HFBoards

Go Back   HFBoards > NHL Eastern Conference > Atlantic Division > Montreal Canadiens
Notices

Sabremetrics/Microstats in Hockey

Reply
 
Thread Tools
Old
03-30-2012, 03:50 PM
  #26
Poulet Kostopoulos
Registered User
 
Poulet Kostopoulos's Avatar
 
Join Date: Oct 2009
Posts: 4,843
vCash: 500
Quote:
Originally Posted by Goldthorpe View Post
The microstats (in hockey, this tend to be the term for "sabremetrics", I don't know why) that already exist and have been compiled in the last few years don't go that far, but they still go farther than the traditional stats we all know (goals, assists, hits, etc.). Some of them have been shown to be pertinent over a long period of time (like many years).

For anyone who is just starting on the topic, I highly suggest the links provided by Mathman above.
I'm actually surprised that no one went "that far" yet. I'm far from being an expert on the subject but it seems to me that the important ingredients are simply time and computational power.

Poulet Kostopoulos is offline   Reply With Quote
Old
03-30-2012, 03:52 PM
  #27
Cyclones Rock
Registered User
 
Join Date: Jun 2008
Posts: 1,451
vCash: 500
Quote:
Originally Posted by Roke View Post
There's no salary cap on front office operations. For less than what Georges Laraque cost the Habs they could probably put together an elite stats department to compliment the rest of the organization. As with scouting they'd get things wrong, but having another perspective within the organization probably wouldn't be a bad thing
This would be a very intriguing experiment. If such a staff could create models which were effective, the savings from not signing a bad UFA could more than justify their expense.

Back when I played high school hockey-when dinosaurs roamed the earth-I distinctly remember a hit which altered my (and a few other players) play that couldn't be quantified by any metric of which I'm aware. (The change in play could be quantified but the reason behind it couldn't be. Not in any standard manner.)

In a game early in my senior season, a guy (who will appear in a linked video), charged from the blue line in full stride, left his feet and knocked out about 3 or 4 teeth of one of my teammates. I had a tendency to hold on to the puck a bit more than a should have (I guess I would now call myself a puck possession forward). After seeing this hit, for the rest of the game-and every time I played against this guy thereafter-my style abruptly changed from puck hog (I mean puck possession) to passing teammate (I mean 'dump and skate away'.

Intimidation can't really be measured, but it's very real. A lot of players won't go into the corners as zestfully or as often as they should do to fear. From lower level hockey to the NHL. Quantifying stuff like this is beyond the realm of micro stats, imo.

And here's my former tormentor. R.I.P. He was one scary dude. LOL

http://www.youtube.com/watch?v=68Pmcq-ZA2M


Last edited by Cyclones Rock: 03-30-2012 at 04:00 PM.
Cyclones Rock is offline   Reply With Quote
Old
03-30-2012, 04:00 PM
  #28
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,891
vCash: 500
Quote:
Originally Posted by Cyclones Rock View Post
This would be a very intriguing experiment. If such a staff could create models which were effective, the savings from not signing a bad UFA could more than justify their expense.
We know for a fact that certain GMs already do things like this: The Bruins, because Chiarelli has said so himself. Doug Wilson of the Sharks. The Penguins as well. Detroit at least counts scoring chances and probably more. Teams are, understandably, loathe to tell exactly what they collect (as this would give competitive advantadge to other clubs) but we do know that they gather data.

Edmonton is, I believe, looking to invest in this. Particularly ironic since a lot of the analytics work was started by Oilers bloggers and they are generally extremely unimpressed with the Oilers front office.

Heck, Martin would regularly quote the Habs' scoring chances numbers. Funny though, I don't recall Cunneyworth making mention of this.

MathMan is offline   Reply With Quote
Old
03-30-2012, 04:00 PM
  #29
Roke
Registered User
 
Roke's Avatar
 
Join Date: Jul 2003
Location: Winnipeg
Country: Canada
Posts: 1,885
vCash: 500
I've actually been thinking about the problems with the way hits, giveaways, and takeaways are counted in the NHL. Like MathMan said the tracking of them is very inconsistant from building to building. It's not all that different from a hook being a penalty at one point in time and not later in the game/in the season/ with different referees.

