preventing performance breakdowns. While the study focuses on basketball, the implications extend to various domains where teamwork is essential.