Is corsi real?

[MessierII]

For whatever reason a huge chunk of the analytics community chose corsi as it’s hill to die on when the stat is outdated and as the sample size has increased with time it’s shown to be less predictive even as a team stat.

 

[MessierII]

Trying to emulate anything the 80s Oilers did is kind of pointless since goaltending was pretty atrocious until the early 90s. You could take ok open shots and you could come away with a goal pretty often, where a goalie saves it 95 percent of the time today.

If we could transport Al Montoya back to the 80s, we’d think he was the greatest goalie ever.

No. Give Al Montoya tiny equipment that weighs 40x more and you think he has the same numbers? The 80’s was the golden age of scoring stats toddy’s era is the golden age of goalie stats. Tukka Rask has the best save % of all time and he’s not in anyones convo regarding the best goalie of all time.

 

[Doctor No]

a huge chunk of the analytics community chose corsi as it’s hill to die on

If you truly believe this, then I'd recommend finding a new cohort of analytics people to follow.

Everyone who matters has moved past raw Corsi.

 

[Setec Astronomy]

No. Give Al Montoya tiny equipment that weighs 40x more and you think he has the same numbers? The 80’s was the golden age of scoring stats toddy’s era is the golden age of goalie stats. Tukka Rask has the best save % of all time and he’s not in anyones convo regarding the best goalie of all time.

It has very little to do with the equipment. The goalies are bigger, stronger and faster than they were in the 80s, by a lot. The skaters are bigger and faster too, but it cancels each other out so far as defensemen and forwards.

In the live puck era you had goalies win the Vezina with save percentages that would get you sent to the ECHL today. It’s not like the forwards had better shots than forwards do now, in fact, I would venture a guess that it’s the opposite.

 

[lomiller1]

For whatever reason a huge chunk of the analytics community chose corsi as it’s hill to die on when the stat is outdated and as the sample size has increased with time it’s shown to be less predictive even as a team stat.

This is false. Score adjusted Corsi remains the most predictive publicly available stat.

 

[MessierII]

It has very little to do with the equipment. The goalies are bigger, stronger and faster than they were in the 80s, by a lot. The skaters are bigger and faster too, but it cancels each other out so far as defensemen and forwards.

In the live puck era you had goalies win the Vezina with save percentages that would get you sent to the ECHL today. It’s not like the forwards had better shots than forwards do now, in fact, I would venture a guess that it’s the opposite.

So your saying there has been a huge jump in human evolution since the 1980’s? Goalies are bigger that’s part of it for sure. But equipment is far and away the biggest reason save %’s are higher. Not a jump in human performance.

 

[MessierII]

This is false. Score adjusted Corsi remains the most predictive publicly available stat.

No. Goal based stats are better in every way.

Goal-based Metrics Better Than Shot-based Metrics at Predicting Hockey Success

 

[Setec Astronomy]

So your saying there has been a huge jump in human evolution since the 1980’s? Goalies are bigger that’s part of it for sure. But equipment is far and away the biggest reason save %’s are higher. Not a jump in human performance.

Not human evolution, but nutrition and conditioning absolutely. Even without steroids, the players are bigger, stronger and faster than they were in the 80s in pretty much every sport.

I don’t have the inclination to look it up, but there was a picture of Cory Schneider next to Chico Resch, who was a decent goalie in his day. Cory towers over him. Just look at the way the goalies play as well.

They’ve tried every trick in the book to get more scoring and it has only ticked up a little. If you just look at save percentages now versus the 80s, you can’t tell me that defensive schemes were just so awful that breakaways were almost routine or no one knew how to defend power plays or that Wayne Gretzky’s shot was worlds better than McDavid, Crosby or Ovechkin. Occam’s razor tells you if the goalies are saving a much bigger percentage of shots than their precedessors, then it’s the goalies that are actually better.

 

[lomiller1]

No. Goal based stats are better in every way.

Goal-based Metrics Better Than Shot-based Metrics at Predicting Hockey Success

Nope. Contrary to what it’s title claims The paper is a not evaluating predictiveness, or at least not doing so correctly.

