Expected Goals Models – What’s the Use?

Last week I posted about one of my own expected goals models and briefly touched on how it’s applied. I wanted to go in to greater depth about its possible uses and had been thinking about it during the week when I finally got round to watching the ‘football as a dynamic system’ presentation by Prozone’s Paul Power (@counterattack9).

If you have half an hour spare it’s well worth a full watch. If not, skip to 10mins in which is where it gets really interesting. Paul starts to talk about creating ‘overloads’ in the right areas. An overload is basically where you have attackers in greater numbers than defenders. He specifically talked about creating the overload in the danger zone where the ball is going to end up (in the middle of the box) rather than creating it where the ball is delivered from (out wide). Simply put, have a stack of players out wide creating the chance and its likely you’ll have less team mates in the middle to finish it off.

Being an Everton fan I immediately started to think of David Moyes at this point in the presentation. One of Moyes main tactics was to create overloads on the left hand side – stationing Osman or Fellaini towards Pienaar and Baines. The ball would be worked out there for a delivery into the box. It was effective in that it enabled two things – Everton to have a lot of ball in the opposition’s final third, and also, the opportunity to create bags of chances.

The model I concentrated on last week was purely based on shots on target. Another model I use looks at ALL the shots a team has, but more importantly in this instance HOW they were created. Here’s a graphic of the number and type of chances Everton created in the box under Moyes during his last season at Everton, and the number and type of chances created this season under Roberto Martinez:

EvertonComparedIn Moyes’ last season, Everton took nearly 100 more shots in the box than it did under Martinez. According to my Chance Creation Model, the benchmark for a team taking the shots Everton did in the way they did in Moyes’ last season would be 60 goals. Everton managed 52. This season the benchmark for the shots Everton took would be just 50 goals, but they scored 58 (own goals aren’t counted in the model). Everton turned from one of the least efficient attacking sides in the league to one of the most efficient.

Despite Everton dominating the final third in their games under Moyes, they were constantly bombarding a box that was heavily defended. They failed to create overloads in the right areas. You could see that from simply watching. What you couldn’t see with your eyes is an actual goal figure of how much this was costing Everton. With an expected goals model you can start to do this.

It isn’t immediately apparent (as it’s not labelled) but the example Paul uses of creating an overload in the actual danger zone rather than out wide, is taken from Everton’s game against Chelsea earlier this season. Under Martinez, Everton are much less likely to just sling crosses in. Here’s a video of that goal Prozone have simulated in Paul’s presentation:

An expected goals model can be the start of team (and player) diagnostics. If it ain’t going right, it can start to tell you where it’s gone wrong and put a figure on how much it’s costing you. It can start to tell us a lot about the effectiveness of how a manager is setting his team up. Despite displaying such attacking inefficiency, Moyes’ set-up still managed to better previosu years’ shot numbers, goal tally and points tally by playing this way. Put simply, it worked. Martinez, however, found a way to buck the system last season and improve again.

Just how well is your manager performing? Expected goals models provide a benchmark for this. Before you think of hiring or firing your next manager, you might want to give these numbers a crunch. It’s the first thing I’d do if I was running a football club.

Follow me @footballfactman here on twitter.


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

4 Responses to Expected Goals Models – What’s the Use?

  1. Tery Whenett says:

    Great stuff, thanks for sharing!

  2. This is very interesting.
    I was wondering though how much of the 16 goal difference in efficiency could be attributed to having Lukaku instead of Jelavic or just statistical variation?

  3. Its a good point. It could be either. I do these numbers and have watched the vast majority of everton games over the last 5-6 yrs. Some of it is system, some of it is player, some of it will be random.

    I picked one feature out to highlight here sparked by that presentation. I also think everton won’t be so lucky next season and have a hell of a job on to match this yrs outcomes next season.

    The general point is, here you have a benchmark to judge ur own efficiency. It’s up to you to find the reasons – whether through analysis, trial and error or whatever.

    I approach this with a football head on rather than a maths/stats head. You can let the numbers cripple you saying its all random or you can choose to do something and see what the outcome is.

    I think lukaku in the moyes system would have been a poor fit – I did a piece around this if you search the site.

    Thanks for the comment. It was a good one!

  4. Pingback: Three Points : TheChels.co.uk : Chelsea FC Statistics by @ChelseaStats

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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