Hockey Analytics: What Does It Mean For Cornell?

[Trotsky]

There are people who say ... "so-and-so has a Corsi of X, he's terrible" - or "he's great" - and use it to prove one player is better than another. I see that all the time.

Those people are idiots.  I don't think the people here are idiots.

[adamw]

There are people who say ... "so-and-so has a Corsi of X, he's terrible" - or "he's great" - and use it to prove one player is better than another. I see that all the time.

Those people are idiots.  I don't think the people here are idiots.

I think I made it clear I wasn't referring to anyone here

[Josh '99]

Even WAR is not a be all and end all of value, which is why there are 2 types of WAR for non-pitchers and they're still most useful only when comparing players who play the same position.

Assuming you're talking about fWAR and rWAR, there are two types because two different entities set out to create a metric that "attempts to encapsulate a player's total value to their team in one stat" (in other words, it IS supposed to be a be all and end all, it's just imperfect at this point) and went about it somewhat differently.

[Robb]

Also, hockey's fluidity is much harder to model than baseball's discrete events.

In hockey (and other similar sports, such as soccer, lacrosse, field hockey, and even basketball) offense and defense are also much more "integrated" than they are in baseball.  Trying more home run passes in hockey necessarily means your forwards are not backchecking as effectively, and vice versa - if your forwards are down low, they're not going to be in a position for receiving clean breakout passes at speed.  Pinching the defensemen helps the offense but hurts the defense, etc.  In baseball, swinging for the fences has no effect on your ability to turn a double play later on, with the possible artificial linkage of salary limitations (you may not be able to afford golden glove defenders who are also fantastic hitters).  Being good at hockey requires a much more delicate balance between offense and defense, because they are tightly integrated into a single game.

[css228]

There are people who say ... "so-and-so has a Corsi of X, he's terrible" - or "he's great" - and use it to prove one player is better than another. I see that all the time.

Absolutely, the numbers are irrelevant on an individual level without looking at zone starts, quality of competition, and quality of teammates, but those can all be factored into different variants of Corsi. I'm not saying Sean Couturier is great because he has a good Corsi, I'm saying he's great because last year he had a great Corsi despite playing the the best competition available with nearly 60% defensive zone starts and four of his five most common linemates being Max Talbot, Mike Knuble, Zac Rinaldo, and Ruslan Fedetenko. Justin Williams is really nice player, but he's only leading the NHL in Corsi % because his linemates are Dustin Brown and Anze Kopitar.

[adamw]

Absolutely, the numbers are irrelevant on an individual level without looking at zone starts, quality of competition, and quality of teammates, but those can all be factored into different variants of Corsi. I'm not saying Sean Couturier is great because he has a good Corsi, I'm saying he's great because last year he had a great Corsi despite playing the the best competition available with nearly 60% defensive zone starts and four of his five most common linemates being Max Talbot, Mike Knuble, Zac Rinaldo, and Ruslan Fedetenko. Justin Williams is really nice player, but he's only leading the NHL in Corsi % because his linemates are Dustin Brown and Anze Kopitar.

No doubt. But can't the same be said for plus-minus? Which is also a more direct correlation to actual goals? You can make all sorts of +/- adjustments to better account of those things too, and in the end, have something more directly useful. Yet Corsi-ites poo-poo +/- as kind of a joke.

[css228]

Even WAR is not a be all and end all of value, which is why there are 2 types of WAR for non-pitchers and they're still most useful only when comparing players who play the same position.

Assuming you're talking about fWAR and rWAR, there are two types because two different entities set out to create a metric that "attempts to encapsulate a player's total value to their team in one stat" (in other words, it IS supposed to be a be all and end all, it's just imperfect at this point) and went about it somewhat differently.

Yeah I guess that one is, but it still can't really be compared across positions, even with the positional adjustments. Not to mention Pitcher WAR and regular WAR are calculated so differently that you can't really compare the value of a pitcher to an everyday player.

[adamw]

I'm not saying Sean Couturier is great because he has a good Corsi, I'm saying he's great because last year he had a great Corsi despite playing the the best competition available with nearly 60% defensive zone starts and four of his five most common linemates being Max Talbot, Mike Knuble, Zac Rinaldo, and Ruslan Fedetenko.

I think he's a good player just based on what I've seen - but aren't you basing the "competition" factor on opponents' Corsi, which just makes it self-referential?  That only works if you assume Corsi is a measure of a quality in the first place.

[Jim Hyla]

It seems to me that the key to a good statistic is how long does it take to show a meaningful difference. If it takes an 80 game season to show a marginal difference between 2 players, then it's not going to be very useful in our game. If you have something that can show a difference in 20 games, then go for it. Anything less than that may not be better than just what a coach thinks. I'm sure that with time better stats, useful to the college game, will come out. I'm just not sure I'm going to see it in my viewing lifetime. So for now, I'll be happy just cheering.

[css228]

Absolutely, the numbers are irrelevant on an individual level without looking at zone starts, quality of competition, and quality of teammates, but those can all be factored into different variants of Corsi. I'm not saying Sean Couturier is great because he has a good Corsi, I'm saying he's great because last year he had a great Corsi despite playing the the best competition available with nearly 60% defensive zone starts and four of his five most common linemates being Max Talbot, Mike Knuble, Zac Rinaldo, and Ruslan Fedetenko. Justin Williams is really nice player, but he's only leading the NHL in Corsi % because his linemates are Dustin Brown and Anze Kopitar.

