Move over Opta. The Times yesterday launched Fink Tank, their unique rankings of football players in the Premiership. A heaven of statisics, I’m going to take a closer look later. First, a few observations:
1) Introducing the rankings with the line: “How did Fink Tank do it? We used a multivariate Poisson log-normal model. I hope you find that information helpful.” was pretty off-putting. Only real serious mathmaticians care about running several Poisson distributions as a method of removing anomalies – if that’s what it is…
2) The model works by identifying “the relationship between goals scored and every kick of the ball made by every player for every club”. I presume this takes into account defending as well as attacking. But it mention tackles, pressure that indirectly relates to goals, or other parts of football. A more detailed methodology is needed.
3) It penalises players for lack of time on the pitch – which is strange, as some players are used deliberately as a “super-sub” or impact player, and their lack of pitch-time is a virtue.
Here is the methodology:
The Fink Tank Predictor provides forecasts and ranking systems for English and European club football, based on a statistical model of matches based on more than five years of football scores.
In looking at player rankings for this season, the phrase “time-adjusted points” means the number of points the player would have added to an average team in the full season, compared with an average replacement. The points are then adjusted to reflect the amount of time spent on the pitch – minimum 400 minutes.
I’ll take a closer look in my longer, research blog later. Time to crunch a few numbers…