Semiparametric scoring rate estimation in European soccer
Date
2021-06
Authors
Ritchie, Robyn
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis studies semiparametric scoring rate estimation in soccer (league-wide, season- and team-specific) under the assumption that the number of goals scored per team in a game follows a nonhomogeneous Poisson process model. Specifically, time varying scoring rates included in the model are decomposed into a scoring pattern, which smoothly changes over the course of the game, and an expected number of goals for each team, based on team abilities and a home field advantage. We explore specific questions related to team performance, including whether scoring patterns differ between home and away games. These methods can also be used to compare scoring rates between leagues and/or seasons and study scoring patterns throughout extra time. Expanding these ideas, we examine the patterns of other events such as player specific scoring patterns, conceded goals, and substitutions. These analyses are based on five seasons of soccer across four of the top European leagues where to-the-second game event times and details have been collected from game commentaries through web scraping, leading to data on over 20,000 goals throughout more than 500,000 comments.
Description
Keywords
Statistics, Kernel Density Estimation, Nonhomogeneous Poisson process, Homogeneous Poisson process, Bernstein polynomials, Soccer, Generalized Linear Models, Scoring Rate, Sports Analytics