What Corsi really translates to (in numbers that are easy to grasp)

[Machinehead]

I know he's not a defenseman but Jesper Fast is a good example.

One of the most painful to watch players I've ever seen offensively, and he's still 6th on the Rangers in points/60, just because he's crazy good defensively and the Rangers are in control when he's on the ice.

 

[Korpse]

Again 22 out of the last 24 Cup winners were top 10 in shot differential. Not sure what part of that people can't grasp.

because shot differential isn't the only thing in play.

 

[Bourdon101]

Again 22 out of the last 24 Cup winners were top 10 in shot differential. Not sure what part of that people can't grasp.

Because that's just dumb?

Statistics are based on the law of large numbers, what you are doing is a sophism. You basically shrink down to a small enough sample that will confirm your theory. Thats now how statistics work: you actually need to do some hypothesis testing, and that involves taking a large sample. Go ahead and calculate the correlation of corsi and wins in playoffs over enough years and then you might have an argument but hint: you won't because the correlation is negligable.

 

[Machinehead]

because shot differential isn't the only thing in play.

Obviously not. Never said it was.

But allow me to repeat: corsi is one of many things you HAVE TO be good at to win.

Because that's just dumb?

Statistics are based on the law of large numbers, what you are doing is a sophism. You basically shrink down to a small enough sample that will confirm your theory. Thats now how statistics work: you actually need to do some hypothesis testing, and that involves taking a large sample. Go ahead and calculate the correlation of corsi and wins in playoffs over enough years and then you might have an argument but hint: you won't because the correlation is negligable.

It's 24 seasons. How big do you want the sample to be?

 

[Sojourn]

because shot differential isn't the only thing in play.

Well, there's also the size of the league. Top 10 out of 30 teams. Sounds a little less significant when you're just talking about the top third of the league. And, apparently, even that "rule" doesn't always apply.

And then we're dismissing 2 out of 24. 24 is a pretty small number already. What if you extrapolated that to a larger number? 20 out of 240? Seems a bit more noteworthy now, doesn't it. Of course, we shouldn't really do that, but let's not pretend 24 is a large number.

 

[RageQuit77]

I get the point of scientific handling of advances statistics in their scope of use, but I'll never get why 'hard simple stats' like Goals and Assists on the score sheet are sometimes ignored (or at least it looks like that) when some stat essentially describing this or that about a groups' play is reflected upon to single player showing 0.31% worse then some other player in completely different team, role, TOI, linemates, playstyle, age and experience etc.

Corsi sucks dramatically most of cases it is placed against an individual contributions on the score sheet (according HFB experiences).

Team A can outshoot Team B in shooting differentials for every lineups. Team B scores goals and win. Where you need Corsis in your determination what team is better?

Reflect straightforwardly a group stat (absolute or relative) to an individual player and you are lost. Hopelessly.

 

[Machinehead]

Well, there's also the size of the league. Top 10 out of 30 teams. Sounds a little less significant when you're just talking about the top third of the league. And, apparently, even that "rule" doesn't always apply.

I'm not disagreeing with you. All I'm saying and all I've ever said is that it's highly advisable to be in the top third in CF% if your goal is to win the Cup. Is that so shocking?

 

[bobbyb2009]

We simply don't measure it that way. That's your opinion on what a zone start should be, but corsiRelOlJase is not a stat.

There was a time when "we", the scientific community all agreed the world was flat...

Over time, "we" have been proved wrong repeatedly, no matter how smart and thorough "we" are.

 

[Sojourn]

I'm not disagreeing with you. All I'm saying and all I've ever said is that it's highly advisable to be in the top third in CF% if your goal is to win the Cup. Is that so shocking?

I don't have much of a problem with team Corsi. When you're a better team, you typically have the puck more, and are shooting more.