With the hit/takeaway/giveaway tracked as is by the NHL I think they're pretty much useless. Lafleurs Guy mentioned that a hit by Chara is given the same weight as a hit by Desharnais and he's absolutely right. Just like a hit separating a guy from the puck is given the same weight as a hit behind the play a second or two after the player has passed the puck.

Just Thoughts/Chris Boucher's stats do a pretty good job of measuring actual puck battles one/lost and his data on the Habs isn't tracked for any other team in the league. While I'm still overwhelmed by the sheer amount of data he has it's pretty valuable in my opinion.

Along those lines, I think the "hit" statistic would have more meaning if it was actually "hit causing a change in possession of the puck" or at least "hit separating the guy from the puck". Similarly, giveaways/takeaways would be more valuable if all changes in possession were counted, whether it be an unrecovered dump-in or a defenseman chipping the puck off the glass in the offensive zone to the opposition.

If I finally get a PVR for next season and some free time I think I might end up counting something along either lines for the Habs since scoring chances are already being tracked. I'm not even sure if it's feasible without a lot of spare time given the speed of the game.

Roke is offline   Reply With Quote
Old
03-30-2012, 04:08 PM
  #30
Mathletic
Registered User
 
Mathletic's Avatar
 
Join Date: Feb 2002
Location: St-Augustin, Québec
Country: Canada
Posts: 11,748
vCash: 500
Quote:
Originally Posted by Poulet Kostopoulos View Post
I'm actually surprised that no one went "that far" yet. I'm far from being an expert on the subject but it seems to me that the important ingredients are simply time and computational power.
if a team actually gave it time and ressources ... like someone said earlier, only investing the equivalent of BGL's salary on an analytics team ... I'm pretty sure would get pretty far and early. Problem is, there's very few teams who have interest in the subject.

Professors I have talked to who have made research on stats analysis in hockey have had no interest from NHL teams. Most teams don't even bother reading the reserach paper in the first place. So, most of them won't bother wasting their time on doing research in that area.

Other sports have been much more open to it. Most notably basketball. Teams like Dallas, Houston and Boston, among others have been open for different ideas. I believe it has an impact as far as people being much more interested in developing new ideas for analyzing basketball since they know their reserach will be taken into consideration in one way or another.

Though NHL executives aren't stupid by any means, far from it. Most of them don't have trainning in computer science, stats and whatnot. For them it must be pretty new being exposed to the idea that stats may actually help you in decision making since throughout your life in hockey, you learn that stats are as useless as they come.

NHL Owners don't seem to be too involved in decision making either on the hockey side either. For the Dallas Mavericks, Mark Cubban pretty much oversaw the implementation of analytics for his team. The implementation part is probably the toughest part to get through for analysts since you may have all the best research in the world, but if it's not used by NHL teams, there's no way they'll have any proof the **** works.


Quote:
Originally Posted by Roke View Post
If I finally get a PVR for next season and some free time I think I might end up counting something along either lines for the Habs since scoring chances are already being tracked. I'm not even sure if it's feasible without a lot of spare time given the speed of the game.
that would be the problem for most people unless you work for an nhl team, you need tons of free time to do this type of thing by hand. There are softwares now available that may help you build much greater samples at a faster rate. I believe there was a project at UBC of students who build a software that could track the puck, player position on ice and whatnot off video information.


Last edited by Mathletic: 03-30-2012 at 04:13 PM.
Mathletic is online now   Reply With Quote
Old
03-30-2012, 04:13 PM
  #31
Cyclones Rock
Registered User
 
Join Date: Jun 2008
Posts: 1,451
vCash: 500
Quote:
Originally Posted by MathMan View Post
We know for a fact that certain GMs already do things like this: The Bruins, because Chiarelli has said so himself. Doug Wilson of the Sharks. The Penguins as well. Detroit at least counts scoring chances and probably more. Teams are, understandably, loathe to tell exactly what they collect (as this would give competitive advantage to other clubs) but we do know that they gather data.
With a "micro" staff and a coaching staff which was on board with the use of these stats, I could see a great value rendered in adjustments of individual's playing styles. i.e. player x takes 40% of his shots from outside a certain area and these shots result in 2% of his points and 20 % of those shots were blocked and resulted in turnovers. Player x's playing strategy could be altered accordingly to take higher percentage actions. I don't believe that public microstats are this highly specific at this point and probably couldn't be just due to the vast quantity of information which more individual team specific stats would require.

