Name
Session 2, Speaker 4: Elizabeth Upton, Assistant Professor of Mathematics and Statistics, Williams College
Date & Time
Saturday, April 12, 2025, 4:30 PM - 4:50 PM
Description
In team sports, standard ranking metrics do not consider how smaller groups of players interact. Individual performance measures rarely reflect the synergy or friction among pairs or trios, while full-lineup assessments can’t pinpoint the exact contributions of these lower-order combinations. To bridge this gap, we propose a novel adjusted plus-minus (APM) framework that simultaneously evaluates individuals, smaller groups, and entire lineups. Underlying our approach is a link between APM and the hypergraph representation of a team, which captures these overlapping interactions. In this talk, we’ll demonstrate how this perspective applies to NBA data from 2012–2022 and highlight the insights gained from viewing a team as a network.
Location Name
1401
Full Address
Kline Tower
219 Prospect St
13th and 14th Floors, Registration Table in Room 1401
New Haven, CT 06511
United States
219 Prospect St
13th and 14th Floors, Registration Table in Room 1401
New Haven, CT 06511
United States
Session Type
Lecture
Title
Hypergraphs And Hoops: Uncovering Nba Player Interactions Through Apm
Abstract
In team sports, standard ranking metrics do not consider how smaller groups of players interact. Individual performance measures rarely reflect the synergy or friction among pairs or trios, while full-lineup assessments can’t pinpoint the exact contributions of these lower-order combinations. To bridge this gap, we propose a novel adjusted plus-minus (APM) framework that simultaneously evaluates individuals, smaller groups, and entire lineups. Underlying our approach is a link between APM and the hypergraph representation of a team, which captures these overlapping interactions. In this talk, we’ll demonstrate how this perspective applies to NBA data from 2012–2022 and highlight the insights gained from viewing a team as a network.
Speaker Bio
Elizabeth Upton is an Assistant Professor in the Mathematics and Statistics Department at Williams College, where she has been since completing her Ph.D. in Statistics at Boston University in 2019. Her research focuses on applied statistics, with a particular interest in regression models for network data. Before graduate school, she taught high school math and worked in finance analyzing currency markets. Liz loves teaching statistics, collaborating with researchers across disciplines, and talking sports data whenever she gets the chance. Outside of work, she’s usually chasing around her three kids or cheering on the New England Patriots (#54 forever).
Speaker Headshot