I do have a problem with individual Corsi, and the way it's used by some. There are more variables to consider with individual players, with a smaller margin, and less to balance out the results. Thats my issue. You try to apply the same rules to individuals that you do to teams. A glaring example of this was your attempt to argue McIlrath was, and would be, better than Fowler. Am I beating a dead horse? You bet. Is it an example of my issue with the way some people try to use these statistics? Absolutely. I've rarely seen a more glaring example of an individual misusing statistics. And that wasn't the statistics telling you that. It was how you interpreted them.

I also have a very big problem with your "discussion closed" argument. You should be trying to figure out why the numbers are saying things. I don't see that from you at all. I see you positing that the numbers tell you what you need to know, and that's that.

 

[Machinehead]

There was a time when "we", the scientific community all agreed the world was flat...

Over time, "we" have been proved wrong repeatedly, no matter how smart and thorough "we" are.

If 10 years down the line that's the way we measure it and he's right, I'll be the first to bump the thread.

I think his very premise is nonsense because nobody changes lines unless the puck is in the offensive zone our about to be in the offensive zone. But that's not the point.

My point is, he's arguing outside of the parameters of what the stats currently are. What he posited is not what zone starts even are. It's like saying "ok that corsi stat is interesting but what if instead of shot attempts, corsi was this instead."

Maybe down the line we'll have a new way to measure it and he'll be right, and it still wouldn't be zone starts. It would be something else. Maybe on the fly starts.

 

[Machinehead]

I don't have much of a problem with team Corsi. When you're a better team, you typically have the puck more, and are shooting more.

I do have a problem with individual Corsi, and the way it's used by some. There are more variables to consider with individual players, with a smaller margin, and less to balance out the results. Thats my issue. You try to apply the same rules to individuals that you do to teams. A glaring example of this was your attempt to argue McIlrath was, and would be, better than Fowler. Am I beating a dead horse? You bet. Is it an example of my issue with the way some people try to use these statistics? Absolutely. I've rarely seen a more glaring example of an individual misusing statistics. And that wasn't the statistics telling you that. It was how you interpreted them.

I'm not going to rehash the same argument. I'm just going to say I agree with what you say about variables and application, but the evidence against Fowler is conclusive.

 

[Bourdon101]

Obviously not. Never said it was.

But allow me to repeat: corsi is one of many things you HAVE TO be good at to win.



It's 24 seasons. How big do you want the sample to be?

Well first I'd like you to actually make a correlation between wins and corsi, not just using a different (and very flawed) argument because there's no real correlation.

Then I'd say you have to be kidding me with that statement... Hint: Look up normal approximation of binomial distribution. I'm not here to teach you how statistics work, but I'm sure you realise that a hypothesis test with a very small sample (which 24 is for sure, it technically is too small to be even accepted as a model if you read just a little about binomial densities) will have a much lower significance due to the central limit theorem and the effect of small samples on variance.

 

[Sojourn]

I'm not going to rehash the same argument. I'm just going to say I agree with what you say about variables and application, but the evidence against Fowler is conclusive.

Sure it is. In the meantime, I'll be happy with him on the team, because we're better with him.

 

[Machinehead]

Sure it is. In the meantime, I'll be happy with him on the team, because we're better with him.

Well as long as you're happy...

That's why we watch.

 

[Sojourn]

If 10 years down the line that's the way we measure it and he's right, I'll be the first to bump the thread.

I think his very premise is nonsense because nobody changes lines unless the puck is in the offensive zone our about to be in the offensive zone. But that's not the point.

My point is, he's arguing outside of the parameters of what the stats currently are. What he posited is not what zone starts even are. It's like saying "ok that corsi stat is interesting but what if instead of shot attempts, corsi was this instead."

Maybe down the line we'll have a new way to measure it and he'll be right, and it still wouldn't be zone starts. It would be something else. Maybe on the fly starts.

This is not true at all. Teams often change as they exit the D zone, or even while they have control in the D zone. Changing on the fly happens in all kinds of situations. The key being you want to change at a time it doesn't hurt the team. You're over-simplifying.

