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

[Machinehead]

I dont think extremely rare is the proper way to describe this. Your o-zone and d-zone are roughly 25% of your starts. That is a quarter of your shifts that are either starting in the o-zone with a 10% bump in CF% or the d-zone with a 10% reduction in CF%.

The difference in O vs D zone on 25% of the shifts should be enough to sway a players CF% 2-4%. a 4% swing in a players CF% according to "the community" is the difference between Tanner Glass and Patrice Bergeron (hyperbolic example).

They cancel out.

Your OZ and DZ shifts might each account for 25% of your shifts, but if you had 10 OZ shifts and 9 DZ shifts, you've had 1 shift of consequence.

But isn't this quite a banal statement? A good team usually has better shot differential than its opponent most of the time. And more goals. I think it's quite obvious and not all that revealing...

It's meant to be banal and it is quite obvious, so why does everyone freak out about it?

 

[SladeWilson23]

It's meant to be banal and it is quite obvious, so why does everyone freak out about it?

Because it's not a causation of winning cups. Good Corsi is the effect of having good teams.

 

[aufheben]

But isn't this quite a banal statement? A good team usually has better shot differential than its opponent most of the time. And more goals. I think it's quite obvious and not all that revealing...

It can be a better indicator than regular season record. Teams go through coaching chances, injuries, player acquisitions, etc.

 

[Machinehead]

Because it's not a causation of winning cups. Good Corsi is the effect of having good teams.

Never said it was.

But if for example, you're good at everything else, and 27th in CF%, you're not winning the Cup. Not mentioning any names.

 

[Sojourn]

They cancel out.

Your OZ and DZ shifts might each account for 25% of your shifts, but if you had 10 OZ shifts and 9 DZ shifts, you've had 1 shift of consequence.

That kind of contradicts your statement that zone starts are irrelevant. You're actually saying they are relevant, and could add up to a difference over time. Just because it balances out for some players, or even most players, doesn't mean it's not a relevant part. You can't just dismiss such a large portion of a player's shifts just because you assume they cancel out and don't impact the results. That's a big chunk that you're kind of handwaving aside.

As the poster you responded to pointed out, the difference between a perceived good Corsi player and a bad Corsi player is a couple of percentage points.

Edit: Help me out here. Over the course of a single game, how many events can a player expect to be on the ice for if they play 20 minutes?

 

[Sam Spade]

Go to a bookie and bet the first round of the playoffs. Take all of higher CF% teams in a series from the regular season to win.

You will make a lot of money as the only "lower seed" to win was the Sharks (#9) over the Kings (#1).

 

[Machinehead]

That kind of contradicts your statement that zone starts are irrelevant. You're actually saying they are relevant, and could add up to a difference over time. Just because it balances out for some players, or even most players, doesn't mean it's not a relevant part. You can't just dismiss such a large portion of a player's shifts just because you assume they cancel out and don't impact the results. That's a big chunk that you're kind of handwaving aside.

As the poster you responded to pointed out, the difference between a perceived good Corsi player and a bad Corsi player is a couple of percentage points.

I'm not handwaving anything. It's been studied and the conclusion is that the effect is 0.5-1.0%. That's nothing.

I'm just trying to explain why. You guys seem to think I made all this **** up. I haven't. Lengthy studies have concluded that zone starts do next to nothing statistically. I'm just trying to help you understand some of the counterintuitive aspects of it.

Edit: Help me out here. Over the course of a single game, how many events can a player expect to be on the ice for if they play 20 minutes?

Really depends on the game but generally somewhere between 30 and 45.

 

[Korpse]

Go to a bookie and bet the first round of the playoffs. Take all of higher CF% teams in a series from the regular season to win.

You will make a lot of money as the only "lower seed" to win was the Sharks (#9) over the Kings (#1).

please don't.

the lower CF% rank had won 7/8 series in 2014/2015 if you had banked on the lower CF% winning in 2015/2016 you would have lost a lot of money.

Over the last 9 playoffs in the first round the team with the higher CF% has won as often as the team with home ice advantage, 56.9% of the time.

gamble responsibly.

 

[Sniper99]

Go to a bookie and bet the first round of the playoffs. Take all of higher CF% teams in a series from the regular season to win.

You will make a lot of money as the only "lower seed" to win was the Sharks (#9) over the Kings (#1).

No thanks, I refuse to bet with anything corsi related.

 

[SladeWilson23]

Never said it was.

You definitely treat it like it is.

But if for example, you're good at everything else, and 27th in CF%, you're not winning the Cup. Not mentioning any names.

You act as if the Rangers aren't capable of playing the possession game.

 

[Machinehead]

You act as if the Rangers aren't capable of playing the possession game.

They're not. Like, not even remotely.

 

[Addison Rae]

i don't buy the notion that zone starts don't have an impact on possession, but machinehead illustrated perfectly that you barley start your shifts in a specific zone as you're changing on the fly a lot more.

 

[Korpse]

i don't buy the notion that zone starts don't have an impact on possession, but machinehead illustrated perfectly that you barley start your shifts in a specific zone as you're changing on the fly a lot more.

though you see a lot of on the fly changes happen when a team dumps the puck and give up possession so to speak.

