Author: Rory Paul Bunker
Bunker, Rory Paul, 2025 Machine Learning and Data Mining for Sports Performance Analytics: Frameworks & Applications for Match Outcome Prediction and Pattern Identification, Flinders University, College of Science and Engineering
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
Identifying the factors contributing to successful outcomes at different levels is vital for decision-makers in professional sports, which historically have relied on subjective opinion. Technological developments have resulted in rapid growth in data generated in sports, which, with appropriate techniques, ostensibly allows for the objective identification of factors contributing to success.
This thesis presents original contributions to the emerging field of sports performance analytics through six interconnected publications addressing three primary research questions. The first question investigates how conceptual frameworks can be developed for machine learning for match outcome prediction and sports performance analytics as a whole. The second question explores how data mining methods from other disciplines can be effectively leveraged for key pattern identification in sports. The third question focuses on applying machine learning methods from various domains to provide interpretable match outcome predictions in sports.
The thesis highlights the benefits of contemporary machine learning and data mining approaches used successfully in other domains and contexts for match outcome prediction and key pattern identification. In this thesis, relevant literature in machine learning for sports match outcome prediction is critically analysed and synthesised, and a conceptual framework for applying machine learning in sports match result prediction is proposed. This framework is then demonstrated in practice in a subsequent publication in the context of tennis match result prediction. More specific explorative surveys into machine learning for match outcome prediction were carried out in the context of team sports in general and then soccer specifically.
Sequential pattern mining-based classifiers are used to identify interpretable key event patterns from passages of play that discriminate between scoring and non-scoring outcomes, and interpretable rules-based machine learning is used to identify key patterns composed of performance indicator combinations and values that discriminate between successful and unsuccessful tournament stage outcomes. This thesis takes a multi-level approach, investigating performance at various levels of analysis, including the match outcome, tournament stage, and passages of play levels across various sports.
Keywords: sports analytics, sports performance analysis, match outcome prediction, match result prediction, machine learning, data mining, sequential pattern mining
Subject: Engineering thesis
Thesis type: Doctor of Philosophy
Completed: 2025
School: College of Science and Engineering
Supervisor: Matthew Stephenson