 

[RageQuit77]

I don't have much of a problem with team Corsi. When you're a better team, you typically have the puck more, and are shooting more.

I do have a problem with individual Corsi, and the way it's used by some. There are more variables to consider with individual players, with a smaller margin, and less to balance out the results. Thats my issue. You try to apply the same rules to individuals that you do to teams. A glaring example of this was your attempt to argue McIlrath was, and would be, better than Fowler. Am I beating a dead horse? You bet. Is it an example of my issue with the way some people try to use these statistics? Absolutely. I've rarely seen a more glaring example of an individual misusing statistics. And that wasn't the statistics telling you that. It was how you interpreted them.

I share your sentiments.

Adv. stats can be hellish effective tool with real practical and tactical coaching uses, no doubt, some day, but certainly not if they aren't utilized proper in manner for the team and line up level.

Then, also, contemporary statistical system is not the last word inscribed to the eternal stone about hockey, and even year 4890 when 1% of all happenings in a rink aren't yet managed to transform to models and formulas, it requires only one thing that can ruin that conceptual apparatus - a Hockey player - amidst that mathematical alchemy/hodgebodge (at least when someone try explain it to the non-cultivated public).

Good luck in developing better systems of statistical analysis for hockey. O7

 

[Machinehead]

This is not true at all. Teams often change as they exit the D zone, or even while they have control in the D zone. Changing on the fly happens in all kinds of situations. The key being you want to change at a time it doesn't hurt the team. You're over-simplifying.

True but you're still going in transition. Nobody just walks off the ice mid shift in the defensive zone

 

[4thline]

The thing is, there not being enough information to say that zone starts has any significance is not the same as there not being any significance. Logic would suggest that it does matter. If you're starting 180 feet from your own net vs. 10 feet away, well, any one who has played can tell you that one situation is more likely to see a shot against your net than the other.

If the "advanced" stats community feels that it's a closed discussion, I think they are doing themselves a disservice. You shouldn't just be looking at what the numbers say. You should also ask why it is saying it, and if it makes sense. To me, that's the difference between someone who has a real understanding of numbers and how to use them, and someone who lets the numbers tell them what to think.

Fantastic post.
A lot of the stats blogosphere has great application of math, poor application of the scientific method and critical thinking. Many of the "conclusively proven" x has no impacts are based on a couple of horribly flawed circular conclusions.

 

[Elvs]

I'm not going to rehash the same argument. I'm just going to say I agree with what you say about variables and application, but the evidence against Fowler is conclusive.

Sure, as long as you keep telling yourself that all shot attempts are equal. That all goaltending is equal. That special teams don't matter. That every goal is just as important as the other. That defense partners don't matter. Etc etc.

 

[Creativero]

Fantastic post.
A lot of the stats blogosphere has great application of math, poor application of the scientific method and critical thinking. Many of the "conclusively proven" x has no impacts are based on a couple of horribly flawed circular conclusions.

This is my problem with most advanced stats too.

 

[Sojourn]

True but you're still going in transition. Nobody just walks off the ice mid shift in the defensive zone

No, of course not, but how many times do you see bad changes? Or a player stepping onto the ice and then a goal is scored against their team? The game is so friggin fast, and a 30-45 shift means players are coming on and off the ice all the time.

I'd love to see a statistic that included changes. I'm pro statistics, believe it or not. Information is good. I don't come to the same conclusions that you do, and to each his own, but I want more information. Always. I want an actual possession statistic. Not because I don't think Corsi gives us some information on that(it obviously does), but because I'd love to see where they diverge, and then I want to know why they diverge. What types of things show up with individuals, with teams, game to game, long-term...

Cause and effect isn't enough. I want to know why. If there are anomalies, why are they there? Is it something you can reproduce? And if it is, what does that tell you? It's the statistical version of taking apart a phone so I can see how it works. I love that ****.

 

[Machinehead]

Sure, as long as you keep telling yourself that all shot attempts are equal. That all goaltending is equal. That special teams don't matter. That every goal is just as important as the other. That defense partners don't matter. Etc etc.

