TheHockeyRant
Registered User
I wouldn't call that possession, I would call that shot dominance.
which over time can translate to more wins. But not always, just ask the New Jersey Devils (finished 3rd in Corsi in 2014)
I wouldn't call that possession, I would call that shot dominance.
which over time can translate to more wins. But not always, just ask the New Jersey Devils (finished 3rd in Corsi in 2014)
I think it's important to note that this is the context that Darryl Sutter was referring to when he said during the playoffs last year that (paraphrasing) Defensive hockey is a bygone era.
Full quote:
A team can "defend" and still maintain puck possession. Bad teams, like the Leafs (Leaf fan saying this), fall back and cluster in their own zone, and let the other team pepper them with shots when protecting a lead. Good teams will continue to fight to control the puck, and just won't fight for scoring chances, but reserve their energy to control the play and limit opposition scoring chances.
One of the best examples I can think of, historically, for this "new" system would be Bobby Orr.
There are countless examples of Bobby skating the puck up to the opposing blue line, and turning around, going back to his end and restarting a zone exit system because the preceding one didn't work to his liking.
Really good players/teams employ this strategy nowadays. This is why a lot of the advanced stat people are so quick to dump on "dumb-and-chase" fanatics. The evidence shows that dump-and-chase systems are categorically inferior to possession-entry systems.
It's unfortunate that North American hockey became so sterotypically founded on dump-and-chase forecheck hockey. Probably had a lot to do with the USSR being the key proponent of possession-style hockey. So just to be confrontational, we branded our hockey as dump-and-chase and chased away any suggestions that we might be wrong...
Fortunately it only took us 22 years after the fall of the Soviet Union to correct this.
Well played, Ruskies... Well played...
TOML
Registered User
Doctor No
Registered User
TOML
Registered User
Hundreds of 'analytics' articles being written, combined to various hirings by teams to address the issue this summer. This only serves to inflate the value of the numbers in question.
Doctor No
Registered User
So your position is that advanced analytics in hockey are currently "overvalued in a big way"?
I guess that's fine to be your opinion. Mine is that we've still barely scratched the surface.
The difference is probably that most NHL teams aren't talking about what they're doing (at least not to the public). There's some exciting stuff happening.
TOML
Registered User
No. This thread is about the value corsi and fenwick. Corsi and fenwick have value, but no more value than many other numbers. However, with the hype job (coming from the TO area in particular), they are trending towards being overvalued imho.
There will be more stats coming due to advancements in tech. As long as it's remembered that the player makes the numbers and not the other way around, teams will be fine.
TheDevilMadeMe
Registered User
I don't think Corsi and Fenwick are overvalued by team analytics departments that use much better statistical data that isn't publicly available. I agree that they have become overhyped by the media and a portion of the general public, that seems to view them as some holy grail in determining a player's value.
The whole "we suddenly found out that defensive defensemen are worthless because they don't drive Corsi!" thing makes me cringe.
TOML
Registered User
Agreed totally, but then a guy like Tim Leiweke comes on the air and pretty much confuses everyone. Good thing he's not directly involved.
The way I see it, Corsi and Fenwick don't measure possession, they measure shots. Think of them as fancy SOG numbers. That may correspond with possession and it may not because it isn't a measure of possession. Ever watch a team throw around the puck for 30 seconds in the opponent's zone, then lose it and never attempt a shot? That is possession, Corsi and Fenwick don't see it.
The second thing to know about these numbers is they are useless for measuring individuals. They measure 10 players at a time - you cannot get an individual measure from a team stat like that. Picture you getting disciplined at work for something the guy in the next cubicle does. Is that fair, right or accurate? Nope, but that sums up Corsi & Fenwick.
These stats are not "advanced", they are actually regressive as they hinder understanding by not accurately measuring anything of real value. They are little more than fancy SOG and +/- stats, IMHO.
The second thing to know about these numbers is they are useless for measuring individuals. They measure 10 players at a time - you cannot get an individual measure from a team stat like that. Picture you getting disciplined at work for something the guy in the next cubicle does. Is that fair, right or accurate? Nope, but that sums up Corsi & Fenwick.
These stats are not "advanced", they are actually regressive as they hinder understanding by not accurately measuring anything of real value. They are little more than fancy SOG and +/- stats, IMHO.
Hivemind
We're Touched
I agree they're not "advanced" stats, and I hate that monicker. However, I can't agree with much else you say, Norm.
The reason they're referred to as possession metrics is not arbitrary. It's because shot counters correlate very heavily with possession.
http://vhockey.blogspot.com/2008/08/zone-time.html
http://vhockey.blogspot.com/2008/08/zone-time-corsi-and-correlation-to.html
http://www.pensionplanpuppets.com/2013/9/16/4727746/leafs-attack-time-at-the-halfway-mark
In terms of tracking inviduals, that's why sample size must always be considered, and why statistics like relative Corsi/Fenwick and WOWY measurements exist. As the sample size increases, biasing from other players will generally decrease. Relative corsi compared a player's corsi value to that of the rest of the team while he's off-ice, which does a great deal to isolate his individual contribution. WOWY (with or without you) analysis examines how a player performs with and without particular teammates.
