Advanced Stats and Corsi - Another look. (Mod Warning - Keep it Civil)

[duck]

I admittedly don't know too much about Corsi and Fenwick, but I do imagine that there is some correlation between shots against and team success. With the Leafs on pace to have the third most shot attempts against in the past twenty-or-so seasons (the two ahead of the Leafs would be two teams from 93-94: the Kings and the Blues), I don't think the recent results (like the 8 game losing streak) are shocking...instead, it just seems like regression to the mean.

 

[BraveCanadian]

Why are some of the worst canadian corsi teams still in the playoff race (Habs, Leafs) while the two best canadian corsi teams have been eliminated (canucks and senators)?

Because so far imo the predictions based on Corsi don't work out very well. It is pretty great after the fact though.

It was supposed to be unsustainable last season. Unsustainable this season. Supposedly caught up to them in December before they got hot again. It took an unsustainably(!) bad run of games to finally make the "prediction" come true.

Obviously Corsi does indicate there are possible leaks springing in the boat but anyone watching the games or looking at the simple shot numbers could have told you the same.

Then again being in the playoffs/out of the playoffs is a very binary thing. Even after that terrible run the Leafs are still .500 in regulation / OT if I am not mistaken.

Shouldn't they be losing more?

There are some pretty obvious pieces of the statistical puzzle missing. And the advanced stats like Corsi/Fenwick etc. are not very advanced to put it mildly. Although the available measures have gotten a lot better recently with more things being tracked by the NHL.


(survey says they probably will lose more with the way they have been playing lately! )

 

[Disgruntled Observer*]

Wait... you mean getting outshot and outchanced almost 2-1 most of the time was "bad"? Getting hemmed in our own zone 3/4's of every game was "bad"?
These things actually caught up with us?

I thought non of those things mattered? I thought our snipers were 10 times better than every other team so getting out-chanced didn't matter?
I thought our goalies were and ALWAYS WILL BE 10 times better than every other goalie put together?

All those things the majority of leaf fans told me weren't true?
wow.
Shocking.

 

[ULF_55]

Wins 38 - 76 Points:

• ROW 29
• Shootout 9

Losses 41 - 8 points:

• Regulation 33
• Overtime 4
• Shoot out 4

 

[Kurtz]

Can someone provide the argument for why CORSI is better than simply Shots for/against?

Seems to me that CORSI is inferior to its predecessor, because it neglects shooting accuracy and shot-blocking metrics, which are rather important, and which shots for/against does inherently cover.

 

[Gallagbi]

Can someone provide the argument for why CORSI is better than simply Shots for/against?

Seems to me that CORSI is inferior to its predecessor, because it neglects shooting accuracy and shot-blocking metrics, which are rather important, and which shots for/against does inherently cover.

Corsi is supposed to measure possession, so the more shot attempts, the more possession you should have according the Corsi.

 

[Morbo]

that's why I brought up that bad run Volc, it brought them back to the pack that opened the door for a short bad run like this, to do what it did.

It's pretty bad to lose 8 in a row and get no points in Gary's league. that's always going to hurt you really badly, nothing can really mitigate getting no points for that kind of stretch that late in the season. Even the Leafs hadn't done that in 30 years.

that's why they didn't make it.

to what I bolded

maybe you just have issues with people being right, when the news is bad.

I would apply the bolded to the folks that railed against the advanced stats and turned out to be dead wrong.

If I thought they were right I would say so.

not sure what your point is here. they're not right. the Leafs didn't miss the playoffs because of possession, it's pretty obvious.

 

[Kurtz]

Corsi is supposed to measure possession, so the more shot attempts, the more possession you should have according the Corsi.

Oh sure, but shots for/against are correlated to possession also. Is there evidence that CORSI has a significantly higher correlation coefficient to possession than shots for/against?

 

[Mess]

Corsi is supposed to measure possession, so the more shot attempts, the more possession you should have according the Corsi.

Corsi also includes missed shots and blocked shots ie. shot attempts at the net not just shots on net for and and against that ever boxscore will record.

