## Using comparable performance to predict the Premiership standings

In each of the past two seasons I’ve tried to use past performance to predict the Premiership standings. In each case I’ve used a ‘dumb’ system – i.e. all I do is run the numbers, not add knowledge to the system. The basis of those systems was the ratio of shots and shots on target that a team achieved in it’s games the season prior.

This time around I’ll run both of those systems again but I also want to try something new. I’ve no idea how well it’ll perform but I think it’ll help to provide some context.

Given that I’m bound to get some criticism from fans who think their team should be ranked much higher it’s important to stress that, outside of the concept, I’ve had a minimal amount of input here. If someone else were to run the analysis the results would be the same. It should also be mentioned that neither this system, or either of the other two ‘dumb’ systems I’ll be using to predict this years standings, take any player movement into account. They aren’t perfect, nor are they designed to be – I’m simply trying to pull something useful out of freely available information.

The methodology is as follows

Determine the total shot ratio of a given team in the prior season*
Look back historically and find the ten teams who’s performance is the closest match for said given team
See how well these ten ‘comparable’ teams performed in the following season

*An explanation of total shots ratio can be found here and here.

For Premiership teams I have 220 data points to pick from, for those promoted to the Championship the sample is somewhat smaller, at 33.

So lets take a team to demonstrate how this works. Sorry Blackpool fans but I chose you. Anyway this is the report for Blackpool last season:

Now it turns out that Blackpool were actually pretty bad for a team that nearly got promoted so in their case I’ve actually chosen only nine comparables, otherwise the TSR of the comparables would appreciably exceed that of the team we’re studying. The nine chosen teams and their ‘year 0’ TSR is summarised below:

So on average we’ve hit the nail on the head in terms of finding teams that a) performed in a comparable manner to Blackpool and b) expanding our sample size of expected Premiership performance to >300 games.

So how did these teams perform in ‘year 1’ (in the Premiership)?

Instead of reporting this table each time I’m going to summarise it as follows:

So from this table we would expect Blackpool to finish 20th in the league, scoring 38 points. The most points recorded by one of their comparables was 65 by Ipswich, with the least being 11 by Derby County. The positions in the ‘max’ and ‘min’ columns are where the given team would finish if they recorded the ‘max’ or ‘min’ number of points and everyone else in the league was to score the average of their respective comparables.

This isn’t a perfect system but I think it’s my favourite. Now for the fun of getting blasted when I project teams to finish lower than their fans expect…