View Single Post
Old
03-04-2012, 03:45 AM
  #93
BillDineen
Registered User
 
BillDineen's Avatar
 
Join Date: Aug 2009
Posts: 3,009
vCash: 500
Quote:
Originally Posted by SMantzas View Post
Sorry for trespassing as a Devils fan, but I think you're taking CORSI and Fenwick out of context. I dont think these stats are good on a team level, but rather on an individual player level. Corsi is easily the best stat to predict future performance and past performance. I am by no means a "math guy", but advanced hockey stats are years behind baseball sabremetrics.
No offense to you, because that statement probably comes from blogs you have read. But those are the kind of statements I take issue with. I have not seen any study of predicting future performance with Corsi. I would love to see one across teams. Ie. UFA signings or trades. (5v5 only is fine, although Corsi should predict something about special teams if it is a measure of performance).

I can look and see that Briere and Leino had similar numbers in Philly, does that mean Briere's production would fall off a cliff if he went back to Buffalo instead of Leino this season? I doubt it. Could it predict a player like Lupul or Versteeg playing better with Toronto and Florida?

At a lower level, even if corsi numbers tell you JVR against tougher competition (as defined by corsi) might generate more all shots for us and less all shots against us than Briere and the coach should play him more even strength minutes. That says nothing about the additional scoring chances generated. It treats all shots and missed shots and blocked shots the same and judges QoC in the same light.

Note that if a team valued possession like the Pens and Wings (in above article and video), they could just hire two undergrads to track possession daily (day after on video) as the NHL did in the 90s (by zone) and then perform all the same analysis, possession when a player is on the ice vs off, adjust for the quality of teammates and opponents (as defined by possession). At least then the analysis would be sound because the underlying statistic (possession) is sound vs Corsi (which treats all shots the same) which IMO is not sound as a basis.


Last edited by BillDineen: 03-04-2012 at 04:02 AM.
BillDineen is offline   Reply With Quote