Many, many years ago a friend of mine gave me the basis for a soccer prediction model. The paper was called “One Match to Go” and was published in December 2009. The paper focused on trying to predict the score and results for the end of season in the Premiership using expected goals and poisson distributions. I had been sitting on this paper for years, I had flitted with the idea of applying the model but in the last few weeks I finally applied it.
I used Excel as the basis, but Martin Eastwood has the bulk of the R code published if you wanted the model automated.
I used data from Football-Data.co.uk and based the model on the final table from 2016/17 year for the English Premiership. The results for the home games prediction were encouraging.

Home Draw Away
Actual 187 84 109
Predicted 167 13 32
Success 89% 15% 29%
Correct Scores 39 4 6
21% 5% 6%

The next stage is to look at the value the model highlighted versus the price was trading in the market. If you can find something that is mis-priced relative to your assessment of value you should always buy it.

The following table shows the probability of success for the home games that were successful. Generally when the model say a home win had a probability of 60% you should back it.

Leave a Reply

Your email address will not be published. Required fields are marked *