Monday, January 14, 2013

It's All Relative Part I/III: 2011-2012 EPL Midfield Rankings

In soccer, coming up with a player rating system presents quite a challenge.  Well, actually it can be quite easy to create a system, but making a good system has proven difficult.  In particular, how can you provide a single metric that can apply to all different types of field players (forwards, defenders, wingers, etc.)? Quickly the exercise devolves into one of art more than science.  How much should goals be weighted?  How do you rate defenders?

Progress is being made.  Squawka may well yet usurp the dreaded Castrol ratings, and I feel that we may be only a year or two away from a generally accepted standard rating system.  I know I won't be the one providing it, but since the Manchester City/ Opta folks have been so gracious with their 2011-2012 EPL data,  I figured I could at least throw my spreadsheets into the growing canon of online soccer analytics.

I have decided to sidestep the tricky Forward/Midfielder/Defender conundrum by tackling (no pun intended) each position on their own merits.  Naturally, since this is the first of a three parter, I have taken on arguably the most nuanced position group, Midfielders.  On one end of the spectrum, you have the purely attack-minded looking to score goals at every opportunity (Clint Dempsey).  At the other end are the defensive destroyers (Nigel de Jong).  As such, I broke the ratings into three general categories: Attack, Possession, and Defense.  

Deciding what metrics should go into each category was perhaps the most difficult exercise, and one that I won't go into too much detail on as it would bore the heck out of me.  One notable change I made after my initial list was the inclusion of the "successful dribbles" category in the Possession group.  I noticed that wingers were getting abnormally low scores.  Even though the statistic may seem attacking in nature,  I feel it is really more a transition statistic as players carry the ball forward from their end, thereby transitioning from a defensive to offensive stance. 

The most notable/unusual thing to note with these ratings is that they are all in standard deviations.  For the most part, these metrics are normal in their distribution and there was enough of a sample (134 midfielders) to feel comfortable with this approach.  For the non-statistic inclined, what this means in terms of the tables below is that a SD score of 1.00 generally means you are in the 85th percentile and -1.00 is in the 15th percentile, 2.00 is in the 98th percentile and -2.00 is in the 2nd percentile.

I like this approach because it rewards abnormal behavior (both good and bad).  For purposes of the overall ranking I took the assumption that Attack, Possession, and Defense be equally weighted.  This may be debatable, but because of this I adjusted the Attack score to account for the extra variable in that category.  Criteria for inclusion was 900 minutes of play in an EPL league game at a midfield position (according to OPTA).  Players are presented in order of overall ranking from 1 to 134. I apologize in advance for the size of the tables.  If you want an excel version of these just comment below or hit me up on twitter.

No comments:

Post a Comment