Saturday, April 11, 2009
New Post on Puck Prospectus: The Plus and Minus of Plus/Minus
"Hockey is usually not the sport where statistical or analytical innovation takes place. So it's ironic that one of the most useful concepts in sports statistics has been an official NHL statistic for over 40 seasons, leading other sports by at least three decades. The statistic in question is Plus/Minus, which is perhaps the biggest subject of debate in hockey right now. Even though it accounts for the two most important actions in hockey, scoring goals and preventing goals, nobody can agree on what's wrong with it or how to fix it."
http://www.puckprospectus.com/article.php?articleid=70
http://www.puckprospectus.com/article.php?articleid=70
// posted by Hawerchuk @ 4:39 PM 
Comments:
Links to this post:
<< Home
Thanks for blog posting the links here. It's nice to be able to leave comments and get discussion going (which you can't do on Puck Prospectus).
My general issue with using +/- or ANY measure that uses goals (both For and Against) as its primary measure is that goal scoring rates have considerable variance over small sample sizes (i.e. - one season).
The main reason I prefer Corsi, or even Expected Goals (based on some sort of shot quality model), is that it seems more predictive of future production.
The issue with Expected Goals is the inherent problems with the shot quality model - recording bias, lack of descriptive shot data (i.e. - distance from net and not much else), etc.
The issue with Corsi is that it's so heavily team dependent, which makes it difficult to compare players on different teams.
Your "Corsi Rating", which I assume factors out a lot of those team dependent things, seems like a good metric. You mention in the article that all those stats are available on behindthenet, but I can't find Corsi Rating anywhere.
?
Post a Comment
My general issue with using +/- or ANY measure that uses goals (both For and Against) as its primary measure is that goal scoring rates have considerable variance over small sample sizes (i.e. - one season).
The main reason I prefer Corsi, or even Expected Goals (based on some sort of shot quality model), is that it seems more predictive of future production.
The issue with Expected Goals is the inherent problems with the shot quality model - recording bias, lack of descriptive shot data (i.e. - distance from net and not much else), etc.
The issue with Corsi is that it's so heavily team dependent, which makes it difficult to compare players on different teams.
Your "Corsi Rating", which I assume factors out a lot of those team dependent things, seems like a good metric. You mention in the article that all those stats are available on behindthenet, but I can't find Corsi Rating anywhere.
?
Links to this post:
<< Home
Subscribe to Posts [Atom]