To qualify as predictive you acutely need to predict something that is not in the data you used to develop your model. This paper does not do this, instead it correlates raw Corsi with win/loss. As such it’s just re-hashing ground that is already well covered.

What happens when you build a predictive model and run it against data not in the set used to develop the model Cosi does a better job of predicting both goals and wins in that out of set data.

 

[MessierII]

Not human evolution, but nutrition and conditioning absolutely. Even without steroids, the players are bigger, stronger and faster than they were in the 80s in pretty much every sport.

I don’t have the inclination to look it up, but there was a picture of Cory Schneider next to Chico Resch, who was a decent goalie in his day. Cory towers over him. Just look at the way the goalies play as well.

They’ve tried every trick in the book to get more scoring and it has only ticked up a little. If you just look at save percentages now versus the 80s, you can’t tell me that defensive schemes were just so awful that breakaways were almost routine or no one knew how to defend power plays or that Wayne Gretzky’s shot was worlds better than McDavid, Crosby or Ovechkin. Occam’s razor tells you if the goalies are saving a much bigger percentage of shots than their precedessors, then it’s the goalies that are actually better.

Or it’s because the goalies are physically occupying more of the net and PP numbers are down.

Don’t get me wrong there’s some truth to what your saying but put Vaselesky in 80’s gear you think he still has a .930 save %? No chance.

Save % was stablized in the later half of the 80’s after the merger of the WHA etc and the influx of Nordic hockey players to around .880 between 1986 and 1990. It took a huge bump once after Soviet Union fell and influx of Eastern Europeon talent in 1990-91 to about .886 where it stayed until 1993-94 when the big equipment came in. 1993 is considered the last high scoring year it was the year rookie Selanne scored 76 goals. The average save % bumped a full 1% in one year with the equipment change from .885 to .895 and rose steadily as the trap became more prevalent through the dead puck era up until it peaked at .911 in 03-04 before the lockout just slightly lower than it is today. With the lockout they opened up the game cracking down on obstruction and save % plummeted in 05-06 dropping a full 1% to .901 largest change since the big equipment. Since then they’ve loosened up the rules and obstruction has crept back in and pp’s have dropped and save % is back up to .913 a bit above where it was in 03-04.

To sum up. The big equipment and pp numbers have resulted in the largest change in save %. Goalies in 06-07 actually stopped slightly less pucks than goalies in 1997-98. Lots of those goalies in 06-07 are playing today and lots of goalies from 1997 were playing in the 80’s.

 

[MessierII]

Nope. Contrary to what it’s title claims The paper is a not evaluating predictiveness, or at least not doing so correctly.

To qualify as predictive you acutely need to predict something that is not in the data you used to develop your model. This paper does not do this, instead it correlates raw Corsi with win/loss. As such it’s just re-hashing ground that is already well covered.

What happens when you build a predictive model and run it against data not in the set used to develop the model Cosi does a better job of predicting both goals and wins in that out of set data.

So because he didn’t make a bold prediction of the future years to come the results of his experiment which took data from and results from a 10 year sample size are thrown out the door? Is that what your saying? There’s only one thing anyone should care about predicting and that’s wins.

 

[lomiller1]

So because he didn’t make a bold prediction of the future years to come the results of his experiment which took data from and results from a 10 year sample size are thrown out the door? Is that what your saying? There’s only one thing anyone should care about predicting and that’s wins.

To call some predictive you actually have to show how well you can predict something, which he doesn’t do.

You don’t need 10 years of data to know that goal differential is the best indicator of a teams current record. It’s also been show convincingly that score effects case a slight inverse relationship between CF% and win%.

The problem with either is that you don’t need to predict who has the best record, when you can just go look up the record. What you want advanced stats for is to figure out who is mostly likely to win going forward and which players are most likely to help contribute to that. This is where predictive modeling comes in, and it’s WELL documented that CF% outperforms GF% for predicting bot just wins but future GF% as well.

 

[lomiller1]

Good article explain how different statistics are evaluated and what makes one better than another.

Behind the Numbers: What Makes a Stat Good

 

[Setec Astronomy]

Or it’s because the goalies are physically occupying more of the net and PP numbers are down.

Don’t get me wrong there’s some truth to what your saying but put Vaselesky in 80’s gear you think he still has a .930 save %? No chance.

