Coincidentally @the_number_game also published an expected points model this weekend in his most recent Forbes article, this time for the Premier League table. His is based on a measure called mSq£R which comes from the transfer price index mob. I thought it would be fun to look at my simple expected points model based on bookmakers odds for the Premier League and see how similar the results are to those in this alternative model.
This time I have calculated the residuals – a fancy term used by us statisticians to mean ‘difference between reality and the results from a model’ – as this was also how @the_number_game presented his work. We are therefore comparing actual points achieved by a club with what would be expected based on each of these models. I have taken the figures from the mSq£R model from a tweet placed yesterday which updated the original article to include all of Saturday’s matches except Manchester City v QPR.
Here is the comparison. I have included Manchester City v QPR in mine so there is a slight discrepancy there:
The figures shown therefore represent how much better or worse each Premier League team is performing than expected on the basis of a) what the bookmakers have expected and b) what a model using the inflation adjusted squad transfer cost would expect. As you can see, the results are broadly similar.
Currently we must use caution in attempting to conclude anything from this as only a maximum of three matches have been played by each Premier League club. We will return to look at this from time to time. The real challenge, however, will be to try to tease out how much of the apparent improvement/decline in the performances of the 20 Premier League clubs is due to skill or luck.