With the season half over I have taken a look at how the model’s ranking of teams compares with their Premier league position.

As can be seen in the table below most teams’ league position is similar to their algorithm ranking.  A ranking above the league position suggests all flare and no substance or plain bad luck.  For a team to be unlucky over 20 mathces however is unlikely.  It is more likely to suggest therefore a good team with lots of chances but lacking the finishing ability or unable to close out a match.  Conversely a team with a league position above their ranking shows an ability to get results without creating a lot of chances or dominating a game.

Most team rankings and league position are in line.  However there is one major exception - Man City.  Their ranking is relatively low because the model does not take into account defensive qualities whereas it does take into account attacking qualities.  Man City rank low on a combination of the two compared to their league position.  Ericsson’s style is to earn points by being solid at the back without risking much up front.

There may be an opportunity to improve the model by capturing the defensive quality of a team. At the current time however I am not aware of any data available to quantify this.