Friday, 23 November 2012

GW13 Predicted Scores [by SuperGrover]

Weekly collaborator on this site SuperGrover once again brings you his predicted scorelines for the weekend's games.  He also adds some follow up and commentary to his predicted scores from Gameweek 12 wiht observations and lessons learned. Note that SuperGrover and I are developing our own seperate models which are very similar overall but the data used in the regular articles on this site, point projections, etc, are based on my model, not his.  Without any further ado I'll hand over to SG...




Game Week 13 Predicted Scores
Welcome to my Game Week 13 team ratings and score predictions.  Week 12 was certainly interesting, with two early red cards, a few scoring explosions, and a Carrow Road upset highlighting another unpredictable week in EPL.  Below please see my team rankings and score predictions, along with a few notes for further consideration.  As always, comments and criticisms are very welcome.


Game Week 13 Team Ratings
 


Game Week 13 Predicted Scores
 

Commentary on Game Week 12's Predictions (linky)
My prediction for Reading vs. Everton was 1.20 - 1.67 and a goal for each team.  However, while the ultimate result was disappointing, the underlying numbers are more indicative of a 2-1 Everton victory.  It’s not often that 6 shots on target and 4 Big Chances result in only one goal.  Further, Reading’s numbers look like a single goal scoring performance.  Obviously, the penalty skewed things a bit here, but this is a case in which the final score didn’t correlate with the performances on the pitch.

No model will ever predict a 5-0 goal victory, so my model’s 2.28 goal prediction for City shouldn’t be too bothersome.  With that said, I do have to admit my model’s ratings of Villa’s defense are troublesome.  The reason the Villains are rated so strongly has to do with the number of Big Chances allowed.  Coming into this weekend, they had allowed only 14 Big Chances on 93 shots in the box.  This ratio of 6.64 is well above league average and does indicate a defense that may be a bit overrated, although the correlation between Big Chances and shots in the box is not strong enough to adjust the ratings accordingly.  With that said, it may be simply a situation in which Villa’s strategy and performance met up against elite talent performing at their peak.  I’ll have to monitor the situation going forward and look at possible adjustments to the model accordingly.


The Southampton – QPR fixture was the most surprising of the week in my opinion.  While QPR has been awful recently, the model still projected a fairly easy win against an equally bad Southampton team.  Instead, the Saints absolutely dominated the game, generating enough chances to justify 2-3 goals and fully earning the three points.  Needless to say, the model has punished QPR accordingly, dropping their total rating by over 30%.


Hosting Man United, Norwich recorded their 3rd clean sheet in a row and fourth in five games (the other game was a 1 all tie).  Yet, the model still has their defense rated in the bottom five of the league.  This can be explained simply by looking at Norwich’s performance early in the season.  Norwich gave up an average of 3.33 Big Chances per game over game weeks 1-2, and 4-7 (game week 3 was an adequate performance against Tottenham).  The first seven game weeks also included an average of 5.57 shots on target allowed, the worst figure in the league.  To be blunt, Norwich City was the worst defense in the league over the first seven weeks.


In the past five weeks, Norwich has turned their season around.  In those games, they have allowed only 4 Big Chances and an average of just 4.2 shots on target.  These numbers are more indicative of a league average defense and, not surprisingly, Norwich has the 9th ranked defense over the past six weeks (which includes the awful performance against Chelsea).


Implicit in this discussion is the fact I am not currently emphasizing recent performance or team form in my model.  I do want to eventually account for form, but I am not yet comfortable with the initial attempts and would prefer to leave as-is for now.  Please keep this in mind when looking at my rankings and predicted scores.


·

 

No comments:

Post a Comment