Quote:
Originally Posted by Resolute 14
IMO, one problem is that one of the key aspects of how Corsi was used early on by analytics types has been lost: The relative relationship between possession/Corsi and winning is stronger with a large sample size. When there's 80-100 Corsi events in a game, and 3000+ in a season per team, you have enough data points to at least have a decent picture. Even if things we cannot or do not factor still create a larger margin of error.
But so much focus has been put on single games in the last couple years that we've eliminated one of the strengths of the stat. Especially since the effects of ignoring special teams becomes dramatic at the single game level.
Consequently, there is no correlation between winning and Corsi at the level of a single game. To wit, our record when....
CF% <= 40%: 2-0
CF% < 50%: 2-0-1
CF% >= 50%: 8-7-0
CF% >= 60%: 3-2-0
On the other hand, when we look only at the one thing analytics types typically ignore, special teams, we see a much stronger correlation:
When Flames win special teams: 5-2-0
When Flames lose special teams: 4*-4-1
When Flames tie special teams: 1-1-0
*Includes the Rangers win where we scored three goals on delayed penalties. Did not count as power play goals in the stats.
|
Analytics types? That's kind of loaded isn't it?
And of course sample sizes get in the way, but they do with almost any stats. If you looked at out shooting opponents in a ten game stretch it may suggest you're better to take less shots, but that wouldn't be a good game plan.
The problem with your special teams argument is it's too simple. You're pointing to goal differential by saying "win the special teams", so yeah of course it's going to matter. Its like saying if you spot a team a lead you're going to lose more often than not ...
So it doesn't take an analytics type to see special teams is a separate discussion, and doesn't belong in a comparison any more than "on nights where your goalie out plays their goalie you win more often" ... once again ... no kidding!