Save % was stablized in the later half of the 80’s after the merger of the WHA etc and the influx of Nordic hockey players to around .880 between 1986 and 1990. It took a huge bump once after Soviet Union fell and influx of Eastern Europeon talent in 1990-91 to about .886 where it stayed until 1993-94 when the big equipment came in. 1993 is considered the last high scoring year it was the year rookie Selanne scored 76 goals. The average save % bumped a full 1% in one year with the equipment change from .885 to .895 and rose steadily as the trap became more prevalent through the dead puck era up until it peaked at .911 in 03-04 before the lockout just slightly lower than it is today. With the lockout they opened up the game cracking down on obstruction and save % plummeted in 05-06 dropping a full 1% to .901 largest change since the big equipment. Since then they’ve loosened up the rules and obstruction has crept back in and pp’s have dropped and save % is back up to .913 a bit above where it was in 03-04.

To sum up. The big equipment and pp numbers have resulted in the largest change in save %. Goalies in 06-07 actually stopped slightly less pucks than goalies in 1997-98. Lots of those goalies in 06-07 are playing today and lots of goalies from 1997 were playing in the 80’s.

First, obstruction, even if it crept back a little, is nowhere near what it was prior to the lockout. I don’t want to go searching for it, but there’s a clip from game 1 of the 2000 SCF Final where Brodeur makes a really nice save on I believe Hull. If you look at Hull going through the neutral zone, I think it’s Claude Lemeuix who is literally grabbing his arm. Clemente or Thorne don’t even mention it during the replay that’s how accepted it was to do that. There has been nothing like that since the 2004-05 lock out ended.

Also, since the lockout they’ve done virtually everything they can to open up scoring, from getting rid of the redline, the institution of the trapezoid, to making the neutral zone smaller, creating more penalties like puck over the glass, and pretty much doing away with open ice hits. If you had these rules in the 80s, Gretzky and Mario probably would have scored 120 goals in a season at some point. Yet we still get only three one hundred point scorers in a season if there’s even one.

Not just that you have 31 teams now and pretty much the guy who can do nothing but fight is pretty much extinct. In other words there are more goalies than there ever were and more skilled players then there ever were, yet we’re still seeing very high save percentages almost across the board.

I’m sure equipment has something to do with it, but the goalies themselves are just superior atheletes than they used to be.

 

[MessierII]

To call some predictive you actually have to show how well you can predict something, which he doesn’t do.

You don’t need 10 years of data to know that goal differential is the best indicator of a teams current record. It’s also been show convincingly that score effects case a slight inverse relationship between CF% and win%.

The problem with either is that you don’t need to predict who has the best record, when you can just go look up the record. What you want advanced stats for is to figure out who is mostly likely to win going forward and which players are most likely to help contribute to that. This is where predictive modeling comes in, and it’s WELL documented that CF% outperforms GF% for predicting bot just wins but future GF% as well.

I guess you just straight up didn’t read the article. At least not carefully. He took goal stats from 05-06 to 2014-15 and put them against long term shot stats from the same time frame to predict the outcome of 2015-16. Goal stats were better by a large margin.

 

[MessierII]

First, obstruction, even if it crept back a little, is nowhere near what it was prior to the lockout. I don’t want to go searching for it, but there’s a clip from game 1 of the 2000 SCF Final where Brodeur makes a really nice save on I believe Hull. If you look at Hull going through the neutral zone, I think it’s Claude Lemeuix who is literally grabbing his arm. Clemente or Thorne don’t even mention it during the replay that’s how accepted it was to do that. There has been nothing like that since the 2004-05 lock out ended.

Also, since the lockout they’ve done virtually everything they can to open up scoring, from getting rid of the redline, the institution of the trapezoid, to making the neutral zone smaller, creating more penalties like puck over the glass, and pretty much doing away with open ice hits. If you had these rules in the 80s, Gretzky and Mario probably would have scored 120 goals in a season at some point. Yet we still get only three one hundred point scorers in a season if there’s even one.

Not just that you have 31 teams now and pretty much the guy who can do nothing but fight is pretty much extinct. In other words there are more goalies than there ever were and more skilled players then there ever were, yet we’re still seeing very high save percentages almost across the board.

