Has countering puck possession been successfully figured out?

[Channelcat]

Dont they measure time with the puck ? Im sure u got some money puck guys doing it

They used to, but you never hear actual possession stats anymore.

 

[Hobby Bull]

It's kind of a tenuous proxy measurement though. You could shoot the second you cross the blue line every time and it would be counted as "possession" even though that would mean you never had the puck in your control in the offensive zone. I don't understand why hockey doesn't measure actual possession like soccer. Or zone time.

I'd be surprised if someone didn't address this earlier... Sorry, but it's a new thread for me...

Corsi was intended to be a measurement of goalie workload. It was "annexed" to measure possession.

 

[AgentBrodeur]

My personal opinion, there needs to be a new column. The column which the GOAT excelled at more then any others. Hockey IQ.

Hockey broken down is a battle, of 3 period battles, down to a shift of a battle, down to each individual decision a player makes at any given moment and if they might the right move, or not.

Im not saying well he was in position X when the rebound was available at position Y. I am talking about every controllable consortium besides puck luck

• Did the player make a good defensive outlet pass or play.
• Did the player receive a perfectly receivable pass
• Did the player choose to pass when there was already a good shot attempt to take
• Did the player take a shot, when a pass would have been the better move
• Did the player miss the net after huge attempt to get a scoring chance?
• etc

You can argue that these "decision" can be "weighted" from a 1-5 (i.e.defender broke up a scoring chance). The hard part about this stat as it requires at least 10 people per game to enter this information, and they have to be knowledgeable enough to know and grade the person accordingly

 

[Anomie2029]

Corsi is a useful stat that has it's place among all of the other stats.

They detail a picture of a game that is GENERALLY representative of a winning team.

Even with that being said, Corsi itself can be adjusted for further analysis (score adjusted, player combinations, etc.). They can all give valuable information.

But it's only ever a slice of the pie.

 

[Nizdizzle]

My personal opinion, there needs to be a new column. The column which the GOAT excelled at more then any others. Hockey IQ.

Hockey broken down is a battle, of 3 period battles, down to a shift of a battle, down to each individual decision a player makes at any given moment and if they might the right move, or not.

Im not saying well he was in position X when the rebound was available at position Y. I am talking about every controllable consortium besides puck luck

• Did the player make a good defensive outlet pass or play.
• Did the player receive a perfectly receivable pass
• Did the player choose to pass when there was already a good shot attempt to take
• Did the player take a shot, when a pass would have been the better move
• Did the player miss the net after huge attempt to get a scoring chance?
• etc

You can argue that these "decision" can be "weighted" from a 1-5 (i.e.defender broke up a scoring chance). The hard part about this stat as it requires at least 10 people per game to enter this information, and they have to be knowledgeable enough to know and grade the person accordingly

Almost all the categories you listed are judgement based and not black/white. You can't make a reliable model when all the data is arbitrary. What if someone makes a bad pass that results in a good scoring opportunity? What about really risking plays, are they only good if they work out? In a 2-on-1 break, who decides if it was better to pass or shoot? The problem with this stat is that it isn't using concrete data, every data point would be a judgement call and subject to bias or human error.

 

[AgentBrodeur]

Almost all the categories you listed are judgement based and not black/white. You can't make a reliable model when all the data is arbitrary. What if someone makes a bad pass that results in a good scoring opportunity? What about really risking plays, are they only good if they work out? In a 2-on-1 break, who decides if it was better to pass or shoot? The problem with this stat is that it isn't using concrete data, every data point would be a judgement call and subject to bias or human error.

i agree as i listed that as a human to determine these factors would be very difficult to find. This also would not be an official stat of any kind. But if kept and done well could add significant value. (i worked at the NBA this tactic was in R&D)

But the non black and white plays, plays that resulted in fortune or unwanted desire. Should not be counted.

I experimented with this , and I did not capture allot of concrete data in just one single game(due to the facts you laid out). But all be it. There was data. The bad performers and good performers eventually did stick out in numbers.

 

[6 Karlsson 5]

That's the thing, there are no perfect stats. You have to combine as many as possible with what you see to come up with a reasonable conclusion.

No, you'll get hit hard on your adj r2 value if you just add as many stats as possible. You'd have to have a selection process to identify high enough F values to add to your MLR.
ITT: people talk about stats when they literally have no or very little knowledge of statistical theory.

