In this article we are going to look at how the outcome of football games can be predicted using statistics. Before you run away, you should realise that statistics were developed in the first place to make sense of gambling. The early statisticians were avid gamblers.
The approach that was used is Bayesian. This allows us to use any available information about a forthcoming game with a prior distribution. This information can be historical, the current team fitness, weather conditions etc. We then make predictions using a posterior distribution of future scores and outcomes including team performance measures.
The distribution used is the Skellam distribution of goal differences which involves the number of games played and the goal scores of the home and away teams. When we want to predict the goal difference in a future game we use the posterior distribution.
Without going into the mathematics (check the web for that) let’s see how it works in practice. The approach was used to analyse a previous Premiership season using net attacking parameters. The agreement between predictions and reality was remarkably close. It predicted a goal difference of 56 for Manchester United and a point’s score of 87. The actual goal difference was 56, and the points score 89. Many other goal differences were as good or nearly as good, and the final league table was very similar to the predicted one.
This is a remarkable development that could be of huge benefit to sports betters. Unfortunately it is of little use in online casinos where the games are subject to the laws of chance.
However it does have applications in games such a poker. Statistics were developed in order to understand casino games, now they are being developed into predictive models for sports betting – don’t you wish that you had paid more attention to your maths and stats lessons at school?