 

[Gallagbi]

Oh sure, but shots for/against are correlated to possession also. Is there evidence that CORSI has a significantly higher correlation coefficient to possession than shots for/against?

Corsi should be a more accurate measure, since there is more data. There should always be more shot attempts than shots.

Corsi also includes missed shots and blocked shots ie. shot attempts at the net not just shots on net for and and against that ever boxscore will record.

I thought that was covered in his post since his question is why use shot attempts instead of just shots?

 

[Teeder9]

Anyone know of a site that lists the numbers for home and away for each player? I'd be interested to see how other coaches use their best players against us when we're in their building

 

[Kurtz]

Corsi should be a more accurate measure, since there is more data. There should always be more shot attempts than shots.

I kinda see what you're saying, but if each team averages lets say 28 shots per game, that amounts to a total of about 70,000 shots on goal in the NHL through the conventional, old-school measure, during the course of a season.

The margin of error on a sample of 70,000 is tiny, and only negligibly bigger than on a sample of say 140,000 that CORSI may provide (way less than 1%, I think).

But I'm not convinced that that's the correct measure of sample size, since in this approach the sample size might be based on games played, and the shots would simply be a variable inside that population sample size.


On a side note, I've noticed a high degree of direct positive correlation between having Mess on your ignore list and a decline in headaches.

 

[Gallagbi]

I kinda see what you're saying, but if each team averages lets say 28 shots per game, that amounts to a total of about 70,000 shots on goal in the NHL through the conventional, old-school measure, during the course of a season.

The margin of error on a sample of 70,000 is tiny, and only negligibly bigger than on a sample of say 140,000 that CORSI may provide (way less than 1%, I think).

But I'm not convinced that that's the correct measure of sample size, since in this approach the sample size might be based on games played, and the shots would simply be a variable inside that population sample size.


On a side note, I've noticed a high degree of direct positive correlation between having Mess on your ignore list and a decline in headaches.

It should simply give you a more accurate reading. The belief of Corsi supporters is that all shot attempts should signify possession, so whether a shot is blocked, misses the net or ends up as a SOG shouldn't matter.

It should be more accurate and it doesn't penalize a teams shot accuracy when the measurement is possession.

Why do you feel SOG would be a more accurate measure? Personally I don't love either as a measure of possession, but if you're going to use shots, shot attempts seems to make more sense.

 

[Kurtz]

Why do you feel SOG would be a more accurate measure?

In terms of possession ratio, I would guess that SOG numbers would give you nearly the same result as the CORSI ratio.

But in terms of overall correlation to winning, I think the old SOG stat is better.

Consider this. CORSI incorporates blocked shots and missed shots. But the vast majority of blocked shots and missed shots, I would speculate, take place from the point, or in any case, far away from the net. Meaning they are very low % shots to begin with.

SOG stat filters out some of these low quality garbage shots that CORSI would leave in, thus making it a more reliable variable.

 

[Gallagbi]

In terms of possession ratio, I would guess that SOG numbers would give you nearly the same result as the CORSI ratio.

But in terms of overall correlation to winning, I think the old SOG stat is better.

Consider this. CORSI incorporates blocked shots and missed shots. But the vast majority of blocked shots and missed shots, I would speculate, take place from the point, or in any case, far away from the net. Meaning they are very low % shots to begin with.

SOG stat filters out some of these low quality garbage shots that CORSI would leave in, thus making it a more reliable variable.

I think the issue is you're looking at Corsi as a win predictor instead of a possession measurement.

 

[Cap'n Flavour]

The whole point of Fenwick is that it omits blocked shots but includes missed shots. If you're completely convinced that blocked shots shouldn't count as scoring chances or indications of possession then use Fenwick instead.

Also, y'know, there's that fact that Fenwick correlates better with success than SOG or Corsi.

 

[ALEXJD93]

This was posted on the main board GDT 2 nights ago, I found it really interesting

http://www.extraskater.com/reports/team-game-data?sort=corsi_pct&sit=all

16 or the 18 most dominating games Corsi wise, the team who dominated lost.

Switch it to score close and look at the results...

http://www.extraskater.com/reports/team-game-data?season=2013&sit=5v5close&sort=corsi_pct