Xtratime Community banner

1 - 3 of 3 Posts

·
Registered
Joined
·
24 Posts
Discussion Starter #1
I worked up a ranking system for club teams based on statistics that I found to correlate the most heavily to winning. With all the different ranking systems and comparisons made across leagues, I thought it would be interesting to try and come up with a statistically based ranking system. Additionally, I wanted to make sure that my rankings correlated with team success, and I found via linear regression that my ranking system is about 87% correlated with the number of points a team finished the season with.

My rankings and summary can be found here.
 

·
Registered
Joined
·
8,439 Posts
Well, it's a good effort, but the fact that Atleti is like 9th overall on your list, means you have some work to do. This study is another proof that you cannot quantify football with numbers.

Also, you mention "world clubs" but all of the ones you listed are European. Also, there doesn't appear to be much of a correlation between your score and a team's point total. Your study is very biased to be honest. To make something like this universal across various leagues, you'd want to standardize the point totals as some leagues play different number of games in a league season. I could say quite a few more things about this, but I'll leave it at that.
 

·
Registered
Joined
·
24 Posts
Discussion Starter #3
Thanks for the feedback. There will always be teams that "outperform" their statistics, and Atletico are a perfect example. Based on "expected points" statistics (done by several other analysts), Atletico outperformed their expectation by 4-5 points. My rankings are meant simply as a statistical ranking system, to be used with the so called "eye-test", not to replace it.

Yes, calling it a "world" ranking was incorrect, but I did mention in my summary that it was only clubs from the traditional top 5 leagues in Europe.

I also don't understand your claim that my score does not correlate to point total. On my page, I included a linear correlation to both points and points per match, both with very high correlations (R-squared values of .87 and .89). Could you elaborate on that?
 
1 - 3 of 3 Posts
Top