Predicting the English Premier League

Using historical wage data to predict a league position

James Clarke
Software Engineer

Footballers Wages

  • Footballers earn a LOT of money. The average is something like £31,000 per week, with top earners more like £250,000 per week.
  • Somewhat unsurprisingly it turns out that the more a club pays its players the higher in the league it tends to finish.
  • I use historical data link here of wage bills to build a prediction algorithm for the likely league position.
  • Since the clubs are all in competition with each other it makes sense to use the relative spend on wages and not the absolue value.

Plot of wage bill against league finish

plot of chunk unnamed-chunk-1 

## Loading required package: lattice

Accuracy

The mean absolute error has been estimated by training the model on 75% of the data and testing it against the remaining 25% out of sample data. The average number of places that the prediction algorithm was out by is:

## [1] 3.310066

The app also indicates the 95% confidence interval for the prediction. If one could find offered betting odds better than 95% then taking these bets should make money in the long run.

For context the historical highest and lowest placed finishes for a club with that spending rank are also shown.

Links