 

[6 Karlsson 5]

That's part of the formula for xGoals (expected goals), xGoals also includes things like shot types rebounds and even captures some rush chances (which typically have higher shooting %). The methodology was described nearly a decade ago, the problem is that for all the additional complexity it doesn't actually work as well as Corsi. It wasn't a case a of people not thinking of this stuff, when it was looked at and shelved because the simpler metric worked better.

A couple years ago xGolas was revisited by Ryan Stimson and he added a couple things to it. Primarily he integrated historical data on who was taking the shot. His model does outperform Corsi, however he was hired by the Avs at the start of this season and his data has been removed. Some people may still be getting his data but AFAIk it's no longer public domain. The xGoals most people cite is from Corsica.hockey, but but while their model is improving AFAIk it still underperforms Score Adjusted Corsi.

Do you have source for the last statement?
Based on your previous posts, you do seem to have a strong grasp on stats, but I, personally, haven't seen data concluding Score-Adjusted Corsi outperforms xGoals
It'd be really cool if you could link it, so I could take a look.

 

[6 Karlsson 5]

I don't know I think there is too much analysis in this game today. Whatever happened to just get the damn puck in the net and keep it out of yours?

People like to win more than you, i guess.

 

[Volica]

Hockey is the most random number generated sport out of the big 4.
At the start of the NBA season, you can more or less guarantee who will be the last 4 teams remaining in the playoffs.
In Baseball, at the stat of the season, you can normally call who'll be in the top of the division; and can go deep.
In the NFL, you can peg pretenders to contenders pretty easy.

At the start of this season, no one would have pegged the final 4 teams. Maybe Tampa, but for sure not the Western final, for sure not Washington (outside of Caps fans).

I think the good thing about Corsi, and this statistic along with others (HDC%, Sh% in conjuction, etc) is that they've more or less retired the dinosaurs. The players that you were wondered how they were still in the NHL are more or less gone. I think it's painted an excellent picture of 'this player is a liability when he hits the ice' better than +/- could or GA could.

Because hockey is such a game of momentum and emotion, I think this has started a trend for sure.

 

[JoeThorntonsRooster]

Hockey is the most random number generated sport out of the big 4.
At the start of the NBA season, you can more or less guarantee who will be the last 4 teams remaining in the playoffs.
In Baseball, at the stat of the season, you can normally call who'll be in the top of the division; and can go deep.
In the NFL, you can peg pretenders to contenders pretty easy.

At the start of this season, no one would have pegged the final 4 teams. Maybe Tampa, but for sure not the Western final, for sure not Washington (outside of Caps fans).

I think the good thing about Corsi, and this statistic along with others (HDC%, Sh% in conjuction, etc) is that they've more or less retired the dinosaurs. The players that you were wondered how they were still in the NHL are more or less gone. I think it's painted an excellent picture of 'this player is a liability when he hits the ice' better than +/- could or GA could.

Because hockey is such a game of momentum and emotion, I think this has started a trend for sure.

I think Winnipeg wouldn’t have been a stupid pick. For years, they had a strong team on paper, but couldn’t convert it to any more than being a bubble team on the outside looking in.

However, I entirely agree with the premise of your post.

 

[bambamcam4ever]

Corsi is more reliable because typically you see 100 to 130 or so shot attempts per game, while you'll only see 5-7 goals per game.



The difference between good defense and bad defense is actually the number of low danger chances allowed, and NOT the number of high danger chances allowed, because every team actually allows roughly the same number of high danger chances which is between 10 and 13 high danger chances per 60 minutes.

I know it sounds counter-intuitive, but the data doesn't lie.



Individual players are typically at the mercy of their coaches and teammates. So yes looking at raw Corsi data to compare individual players is in fact flawed.



The eye test only works if you actually watch every game of every team and physically compare every team and player. Even then the eye test has a bias, and becomes completely unreliable.



While this is entirely true, Corsi serves as a great foundation. Without that foundation of having good Corsi, the whole structure can collapse quite quickly.

If your data is saying the Kings and Islanders are giving up the same amount of high quality chances your data is severely lacking.

 

[Lap2000]

Corsi is just part of the game that teams chased and it has started to play its course with teams learning what they needed to from the theory. We see teams doing this even in our bantam and midget aaa leagues in the last couple of years as it trickles down although it seems to be a failure at those levels based on results of those teams. A couple teams even rostered up with forwards on defense and it tended to generate more shot differential but bad results even though it seemed like the teams were faster. Obvious theory is chasing leads to more golden opportunities against creating a consistently bad overall team that gets lots of shots but let’s in lots of goals.