 

[SladeWilson23]

They're not. Like, not even remotely.

Wait, why couldn't they play a possession game?

 

[4thline]

They cancel out.

Your OZ and DZ shifts might each account for 25% of your shifts, but if you had 10 OZ shifts and 9 DZ shifts, you've had 1 shift of consequence.

It's meant to be banal and it is quite obvious, so why does everyone freak out about it?

I'm not handwaving anything. It's been studied and the conclusion is that the effect is 0.5-1.0%. That's nothing.

I'm just trying to explain why. You guys seem to think I made all this **** up. I haven't. Lengthy studies have concluded that zone starts do next to nothing statistically. I'm just trying to help you understand some of the counterintuitive aspects of it.

The statements

zone starts have no impact on corsi, conclusive, end of story

and

zone starts impact corsi but the average players usage is balanced enough that the impact is negligible in aggregate over a season

Are very different.

Because the 1st is completely blinds you to contextualizing the performance of players at the top and bottom of many teams depth charts that don't "balance out", that are facing 60/40, 70/30, even 80/20 splits.

The same applies to quality of competition, discounting a fundamental relationship because "all players face the same" is folly when work like that of Micah Blake McCurdy shows that teams do have players at the top and bottom of lineups whose competition varies greatly from baseline expectation.


If player x is facing 5-10% more of their icetime against first line opponents (with the 5-10 being cut out of the 4thline time) and seeing 5-10% more of their shifts start in the d zone thats going to hurt their numbers, to the point where a Rel ranking is largely useless unless its something huge.

 

[nyr__1994]

The statements

zone starts have no impact on corsi, conclusive, end of story

and

zone starts impact corsi but the average players usage is balanced enough that the impact is negligible in aggregate over a season

Are very different.

Because the 1st is completely blinds you to contextualizing the performance of players at the top and bottom of many teams depth charts that don't "balance out", that are facing 60/40, 70/30, even 80/20 splits.

The same applies to quality of competition, discounting a fundamental relationship because "all players face the same" is folly when work like that of Micah Blake McCurdy shows that teams do have players at the top and bottom of lineups whose competition varies greatly from baseline expectation.


If player x is facing 5-10% more of their icetime against first line opponents (with the 5-10 being cut out of the 4thline time) and seeing 5-10% more of their shifts start in the d zone thats going to hurt their numbers, to the point where a Rel ranking is largely useless unless its something huge.

I think this is what I am trying to get at. While the impacts may be small, they are definately there for a player that gets heavy O-Zone or D-Zone starts, and then combine that with either harder or easier minutes based on Quality of Competition, those small differences add up.

So a player that is a 48% CF% player is viewed as a bad player because they have a bad Corgi, but when you dig deeper you find that this player starts 70/30 split in D-zone vs O-zone starts, and is playing against 1st line competition, that 48% doesn't look so bad any more. If the same player reversed their zone starts and played against mainly third line competition, what would their CF% be? But "the Community" insists that zone starts and Quality of Competition don't matter so that 48% player still sucks...

 

[Sam Spade]

please don't.

the lower CF% rank had won 7/8 series in 2014/2015 if you had banked on the lower CF% winning in 2015/2016 you would have lost a lot of money.

Over the last 9 playoffs in the first round the team with the higher CF% has won as often as the team with home ice advantage, 56.9% of the time.

gamble responsibly.

The $1700 I put in my bank account last spring from $160 in bets tells me differently.

 

[Korpse]

The $1700 I put in my bank account last spring from $160 in bets tells me differently.

Well it's a good thing you didn't do it a year earlier

 

[Machinehead]

Wait, why couldn't they play a possession game?

Because they objectively suck at it?

 

[4thline]

I think this is what I am trying to get at. While the impacts may be small, they are definately there for a player that gets heavy O-Zone or D-Zone starts, and then combine that with either harder or easier minutes based on Quality of Competition, those small differences add up.

So a player that is a 48% CF% player is viewed as a bad player because they have a bad Corgi, but when you dig deeper you find that this player starts 70/30 split in D-zone vs O-zone starts, and is playing against 1st line competition, that 48% doesn't look so bad any more. If the same player reversed their zone starts and played against mainly third line competition, what would their CF% be? But "the Community" insists that zone starts and Quality of Competition don't matter so that 48% player still sucks...

Yup, and more and more there is tangible proof that some of the foundational assumptions surrounding "what impacts" corsi were grossly flawed, and led to flawed conclusions.

http://leafshub.com/tougher-than-the-rest/

Kinda ruins the whole "every player plays against every other player approximately equally idea. Corsi based QoC is absolute garbage, been saying it for a long time.

 

[MNNumbers]

Because that's how hockey works.

40% of all shifts have a zone start.

100% of shifts have corsi.

Sorry to jump in here, but this is a strange looking comparison to me. Here is why....

A shift is a physical thing.
A zone start is a physical thing.
"Corsi" is not a physical thing. "Corsi" is a collection of numbers used to analyze physical things.

Better wording, and the actual meaning, might be...."100% of shifts have measurable events which are used to calculate Corsi."