These stats are very far from "regressive." They have gained popularity because they are flexible, predictable, able to be produced via game logs as opposed to manual tracking, and most importantly, correlate with winning hockey games.
The reason they're referred to as possession metrics is not arbitrary. It's because shot counters correlate very heavily with possession.
http://vhockey.blogspot.com/2008/08/zone-time.html
http://vhockey.blogspot.com/2008/08/zone-time-corsi-and-correlation-to.html
http://www.pensionplanpuppets.com/2013/9/16/4727746/leafs-attack-time-at-the-halfway-mark
In terms of tracking inviduals, that's why sample size must always be considered, and why statistics like relative Corsi/Fenwick and WOWY measurements exist. As the sample size increases, biasing from other players will generally decrease. Relative corsi compared a player's corsi value to that of the rest of the team while he's off-ice, which does a great deal to isolate his individual contribution. WOWY (with or without you) analysis examines how a player performs with and without particular teammates.
These stats are very far from "regressive." They have gained popularity because they are flexible, predictable, able to be produced via game logs as opposed to manual tracking, and most importantly, correlate with winning hockey games.
BeardyCanuck03
@BeardyCanuck03
In the minds of the average fan you may be right. But I don't think the hockey braintrusts will overvalue them.
Just like every other stat, it only tells the end of the story. What will separate the top analytic teams/people from the rest will be how they break down how those stats occured. Was it the player (his smarts, speed, etc) or was it how he was coached/deployed (systems, zone starts, etc)?
Ail
Based and Rangerspilled.
Hockey is a team sport, you absolutely can measure an individuals performance based on these stats. If you got in trouble for a project the guy next to you screwed up that you weren't working on, you're right, that would be unfair and ridiculous. However if you were working on that project too, even if the guy next to you screwed up, you take some of the blame for not correcting his mistake, or helping him in the first place since you are working with him.
The 5 guys on the ice at the same time are all working on the project of trying to score, or not get scored on. In small samples sizes, yes it would be bad looking for all 5 guys if 1 of them kept messing up and bringing the other 4 down. (You often see this with poor possession players who bring down line-mates, or pairing mates.) The good news is for advanced metrics, the 5 guys on the ice at the same time are rarely the same over the course of many games, let alone many years. Furthermore, the line-combos, and defense pairings to which most players are measured against also change over time and in large sample sizes start to give you an accurate picture of what individuals look like on the ice, based on comparisons to their teammates.
It's not perfect, but it is relatively accurate enough, especially when all the variables are considered. Including competition and zone starts.
You cannot pin a team stat on an individual. If you want individual stats, measure individuals.
Hivemind
We're Touched
So you're just going to ignore everything I said, continue plugging your ears and repeat the same stuff? You're not going to offer any evidence for your rebuttal?
What I really cannot wait for, is the first time a player sits in the press box because of how bad his corsi numbers are.
My question's are, and maybe I'm just not fully grasping the idea or even wanting to as I am a simpleton because I want to watch with beer in hand as players make puck in net and not in own net, a simple form will you, how does corsi/fenwick work when say a players stick breaks, or loses an edge or a goal post comes into play, things that no one can predict?
This seems like a huge waste of time. Again I'm very uneducated on the subject and doubt I will take the time to ever consider it as a real problem for teams.
In a nut shell, the team that scores more goals wins. Why do teams need corsi/fenwick?
My question's are, and maybe I'm just not fully grasping the idea or even wanting to as I am a simpleton because I want to watch with beer in hand as players make puck in net and not in own net, a simple form will you, how does corsi/fenwick work when say a players stick breaks, or loses an edge or a goal post comes into play, things that no one can predict?
This seems like a huge waste of time. Again I'm very uneducated on the subject and doubt I will take the time to ever consider it as a real problem for teams.
In a nut shell, the team that scores more goals wins. Why do teams need corsi/fenwick?
Hivemind
We're Touched
If things are truly unpredictable, there's no point in buliding statistical models to predict them. Those events will happen, and it doesn't really have anything to do with Fenwick or Corsi.
Goals are a relatively rare event that are impacted by a lot of random variability. Shot attempts are far more frequent, and have been shown to be far more repeatable. As a result, corsi/fenwick have less "noise," and are better predictors of future success than goal-based metrics.
wgknestrick
Registered User
- Aug 14, 2012
- 5,869
- 2,626
I offered common sense - if that doesn't resonate with you, there isn't much to discuss.
