Tools
Tools
Collins, Donal (2024) Identification of Inefficiencies in Football Betting Markets using a Statistical Approach and Artificial Intelligence Techniques. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
Football or soccer matches have long been the focus of analysis, prediction and gambling. The introduction betting exchanges has transformed betting markets into liquid and responsive markets which share characteristics with financial markets. Focusing on the English Premier League, this study develops a methodology to estimate strength or ”Form” of each team prior to each match. Using these Form statistics, a measure of home advantage and implied probabilities of exchange market odds, three models have been developed to predict to outcomes of matches using a modified Poisson distribution method, Extreme Gradient Boosting and Multi-layer Perceptron machine learning methods. To test the performance of the models a simulated betting strategy was employed where model probabilities were compared to Betfair Exchange market odds of bets transacted immediately before match start times. The Extreme Gradient Boosting model generated a 17.2% return on the held-out dataset of the English Premier League 2023-24 season. All three models generated returns of greater than 10% on average when evaluated on held out English Premier League matches. To further demonstrate efficacy, the models were tested on German Bundesliga and Spanish La Liga matches where the approaches were also profitable on average, though, at a lower level. The study demonstrates that football betting markets have inefficiencies which can be identified using the methodology to measure team strength and the machine learning models described.
Actions (login required)
 |
View Item |