View Single Post
Old 04-07-2018, 08:00 AM   #487
Itse
Franchise Player
 
Itse's Avatar
 
Join Date: May 2004
Location: Helsinki, Finland
Exp:
Default

Quote:
Originally Posted by DeluxeMoustache View Post
The idea behind Corsi is quantifying the proportion of behaviour (shots) that lead to the expected result (goals).
This is incorrect. Shots on goal is quantifying the proportion of behavior that's expected to lead to goals. Corsi (or relative corsi to be more specific) is a proxy stat for "possession". Possession is supposed to lead wins. Not goals for or against, wins. (Or, "more goals for than against".)

Two facts:
- From what we know of other sports, some of which track possession directly instead of trying to do it through a proxy (soccer being the best example since it has by-the-second tracking of the ball and all players), possession is a quite poor predictor of wins. While not all sports are the same, I have seen no explanation why possession would be somehow especially relevant in hockey.

- Shots are not a very good indicator of who actually dominated the game. Adding "shots not on goal" to "shots on goal" is adding noise to what is already a rather uninformative stat. In general, increasing sample size is rarely a great way to improve a statistic beyond a certain basic level.

In other words, whether you think corsi is supposed to be an improved shot tracker (questionable whether it's actually improved) or a proxy possession tracker, it's highly questionable whether either of those statistics are that useful. (Plus it's questionable whether corsi is a good proxy for possession in the first place.)
Itse is offline   Reply With Quote
The Following 4 Users Say Thank You to Itse For This Useful Post: