A few weeks ago I posted the method to a new ‘rating’ (for want of a better word) I had for central defenders (here: https://everyteamneedsaron.com/2015/10/01/new-method/). It is clearly not going to be perfect, but it has a logical base behind it, which is more than my previous attempts have had, and it doesn’t seem to suggest anything too ridiculous. With the Opta Pro deadline looming (tomorrow, as I write this) and me unsure whether I should cheekily ask for tens of seasons worth of data so that I can check it more thoroughly, I thought it would be a good idea to look at the ‘results’ in more depth.
Due to the fact that it uses ‘deep completion attempts conceded’ (passes attempted within a deep area of the field), a stat which I don’t have available for any seasons other than the 14-15 Premier League season, that’s the only one I’ve been able to do so far. Here below are 65 players who played at central defence in that season who had a rating. All of them must have had over 10 90s.
A couple of quick points:
1) I believe that Chelsea shielded their central defenders to such a large extent last season that it’s extremely difficult to make Terry and Cahill’s stats workable, and is an unfortunate outlier to appear in the only properly tested season.
2) I’m not claiming this to be a perfect model, as those (I presume) require a lot more complex methods than this, but I think that this is a decent start and that assessing centre-backs statistically shouldn’t require tracking data.
3) Dier scores so lowly because he played games away from CB last season, which affects various parts of the method (eg more likely to lose the ball which is one of the adjustments). Others also may be affected by this, but not to such a notable extent as Dier. I should probably take him out of the dataset, although if doing this for other leagues and seasons there would probably be similar cases that I would be unaware of (ie I’d just be faced with a player who scores very very badly) and so keeping Dier in can act as a reference point for other leagues.
It’s also notable that there isn’t a lot of difference between an awful lot of these players. Eg a group from Ashley Williams to Emre Can. In a way, I see this as a positive, because in reality there probably aren’t huge differences between players (especially when they are then to be put within a system).
As I mentioned in the original post, the more logical and better laid out way I have done this means that I can see why players who are liked by the mainstream don’t do that well on this. For example, Mamadou Sakho, who seems to be generally well liked, loses the ball quite a lot (being dispossessed, bad touches, inaccurate short passes), and therefore his ‘score’ gets adjusted downwards because of this. Toan extent, therefore, it seems that were he to take better care of the ball then he would score higher, and be a ‘better’ central defender’
There are some concerns in this, even after looking at it for a while, mainly the clumping of Manchester United CBs towards the top of the list and high scoring Leicester City CBs, plus the high score of Jonny Evans who, according to this, is probably quite good but instead was sold to West Bromwich Albion.
I also created a table of the average centre-back score for each team, based on their two CBs who played the most time in my sample. Generally, this will probably be the ‘first choice’ pairing, but not always (eg Demichelis and Mangala were used from Manchester City above Kompany). For Liverpool and Spurs I also didn’t use Can or Dier, as I they were not always used as CBs.
Chelsea, of whom I have already mentioned my reservations (/excuses), aside I am not too displeased with this table. Villa and Leicester being so high is surprising, though Leicester’s CBs were a concern anyway and both Ciaran Clark and Jores Okore for Villa have both previously been fairly well considered by the mainstream. Manchester City would also score higher had Kompany been used; how much higher depends on who he is replacing.
I figured that it might be useful for people to see how everyone scores. Thanks for reading. If you wish to comment you can do so in the comments below, on Twitter @ETNAR_uk or my email is firstname.lastname@example.org