AFL football: how much is skill and how much is chance?

It is widely accepted that luck has some influence on all sporting events, whether it be an injury at a crucial time, an unfortunate bounce of the ball or a controversial decision by an official. But just how much does the result of a match come down to bad or good luck i.e. the random events within a game? A Markov process model with 18 states has been developed to approximate AFL football using match statistics. The models have been applied to matches from the 2003 and 2004 AFL seasons and displayed very encouraging results with minimal errors. Using transition probabilities derived from these matches, simulated goal distributions and probabilities of victory have been calculated. Using these distributions and the standard errors for each team’s match scores, the randomness that occurs in AFL matches is identified. Some of these matches have been investigated with the results presented in this paper.