I would think that an owner like Molson might be ripe for an experiment along these lines. He's young and very wealthy and the team is highly profitable. The new GM, if so inclined on the notion of pursuing micro stats at a high level, should hit Molson up on this idea right after Gomez is bought out. The cost will be peanuts by comparison.

Cyclones Rock is offline   Reply With Quote
Old
03-30-2012, 04:15 PM
  #32
Mathletic
Registered User
 
Mathletic's Avatar
 
Join Date: Feb 2002
Location: St-Augustin, Québec
Country: Canada
Posts: 11,748
vCash: 500
Quote:
Originally Posted by Cyclones Rock View Post
With a full staff and a coaching staff which was on board with the use of micro stats, I could see a great value rendered in adjustments of individual's playing styles. i.e. player x takes 40% of his shots from outside a certain area and these shots result in 2% of his points and 20 % of those shots were blocked and resulted in turnovers. Player x's playing strategy could be altered accordingly to take higher percentage actions. I don't believe that microstats are this highly specific at this point and probably couldn't be just due to the vast quantity of information which more individual team specific stats would require.

I would think that an owner like Molson might be ripe for an experiment along these lines. He's young and very wealthy and the team is highly profitable. The new GM, if so inclined on the notion of pursuing micro stats at a high level, should hit Molson up on this idea right after Gomez is bought out. The cost will be peanuts by comparison.
that you can actually do with play-by-play data, given you adjust for certain arenas like MSG

you can get the spot where a player made his shot, shot type (slap shot, wrister ...), whether it was blocked or missed the net and wheter it hit the net.

Mathletic is online now   Reply With Quote
Old
03-30-2012, 04:27 PM
  #33
Cyclones Rock
Registered User
 
Join Date: Jun 2008
Posts: 1,451
vCash: 500
Quote:
Originally Posted by Mathletic View Post
that you can actually do with play-by-play data, given you adjust for certain arenas like MSG

you can get the spot where a player made his shot, shot type (slap shot, wrister ...), whether it was blocked or missed the net and wheter it hit the net.
Would you see any use in a team gathering information in this area or do you think that most useful information pertaining to this specific is already out there?

Cyclones Rock is offline   Reply With Quote
Old
03-30-2012, 04:31 PM
  #34
The Gal Pals
Breaking Hab
 
The Gal Pals's Avatar
 
Join Date: Oct 2006
Country: Canada
Posts: 1,822
vCash: 1549
Quote:
Originally Posted by Roulin View Post
This is actually very much tangible/measurable. Results with specific teammates are available at the timeonice site. More general quality of teammate stats are available at behindthenet.
What I'm saying is that you can't use the Billy Beane method to stack your team with players that will give you a statistical advantage when you don't even know how they will react when playing with one another. Sure there are stats on how Player X may react with his teammates on one team but how does that help me if I want to bring him to my team? He'll have a whole new set of linemates. All I'm saying is that it's completely unpredictable how players will interact with one another before you actually see them playing together.

The Gal Pals is offline   Reply With Quote
Old
03-30-2012, 04:32 PM
  #35
Mathletic
Registered User
 
Mathletic's Avatar
 
Join Date: Feb 2002
Location: St-Augustin, Québec
Country: Canada
Posts: 11,748
vCash: 500
Quote:
Originally Posted by Cyclones Rock View Post
Would you see any use in a team gathering information in this area or do you think that most useful information pertaining to this specific is already out there?
it would be of great use for a team. Even though the information is available I seriously doubt NHL teams exploit that info. Michael Shuckers built a goalie stat of this. He quatified quality of shot a goalies faced in order to measure a more accurate SV%. With his stat it's easier to guess how many goals a goalie would give up given a certain defense in front of him.

