Three expected points models for the Premier League compared

I have mentioned my naïve base model in tweets and articles over the past week and as @the_number_game has again published his expected points model again at Forbes (see explanation of where that comes from in my earlier article here) it is worth placing my two models next to his to see how we are progressing.

Firstly though an explanation of a naïve base model. It is a pretty simple concept. We just make the assumption that a club takes exactly the same points from its opponents as last season and thus ends with the same points in total at the end of the season. In order to deal with the fact that three Premier League clubs from last season were relegated, we replace them with the three promoted clubs as follows; Reading are assumed to have Bolton’s results, Southampton Blackburn’s and West Ham Wolves’. In other words the best promoted club replaces the best relegated etc. This enables us to track these three clubs by comparing them directly to the relegation candidates last season as avoiding relegation must be their first goal.

The reason for using a naïve base model is as a check on my expected points model. If a club is doing better (or worse) in my expected points model and is also taking more (or less) points from its exact same fixtures as last season, we can be more sure of the result of the expected points model which, at this stage, can be affected by a couple of surprise results. It is also an interesting way to look at a club’s start to the season and, like many simple concepts, it is absolutely amazing to me that no one does these sorts of comparisons.

Anyway, the residuals of the three models are consistent for Chelsea who have taken a remarkable eight points more from their six matches to date in comparison to last season’s exact same fixtures, a 100% improvement. West Bromwich Albion have taken four more points than last season from their six opponents to date, a total which is mirrored in both my expected points model and Zach’s MSq£ model. Everton are an interesting anomaly which I covered in my previous article. Newcastle United are too but fans of that club have told me this week that their lagging performance in the naïve base model is due to their failure to repeat an unlikely victory at Stamford Bridge last season and is thus an unfair representation. The expected points model would appear to confirm this.

All in all, my expected points model and Zach’s MSq£ model have broadly similar residuals with the odd smallish difference here and there. The two Manchester clubs are the most interesting with Zach’s model rating them as currently performing beyond expectations and mine suggesting they are just about performing as expected.

This entry was posted in Uncategorized and tagged , , , , , . Bookmark the permalink.

One Response to Three expected points models for the Premier League compared

  1. mark says:

    great stuff, Simon,
    I love this kind of analysis. It often leads in lots of different and useful directions, Respect for putting the work in to keep it updated :-).
    i’m currently looking at how quickly teams get re evaluated by the books on the back of unexpectedly good or poor runs for a guest post. Last season Newcastle’s odds to beat Wolves at home was a lot shorter compared odds available on their early season meeting, even allowing for venue change. Newcastle upgraded, Wolves left pretty much the same? But also lots of cases where early season “improved” form failed to persist.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s