Taylor Hall is the best player on the Oilers by a good margin, Corsi says he is one of their worst players. That is completely flawed BS. Giving Hall a - when Jeff Petry passes the puck to an opponent in front of the Oilers goal further underlines what crap Corsi is.
The argument that over a large sample size, it works itself out is garbage. Bad data multiplied just becomes worse data. Corsi advocates are just grasping at straws - they are so desperate to claim they have "advanced" stats, they take complete garbage and try to convince the world it provides insight. The naive are fooled but, those with common sense see holes you can drive a truck through.
If you want to measure an individual player's effectiveness, measure the individual for Heaven's sake. Taking a team measurement (that is lille more than a glorified SOG count) and trying to shoe horn it on individual players is wrong, inaccurate and futile. Is it laziness that drives people to use Corsi rather than really evaluating individual performance? I don't know but, I do know that Corsi provides nothing of value as an individual measure.
Doctor No
Registered User
On the other hand, this is a *discussion board*, so I typically expect discussion.
Why are you looking at Corsi by itself, in isolation, to define Taylor Hall and his abilities?
So the CorsiRel for Taylor Hall over the last couple years is +15.2, +20.3, and +2.4 that seems like it suggests he is a pretty good player, no?
Not that I think that CorsiRel is the be all and end all but to say that it doesn't like Hall is just not accurate.
Not that I think that CorsiRel is the be all and end all but to say that it doesn't like Hall is just not accurate.
hatterson
Registered User
You're absolutely right, you're not able to predict things like sticks breaking, or losing an edge, or a bad bounce off the stanchions. However, one of the underlying ideas behind stats in general is seeing what happens over the long run. Bad breaks happen (just like good breaks), but in general they even out over the long run. You go on a lucky streak, then on an unlucky streak.
That happens when looking at both "advanced" stats as well as regular stats. Crosby having a 1.3 PPG doesn't mean he can't go on a stretch of 4 games without a point, or that he can't have a breakout game with 6 points. Similarly a player that drives a high corsi can have bad games, go on a cold streak, and not score for a week while at a same time a player who severely harms corsi can go on a hot streak and score 6 points in a couple games.
No amount of measurement we do will ever be able to perfectly predict a play, shift, game or series, but the overall idea is that we can use the data to compliment our observations about the sport and provide a better overall picture that will hold up better in the long run.
Corsi isn't really interested in predicting if Stamkos will score 3 goals on November 5th in the same way that a scout isn't really interested in telling his GM that a player is likely to have a good game 3 weeks from now. They're looking for longer term, bigger picture things, trusting that the bounces will even themselves out over the long run.
First, Corsi, like any other stat, doesn't tell you who the best and worst players are on a team. It tells you who was the best and worst at that stat. That's a very significant difference. That's like saying the goals stat is useless because it says David Perron is the best players on the Oilers.
Second, I'm not sure where you're getting your info from but Hall has typically excelled at corsi measures in his career. In his first 2 years he led the Oilers in corsi rel, the second year by a large margin. In 2012-2013 he was 2nd behind only Eberle, and him Eberle and RNH were miles ahead of the rest of the team. Last year is the only year he hasn't excelled, and even then he still performed well above average.
If you want to pick an example of a good player who has poor corsi, Hall is one of the worst to use.
It sounds like you're not actually understanding what the argument about sample size is. It is not multiplying bad data over and over, it's running a large sample so we can eliminate outliers. That's stats 101.
If I flip a coin 5 times and get heads all 5 times, saying that I should flip it more isn't multiplying bad data, it's trying to get a large sample size so we can see if those 5 flips were a fluke or if the coin is actually weighted heavily.
Corsi, on it's own, doesn't provide much for an individual measure, you're right. That's why we also look at corsi rel, qualcomp, qualteam, zone starts, etc. Looking at data in context helps us deepen our understanding of players and teams. Looking at it in a vacuum is a great way to get absurd results.
Giving Petry an assist because he passed the puck to Hall behind his own net and then Hall went end to end to score underlines what crap assists are.
Giving Hall a goal because a puck ticked off his stick and went into the neutral zone, and then the other team passed it back to the goalie, who then fumbled it and let it in underlines what crap goals are.
Do you see how absurd those statements are? You can strawman out a scenario for any stat you want and make it sound stupid, that says nothing about the stat itself and much more about how you're using it (or in this case, burying your head in the sand and ignoring it)
BeardyCanuck03
@BeardyCanuck03
In a nut shell, how do you score goals? You shoot the puck. Corsi tracks shot attempts, which are shots on goal + shots that missed the net + shots that were blocked. Fenwick tracks shots on goal + shots that missed the net.
Ad
Upcoming events
-
BET ON ONLY ONE HORSE Kentucky Oaks (Gr. 1) - USA NetworkWagers: 6Staked: $1,140.00Event closes- Updated:
-
-
-
-