What’s so wrong with the Capello Index?

The Capello Index has generated a lot of criticism of England coach Fabio Capello since word of the rating system was released before the World Cup.  The criticism has intensified since the index was published today, with Capello now demanding that the site be removed from the internet.  So what is so wrong with the Capello Index?  Other than the fact that it was created in conjunction with a gambling company and produced results embarrassing to the English FA, it's actually a fairly decent index that has fewer of the defects of similar rating systems.

An explanation of the Capello Index is posted here (although I don't know how much longer the site will be available).  It is really quite a document and it does a good job of placing the index within the general area of player performance evaluation and why it is important to have a systematic approach in a multi-billion euro football industry.  It's not very well formatted so if reads like a few very long paragraphs, but I recommend reading the whole thing.

Explaining how the index works will require some reverse engineering, and since the low-level details are not revealed I can't be sure that they're actually doing it as I describe, but it is a reasonable educated guess. The index is created by dividing a player's performance into a series of actions that he carries out on the pitch — over 500, according to the index's creators.  The performance takes into account these events, where on the pitch they occurred, and at what time they occurred.  For each event, a value between 0 and 100 is assigned.  The actual Capello Index is calculated by taking into account all of the player's events over the match, on a scale between 20 and 100.  It appears to be a weighted average of player events, and those events have complex scalings of their own.  Perhaps there is some type of regression to arrive at the scale factors for various events, but there also might be some scaling expressions that were arrived at through input from various football experts.  Enter Fabio Capello.  He would have the experience as a player and as a manager to know what events matter more in a football match than others, as well as knowledge of the small events on which a match can turn. 

Despite the index's creators saying that the index is purely objective, it is still based on a rating of match events that is subjective in nature.  If they used a regression technique, it would remove a lot of the subjectivity from the analysis, but it would require a lot of data to calibrate.  My guess is that the team used a scaling function that they arrived at from discussions with Capello or other experts they had available.

The Capello Index produces results that are different from Castrol's index or the Footballer rating developed by the Amaral group at Northwestern, but very illuminating and, in my opinion, more credible.  The top-ranked World Cup player on the Index was the Golden Ball winner, Diego Forlán, followed by Miroslav Klose, Thomas Mueller, Andrés Iniesta, and Xavi.  All very believable.  In contrast the Footballer rating's top five were Lionel Messi, Felipe Melo,  Frank Lampard, and Yaya Touré, and the Castrol index top five were Sergio Ramos, Joan Capdevila, Carles Puyol, Philipp Lahm, and Gerard Piqué.  The Castrol and Footballer indices are similar in how they compute their ratings, but it is interesting that the Footballer index is heavily dominated by offensive players, while the Castrol index is dominated by defensive players.  (The Castrol index takes into account tackles and "player movement" — whatever that might mean.)  The Capello index appears to take into account offensive and defensive events in a more balanced way, and that is reflected by a more even distribution of position players in its top 20. 

The index makes for very interesting reading when you consider the group stage.  Diego Forlán and Landon Donovan have nearly identical ratings over their three matches, and Michael Bradley is the highest-ranked American player at #12.  Xavi, Schweinsteiger, Higuaín, and Messi were the top four players in the group stage, and three goalkeepers were in the top ten.  No England player had an index rating in the top 60, and just one in the top 100 (Steven Gerrard at #63).  That is perhaps the most embarrassing statistic of all in the eyes of the English FA.

The Capello Index isn't a perfect rating system by any means — which one is!  — but it appears to do a better job of capturing offensive and defensive contributions than any of the all-inclusive player performance rating systems out there.  It also produces results that are credible in that the top-ranked players are the ones that most observers would expect, while at the same time revealing who might be over or underrated in world football circles.  There's nothing wrong with the Capello Index aside from the fact that it makes Fabio Capello's employers look really bad.