| Article ID: | iaor20051442 |
| Country: | United Kingdom |
| Volume: | 52 |
| Issue: | 4 |
| Start Page Number: | 591 |
| End Page Number: | 602 |
| Publication Date: | Oct 2003 |
| Journal: | The Statistician |
| Authors: | Wright Mike, Hirotsu Nobuyoshi |
| Keywords: | markov processes, statistics: inference |
The paper proposes a statistical model of a football match that is useful in providing insights into the characteristics of teams. We evaluate the characteristics by means of maximum likelihood estimators for factors such as home advantage, offensive and defensive strength and their interactions, by taking account of not only goals but also possession of the ball. We build the model on the basis of the real data from the 1999–2000 season of the English Premier League and illustrate the characteristics of the teams in ways that allow us to visualize offensive and defensive strength, preference for playing at home and relative success against particular opposition teams.