No doubt. But can't the same be said for plus-minus? Which is also a more direct correlation to actual goals? You can make all sorts of +/- adjustments to better account of those things too, and in the end, have something more directly useful. Yet Corsi-ites poo-poo +/- as kind of a joke.

It's just that the sample size on +/- is so small because goals are such a rare event. They occur on less than 10% of shots, so a player having just stepped out on a bad change that wasn't really his fault, that's going to bias the results more than a single shot attempt off of a bad change. Furthermore, shot attempts are more directly in the control of the skaters, whereas a bad goalie can really hurt a players +/-. Crazy bounces have much more effect on +/-. The sample size on goals is just too small for it to have any useful predictive value. Think of it like Defense Independent Pitching Statistics like FIP, which seek to eliminate the effects defenses have on a pitchers numbers. Hockey is more fluid, so its difficult, but at least Corsi and Fenwick attempting to do something +/- can't really do by increasing the sample size and focusing on the events skaters have the most control over. For the record, I think Fenwick is far more valuable as a predictive tool than Corsi, because shot-blocking is a skill representative of good defensive positioning.

[adamw]

Right ... I'm just saying that plus-minus directly counts a result that has to do with winning and losing ... whereas the connection of Fenwick/Corsi to winning and losing is more tenuous and indirect - however much it may correlate. So that would seem to offset.

Despite the correlation, it's hard to measure the effect a shot has on creating an actual goal. There's so many other factors that play into scoring a goal - quality of goalie, where the shot was taken from, luck, and so on.  Whereas with baseball stats, you can directly measure the increase in likelihood of scoring a run based upon whether you got a single, double, out, etc...

[css228]

I'm not saying Sean Couturier is great because he has a good Corsi, I'm saying he's great because last year he had a great Corsi despite playing the the best competition available with nearly 60% defensive zone starts and four of his five most common linemates being Max Talbot, Mike Knuble, Zac Rinaldo, and Ruslan Fedetenko.

I think he's a good player just based on what I've seen - but aren't you basing the "competition" factor on opponents' Corsi, which just makes it self-referential?  That only works if you assume Corsi is a measure of a quality in the first place.

It's a measure of possession ability rather than quality. I think that Corsi describes what is occurring on the ice when a player is out there. Corsi doesn't really tell me that Couturier is valuable as much as it tells me that when he is on the ice, the Flyers outshoot their opponents more often than not. What QOC tells me is whether or not the players he's going against tend to outshoot their competition when their on the ice. So really what I should say we've learned is that when Couturier is on the ice, the Flyers tend to get the majority of the shots, despite the fact he plays against competition that usually gets the majority of the shots when they are on the ice, and the fact that he is playing with players who do not tend to outshoot their opponents when they are not playing with Couturier. In other words, not matter who Couturier is against, the puck tends to end up in the offensive zone, even though he plays against competition that tends to accomplish that themselves. This is a much more measured statement than what I said before, and I think it lays out the logic of my thought process on QOC. You can imply whatever amount of value you like from that, but I happen to believe that matters.

[css228]

There's so many other factors that play into scoring a goal - quality of goalie, where the shot was taken from, luck, and so on.

Which is why +/- is such a bad statistic. On a team level sure, goal differential matters, but its like goalie wins, it accounts for a lot of factors which are completely out of a players control and due to chance. What you seem to be taking issue with overall is the correlation between puck possession and winning via goal scoring. I think you're right that probably does need to be explored more deeply, though since I believe a lot of hockey comes down to random chance and the puck sitting right in any given game, I tend to think about good process leading to good long term results. When so much is dependent on chance, the team with more balls in the lottery draw has better odds of winning. It does not mean they will win every time, because good process doesn't guarantee good results. However, even if there is reason to believe that maybe the correlation isn't causal, that doesn't mean that +/- is in anyway a stat that doesn't belong on the scrap heap along with saves, pitcher wins/losses, and goalie wins/losses.

Ultimately, I think this really comes down to whether or not you believe that a quality shot and scoring is a consistently reproducible and significant phenomenon across the team from game to game, in the sense that a team can play a certain why that they may take fewer shots, but almost all of their shots are quality. I've seen no evidence to believe that it is, at least on the team level.

[KGR11]

Here is the main difference between OPS and Corsi.  OPS is a direct building block to scoring runs.  You would never tell a baseball player "get on base less" or "please hit home runs only to this spot, or at this time".  Whereas with shots ... you can't say that the solution to improving your game is simply shooting the puck more.  That may or may not be the case, but you can't know that.

Really nit-picking here, but the purpose of the sacrifice bunt is to give up trying to get on base to move base runners over.

[css228]

Here is the main difference between OPS and Corsi.  OPS is a direct building block to scoring runs.  You would never tell a baseball player "get on base less" or "please hit home runs only to this spot, or at this time".  Whereas with shots ... you can't say that the solution to improving your game is simply shooting the puck more.  That may or may not be the case, but you can't know that.

Really nit-picking here, but the purpose of the sacrifice bunt is to give up trying to get on base to move base runners over.

And the sacrifice bunt is a stupid play which lowers run expectancy. Years of studies have consistently shown that you can expect to score more runs per inning with no outs and a runner on first than 1 out and a runner on 2nd. There may be some situations with particular hitters when a sacrifice bunt might increase run expectancy (e.g. pitchers), but overall its dumb and shouldn't be done.