Document Type

Thesis

Degree Name

Doctor of Philosophy (PhD)

Department

Mathematics

Faculty/School

Faculty of Science

First Advisor

Zilin Wang

Advisor Role

Supervisor

Second Advisor

Mary Thompson

Advisor Role

Supervisor

Abstract

As soccer is widely regarded as the most popular sport in the world there is high interest in methods of improving team performances. There are many ways teams and individual athletes can influence their own performances during competition. This thesis focuses on developing statistical methodologies for improving competition-based decision-making for soccer so as to allow professional soccer teams to make better informed decisions regarding player selection and in-game decision-making.

To properly capture the dynamic actions of professional soccer, Markov chains with increasing complexity are proposed. These models allow for the inclusion of potential changes in the process caused by goals and substitutions, thus leading to a more complete and informative picture. Computer tracking data containing event descriptions and locations from La Liga games involving Futbol Club Barcelona have been used for model implementation and validation. Validations conducted through simulations show that the more complex models fit the data well. The most complex models including goal differential and substitutions are used to develop conditional substitution strategies. This is done through simulations and applications of general game theory.

Drafting is a common way for many North American professional sports teams to obtain new players. With no exception, the Major League Soccer (MLS) SuperDraft takes place prior to the start of each season to select potential new players. Being able to make well informed decisions surrounding draft selections is an important aspect of managing a team. The second major portion of the thesis seeks to identify desirable characteristics of players drafted by MLS teams. Modelling the probability of playing at least 30 MLS games, mixed effects logistic regression models are used to identify desirable characteristics and attempt to predict the outcomes of future drafted players. Results show significant player characteristics as well as multiple significant sources of variability. Furthermore, predictions are made for players who were selected in the 2018 and 2019 MLS SuperDrafts. Additional analysis is conducted on a series of rule changes made by Major League Soccer during the period of 2006 – 2008. These changes include: creating Designated Player roster spots, increasing the number of international roster spots, and allowing clubs to sign players from their academies. This work examines the impact of those rule changes on the MLS careers of players selected in the MLS SuperDraft, while taking into account that the careers of players selected before the changes will have been observed for longer than the careers of players selected after the changes. Survival analysis is used to combat the potential censoring of the careers of more recently drafted players. Statistical methods to compare two groups and logistic regression are used to examine possible impacts. Results suggest that there is a significant decrease in the probability of players selected after the rule changes going on to play competitive games in the MLS, potentially impacting how clubs choose to allocate resources for player recruitment.

Convocation Year

2023

Convocation Season

Spring

Download

Share

COinS