Nelson Mendels ’27 presents research at the Carnegie Mellon Sports Analytics Conference

Nelson Mendels ‘27, a Quantitative Social Science (QSS) major, recently traveled to Pittsburgh, PA. to present his research at the Carnegie Mellon Sports Analytics Conference. Accompanied by his advisor, Professor Michael Herron, Nelson presented his poster, "Fighting and the Flow of Violence in the National Hockey League." See the abstract below.

The research, co-authored by Mendels, Professor Herron, and Alex Rockmore '27, began as a simple problem set question in Professor Herron’s Sports Analytics course (QSS 30.01). When Mendels was awarded a James O. Freedman Presidential Scholarship, the team was excited to expand on the topic, leading to their working paper, “Fighting and the Flow of Violence in the National Hockey League.” 

“What I loved about my time working with Professor Herron is that there was always a reason to chase more answers,” Mendels says. “The models we built came in many forms before we settled on one. It taught me to keep learning and questioning.” 

Mendels, who aspires to pursue a career in sports analytics, added, “This has become a hallmark project for me to demonstrate my skills, technique, and understanding of how to use statistical analysis to evaluate trends and patterns in sports.”

Professor Herron continues to serve as Mendels’s advisor for his QSS Honors Thesis, where he plans to explore the effect of discretionary mid-game goalie substitutions in professional ice hockey.

 

Abstract:
The National Hockey League (NHL) is unique among major professional sports leagues in North America in that in-game fighting is penalized but generally does not result in ejection or a substantial fine. Three theories may explain this: fighting entertains fans, it serves as a bona fide in-game strategy, or it functions as an accountability mechanism in an otherwise violent sport. The accountability theory implies that violence should decrease after a fight, and we explore this dynamic using time-stamped play-by-play data from NHL regular-season games from 2010–11 through 2025–26. We model the occurrence of violent events as a Poisson process and find that, in a majority of NHL games (60-70%), violent events are consistent with process homogeneity, implying that violence in these games evolves (or flows) without momentum. However, games with more fights are disproportionately likely to depart from homogeneity, and, among games with exactly one fight, the estimated rate of violent events decreases by close to four percent after the fight, controlling for observed game characteristics. These findings are consistent with the theory that fighting in NHL games can function as an accountability mechanism that reduces subsequent violence.