Differentgame is taking an Everton turn this coming season. The aim is to focus on how ‘advanced’ data analytics could be applied to a football club in a meaningful way.
Lots of people dislike the use of statistics in football but it is a big part of the future. Advanced data analysis increasingly influences American sport franchises and in the Premier League it’s some of the clubs with the most money (Manchester City, Liverpool) that are dedicating most resources to it.
Ironically, it’s the clubs with the least money (i.e. Everton) that could probably benefit most from the process. That said, Everton’s use of more usual performance analysis techniques over the last decade is second to none.
You’ve probably read articles like this detailing Roberto Martinez’ home set-up for analysis. The Nowhere Men by Michael Calvin also hails the meticulousness of Everton’s approach under David Moyes too.
Yet earlier this year at the Elite Minds in Sports conference, James Smith, Everton’s former Head of Tactical Scouting had this to say: “We’re not using sophisticated regression analysis when scouting players – that’s what bigger clubs do. We do more GSCE maths stuff – averages, bar charts – when it comes to player analysis but that will change. We don’t spend enough on staff in English football. We spend so much on wages, agents and transfer fees. It’s crazy.”
Regression analysis is the kind of stuff you’ll find here on this site. As the season progresses we’ll be using many more tools to try and provide insight (or simply confirm what we think we know already) into what’s going on on the pitch at Goodison Park.
Barry, McCarthy and Gibson
Gareth Barry was signed for another 3 years this week and I wanted to quickly (ahem) demonstrate something vaguely interesting about the three central midfielders now on the clubs’ books.
At the inaugural Opta football analytics conference earlier this year (attended by numerous guys from Premier League clubs, including Everton), scientist Marek Kwiatkowski presented a session on ‘passing dissimilarity’.
The data collected on games now allows each players passing to be minutely analysed using sophisticated techniques. Marek was able to compare players’ passing based on the positions they made the pass from, the angle and length of pass and the volume of passes each player makes. Marek used data from the 2012-13 season when Gibson was playing regularly for Everton, Barry was at City and McCarthy was still at Wigan.
After the conference I asked Marek for the details of Everton’s central three. Here’s how Barry compared to other midfielders around Europe in his last season at City. Angle and length of path is on the verticle axis of the graph meaning the more similar the player is with his passing, the lower down on the image he will be:
The player on the chart that Barry was least like in the Premier League was Gerrard (maroon dot top leftish). Future team mate, McCarthy is at the bottom right of the chart (yellow dot). In terms of passing angle and length, both Barry and McCarthy were pretty similar.
We can see from Marek’s 2012/13 data that Gibson liked to stretch the play more than the other two:You can see here the huge difference between him, Barry (top left) and McCarthy (top right). We saw the striking difference between Barry and Gerrard in the first plot and here you see Gibson’s similarity to Gerrard (turquoise dot in the cluster of players at the bottom).
I’ve started to go through Everton footage from 2013/14 and have been struck by Barry’s passing range. I checked the data and found that he played almost three times as many long balls last season than he did in his final year at Man City (he played an extra 400 or so minutes but this doesn’t account for the leap). Also, the average length of Barry’s passes had gone up by over 4m since his move to Everton. Considering he played well over 2000 passes last season that’s some amazing extra distance.
Similarly, McCarthy’s average pass length has gone up nearly 3m since his move from Wigan to Everton. Again that’s over a large number of passes (1700+). The whole range of both players’ passing (backwards, forwards and sideways) has become longer. What’s more, both have maintained passing accuracy at 86-87% during the change.
The Squawka Comparison Matrix is a nice tool to allow us to show average pass length figures. Under Martinez at Everton, Barry and McCarthy’s passing has become more like Gibson and Gerrard’s:
You could envisage Marek’s data being a useful tool for replacing departing talent with like-for-like players. If scouts are looking for players comfortable in the current system then this could confirm or deny their reported findings.
That said, the change in the data from one season and one club to the next for Barry and McCarthy show that at least some players are flexible. What a player has done in the past is not necessarily what he’ll do in the future. A player is a product of what he’s asked to do by coaches and management. Multiple seasons of data across multiple leagues analysed with Marek’s techniques would be extremely useful in identifying flexible players quickly.
In Barry and McCarthy, Everton have two players with discipline for the short game who are more than capable of maintaining accuracy when going long too. Add Gibson into the mix and Everton have a pretty enviable passing range from the middle of the park. If Gibson can stay fit this coming season, Martinez has the options to cover all bases.
If you’d like to compare players your self using Marek’s data then go over to Steve Fenn’s site where he’s made an interactive viz for that very purpose.
Follow me on Twitter @footballfactman