I’m sure equipment has something to do with it, but the goalies themselves are just superior atheletes than they used to be.

Obstruction is worse than it was before the lockout. Last year had lowest number of pp’s since the early 50’s. I think it’s gone up a bit this year but make no mistake the current era is on par with the peak of the dead puck era.

Like I stated earlier the goalies are bigger and better athletes. But that’s only a part of the picture. Equipment and pp numbers have accounted for the biggest sudden changes in save % historically.

 

[Setec Astronomy]

Obstruction is worse than it was before the lockout. Last year had lowest number of pp’s since the early 50’s. I think it’s gone up a bit this year but make no mistake the current era is on par with the peak of the dead puck era.

Like I stated earlier the goalies are bigger and better athletes. But that’s only a part of the picture. Equipment and pp numbers have accounted for the biggest sudden changes in save % historically.

There might be less power plays because they’re committing fewer penalties, which is one of things the rule was designed to do. You just have to watch a meaningful game from before the lockout. It isn’t even close. By the way, this was the play I was talking about. Pre lockout this was routine, you could not get away with it now, so it does happen.

 

[MessierII]

There might be less power plays because they’re committing fewer penalties, which is one of things the rule was designed to do. You just have to watch a meaningful game from before the lockout. It isn’t even close. By the way, this was the play I was talking about. Pre lockout this was routine, you could not get away with it now, so it does happen.

I suggest you watch Connor McDavid play an nhl game. What you just showed me happens to him every other shift. Sometimes multiple times on one shift.

 

[Setec Astronomy]

I suggest you watch Connor McDavid play an nhl game. What you just showed me happens to him every other shift. Sometimes multiple times on one shift.

I assure you, it’s nothing close to what you saw in the video, a hold with the arm and a hook, and no one even thought it was a penalty. Stuff like that was routine back then.

 

[MessierII]

I assure you, it’s nothing close to what you saw in the video, a hold with the arm and a hook, and no one even thought it was a penalty. Stuff like that was routine back then.

I’ve been watching hockey since 1989. Watch a Mcdavid shift. Literally any shift where he makes a rush up the neutral zone. Even Taylor Hall or Larkin. Either way it has nothing to do with the discussion. PP numbers are way down since the lockout and wouldn’t you know it save % is way up. That has a bigger effect than skill of the goalies. Unless you think there’s been another genetic jump since 2007.

 

[Michael Farkas]

Coaching / short shift game has limited scoring. Goaltending evolved slower than the other positions coming out of the proliferation of pro teams (6 in 1966 vs like 28 in 1972 or whatever it was). As opposed to other times in history where goaltending out-performed other positions in terms of evolution (1920's, 1950's and 60's, etc.). As goaltending and coaching caught up and the talent pool evened out across positions, scoring went down...that's the sign of a stable, well-maintained league, generally speaking...the lesser the overall talent a league has (within reason), the higher scoring becomes...which sounds counter-intuitive at a glance, but don't confuse talent with stickhandling...

 

[LEAFANFORLIFE23]

just like all advanced stats Corsi is not real

 

[Doctor No]

just like all advanced stats Corsi is not real

Of course it's real.

 

[Canadiens1958]

just like all advanced stats Corsi is not real

One of the first lessons in a Stats 101 course is that very few stats are "real". Supported with an example that a typical family fitting certain criteria has 2.3 children. Nobody has ever seen 0.3 of a child, does not exist. However the stat is extremely representative of families in the group, revealing birth rates and other key data.

 

[Canadiens1958]

Coaching / short shift game has limited scoring. Goaltending evolved slower than the other positions coming out of the proliferation of pro teams (6 in 1966 vs like 28 in 1972 or whatever it was). As opposed to other times in history where goaltending out-performed other positions in terms of evolution (1920's, 1950's and 60's, etc.). As goaltending and coaching caught up and the talent pool evened out across positions, scoring went down...that's the sign of a stable, well-maintained league, generally speaking...the lesser the overall talent a league has (within reason), the higher scoring becomes...which sounds counter-intuitive at a glance, but don't confuse talent with stickhandling...

Or don't confuse entertainment with talent.