Deriving a profit from Australian rules football: a statistical approach

Using match information gathered from 100 seasons of Australian Rules Football played prior to 1997, multiple linear regression model was used to identify and weight numerical features that could independently explain statistically significant proportions of variation associated with the outcome of matches. Prediction models constructed at both a team and player level were applied to matches played between 1997 and 2003 with results compared against an existing benchmark for AFL prediction and bookmakers’ prices. Although bookmakers have appeared to improve in their price setting processes over the past seven years, it is still possible to derive an annual profit, with statistically significant improvement coming through the use of data derived at an individual player level