Concerning O-zone and D-zone starts, You have the numbers right there.. 39.xx CF on d-zone starts. 58.xx whatever on o-zone starts. 47.xx on neutral zone starts....
What doesn't someone incorporate those numbers into individual Corsi?
So, for individuals.....
Adjusted Corsi goes something like this:
% of d-zone faceoff starts * Corsi on those shifts/39.xx (or whatever the exact number is)+
% of o-zone faceoff starts * Corsi on those shifts/58.xx (or whatever the number is) +
% of n-zone faceoff starts * Corsi on those shifts/47.xx (or whatever the average is) +
% of on-the-fly starts * Corsi on those shifts/50.xx (whatever the avg number is).

That should take all of that into account, without the advanced stats enthusiast needing to say "The individual difference in Off vs Def starts is statistically insignificant, so we ignore it." Saying that sounds like an excuse. So, instead, incorporate it.

Corsi is nothing more than a way to count anyway. Just find a more precise way to count it.

On a side note, I find those %ages interesting. O-zone + D-zone < 100. N-zone <50. On-fly >50. The only way that can physically be true is if the team which wins the faceoff dumps the puck into the o-zone, and changes players before the other team has a change to do so, and then those 'new' players create a + Corsi. And, of course, that makes perfect sense. But, thinking it through like that is a good exercise in that it forces a person to ask, "What are the numbers REALLY saying?" Not, "How to these numbers confirm my bias?"

 

[RageQuit77]

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

I don't go now to the numbers grasping, but I have a grasp that every even mildly educated and scientifically oriented hockey fan should have alarm bells ringing loud with fundamental conceptual problems related to the Corsi as a player evaluation system:

https://morehockeystats.blogspot.fi/2016/12/p.html

Thanks for that HFer who brought that link to me elsewhere.

 

[zeke]

Sorry to jump in here, but this is a strange looking comparison to me. Here is why....

A shift is a physical thing.
A zone start is a physical thing.
"Corsi" is not a physical thing. "Corsi" is a collection of numbers used to analyze physical things.

Better wording, and the actual meaning, might be...."100% of shifts have measurable events which are used to calculate Corsi."

Concerning O-zone and D-zone starts, You have the numbers right there.. 39.xx CF on d-zone starts. 58.xx whatever on o-zone starts. 47.xx on neutral zone starts....
What doesn't someone incorporate those numbers into individual Corsi?
So, for individuals.....
Adjusted Corsi goes something like this:
% of d-zone faceoff starts * Corsi on those shifts/39.xx (or whatever the exact number is)+
% of o-zone faceoff starts * Corsi on those shifts/58.xx (or whatever the number is) +
% of n-zone faceoff starts * Corsi on those shifts/47.xx (or whatever the average is) +
% of on-the-fly starts * Corsi on those shifts/50.xx (whatever the avg number is).

That should take all of that into account, without the advanced stats enthusiast needing to say "The individual difference in Off vs Def starts is statistically insignificant, so we ignore it." Saying that sounds like an excuse. So, instead, incorporate it.

Corsi is nothing more than a way to count anyway. Just find a more precise way to count it.

On a side note, I find those %ages interesting. O-zone + D-zone < 100. N-zone <50. On-fly >50. The only way that can physically be true is if the team which wins the faceoff dumps the puck into the o-zone, and changes players before the other team has a change to do so, and then those 'new' players create a + Corsi. And, of course, that makes perfect sense. But, thinking it through like that is a good exercise in that it forces a person to ask, "What are the numbers REALLY saying?" Not, "How to these numbers confirm my bias?"

Current zone-adjusted corsi numbers do exactly what you are saying they should do here - and more.

Corsica.hockey is a good source here: http://www.corsica.hockey/blog/2016/06/19/adjustments-explained/

 

[zeke]

Yup, and more and more there is tangible proof that some of the foundational assumptions surrounding "what impacts" corsi were grossly flawed, and led to flawed conclusions.

http://leafshub.com/tougher-than-the-rest/

Kinda ruins the whole "every player plays against every other player approximately equally idea. Corsi based QoC is absolute garbage, been saying it for a long time.

Agreed that the failure to account for quality of competition and workload is the biggest current failing of advanced stats

But this doesn't render them useless.

You just have to acknowledge that the numbers work best when comparing players with similar roles and workloads - the easiest thing to look at here is pure minutes played, as it's harder and harder to sheltef someone the more they play.

So myself i always check workload before comparing players' possession numbers.

There is no good workload-adjusted corsi yet but for now you should always mentally asterisk somebody with great corsi numbers if they are not getting full frontline minutes.

just because a guy is putting up grewt numbers in a secondary or tertiary role doesn't mean he won't struggle in a primetime role. You might argue that the guy deserves a shot at getting more responsibility - but you can't conclude his better possesskon numbers make him a better possession player than someone with a significantly bigger role until he shows he can do it in that bigger role.

 

[CanadienShark]

22 out of last 24 cup winners have an A in their name. A smart owner would name their team with an A in it to increase odds of success.

Oh come on... .

If you think letters in a team name are as relevant to the discussion as actual aspects of the game, then well...