Never said any of this. These threads always devolve into madness because they're one assumption after another.

 

[Machinehead]

No, of course not, but how many times do you see bad changes? Or a player stepping onto the ice and then a goal is scored against their team? The game is so friggin fast, and a 30-45 shift means players are coming on and off the ice all the time.

I'd love to see a statistic that included changes. I'm pro statistics, believe it or not. Information is good. I don't come to the same conclusions that you do, and to each his own, but I want more information. Always. I want an actual possession statistic. Not because I don't think Corsi gives us some information on that(it obviously does), but because I'd love to see where they diverge, and then I want to know why they diverge. What types of things show up with individuals, with teams, game to game, long-term...

Cause and effect isn't enough. I want to know why. If there are anomalies, why are they there? Is it something you can reproduce? And if it is, what does that tell you? It's the statistical version of taking apart a phone so I can see how it works. I love that ****.

Agree with everything here.

My point is that a "zone start" as it is written by the NHL itself, does not affect corsi in aggregate.

There's certainly aspects of the zonal game we haven't uncovered yet and I'm very interested in and open to those effects.

Zone starts as defined right now in 2016 aren't doing that. What that poster proposed isn't even in the book yet and wasn't part of the original debate which is about zone starts as we know them.

Saying "but what if there's some other factor we don't know about?" is fine, but it doesn't change the argument as far as current statistics.

 

[4thline]

Always. I want an actual possession statistic. Not because I don't think Corsi gives us some information on that, but because I'd love to see where they diverge, and then I want to know why they diverge. What types of things show up with individuals, with teams, game to game, long-term...

You mean the kinds of things the corsiples have deemed no impactful because they get confounded out of the picture when you look at mountains of ****** uncontrolled data rather than trying to analyze static conditions in a close to experimental settings as the NHL allows?


Allelujah, amen! The whole "actual time of possession doesn't matter because corsi is a close enough proxy and actually matters more" was the simply the laziest self-serving ********. It measures something that while related is quite different, and hybriding it with corsi would provide way more useful metrics.

50CF% player meh. useless.

50CF% with descriptors on their ozone/dzone time split and the in zone corsi rates for and against, thats something usable.

 

[SladeWilson23]

Again 22 out of the last 24 Cup winners were top 10 in shot differential. Not sure what part of that people can't grasp.

1/3 of the league is a very big portion of the league. Not to mention when you look at CF% for those middle of the pack teams, some are separated by less than 1%.

Obviously not. Never said it was.

But allow me to repeat: corsi is one of many things you HAVE TO be good at to win.

It's 24 seasons. How big do you want the sample to be?

No, you got it all wrong. If you're a good team, you're gonna have a good Corsi. GM's don't build their teams around Corsi. Good teams end up having good a good Corsi because they're good teams.

Well, there's also the size of the league. Top 10 out of 30 teams. Sounds a little less significant when you're just talking about the top third of the league. And, apparently, even that "rule" doesn't always apply.

And then we're dismissing 2 out of 24. 24 is a pretty small number already. What if you extrapolated that to a larger number? 20 out of 240? Seems a bit more noteworthy now, doesn't it. Of course, we shouldn't really do that, but let's not pretend 24 is a large number.

Very well said.

I'm not disagreeing with you. All I'm saying and all I've ever said is that it's highly advisable to be in the top third in CF% if your goal is to win the Cup. Is that so shocking?

Again, teams don't aim for good Corsi. Teams simply end up with good Corsi because they're good teams.

I'm not going to rehash the same argument. I'm just going to say I agree with what you say about variables and application, but the evidence against Fowler is conclusive.

No it's not. Fowler is a great player.

Sure, as long as you keep telling yourself that all shot attempts are equal. That all goaltending is equal. That special teams don't matter. That every goal is just as important as the other. That defense partners don't matter. Etc etc.

One thing I will say is that taking a lot of low danger shots does usually lead to more high danger chances.