Michael Shuckers was part of the hockey pannel at the MIT Conference and he pretty much was blown off by Burke and Milbury. Can't say I didn't see this coming but anyways.


Quote:
Originally Posted by airic000 View Post
What I'm saying is that you can't use the Billy Beane method to stack your team with players that will give you a statistical advantage when you don't even know how they will react when playing with one another. Sure there are stats on how Player X may react with his teammates on one team but how does that help me if I want to bring him to my team? He'll have a whole new set of linemates. All I'm saying is that it's completely unpredictable how players will interact with one another before you actually see them playing together.
that's why I tend to prefer adjusted +/- stats. More specifically, what you may call team-fit adjusted +/-. I have been working on this for some time. A team of researchers presented a similar paper at the MIT Conference last year. They optimized lineups based on various stats like giveaways, takeaways, rebounds, points, assists and whatnot for Basketball. What they found in genral was that you were better off leaving your rebounders with bench players since they miss more shots. You'r also better off having at least 2 players who create turnovers together on the court at the same time since their abilities sum is greater than the sum of their parts... not sure if it's clear enough, basically means that if you expect 1 steal per half from a player. If you match that same player with another player who steals 1 ball per half then you can expect something like 2.5 steals for the half since they complement one another ...

The use of that stat also makes for more possible win-win situations for trades. A team may want to add a rebounder whereas another team may want to add a better defender and whatnot. That way you can quantify expected results for each team so they can make fair trades.


Last edited by Mathletic: 03-30-2012 at 04:40 PM.
Mathletic is online now   Reply With Quote
Old
03-30-2012, 04:45 PM
  #36
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 20,298
vCash: 500
Quote:
Originally Posted by Roke View Post
I haven't seen the movie, but the book is a very good read. I actually got less out of the Billy Beane/Oakland A's parts of the book than the bits on Bill James. This quote from one of Bill James' abstracts still stands out for me:

Those are the two most powerful lines in the book for me.

I don't follow basketball but I understand the analysis is fairly in depth (and the MIT conference was created, at least in part, by the Houston Rockets GM). MathMan mentioned Gabe Desjardins earlier and some of the early hockey analysis he did was informed by the basketball stuff a few years ago. Desjardin's also done consulting work on football arbitration and with hockey clubs and he's venturing deeper into soccer analysis. Basketball's advantage over hockey is that you have pretty distinct possessions.

The advantage baseball has over the other sports (especially soccer) is that most of the analytical developments happened outside the front offices so it's not proprietary and available to everyone.

Hockey Analytics are no replacement for scouting (as the Toronto Blue Jays discovered under Riccardi... though there were other decisions made by him that hamstrung the team) but at least I think they're an important complimentary tool. At the least you can get context for what you're seeing on the ice and on the basic stat sheet such as quality of opposition or where a player starts their shifts.

If a team was to say, hire a bunch of University students to watch games and count "things" (I'd go with scoring chances in every NHL game) you could have a pretty good advantage for evaluating players outside your organization and only you would have the data.

We don't really know what teams track but San Jose's assistant GM Joe Will uses some proprietary software of some sort, at the Sloan Conference Peter Chiarelli mentioned at the Sloan Conference that the Bruins use some form of analytics, and I've heard 3rd-hand the Islanders consulted with some stat-minded types when they signed Matt Moulson out of the AHL (well, the Kings organization) and when they took a look at the top prospects in the 2009 draft.

There's no salary cap on front office operations. For less than what Georges Laraque cost the Habs they could probably put together an elite stats department to compliment the rest of the organization. As with scouting they'd get things wrong, but having another perspective within the organization probably wouldn't be a bad thing
I agree. But see my question immediately below:
Quote:
Originally Posted by Kjell Dahlin View Post
"... Size and grit ARE attributes that GMs simply can't ignore. And I'm not sure how you capture this via analytics."

Size... well... pounds or kilos should do the trick!

As for grit: fighting majors, roughing minors, hits and (Giveaways- Takeaways differential) are a good start imo.
Problem is that giveaways and takeaways vary hugely from rink to rink. And that leads to another big problem in analytics for a fluid sport such as hockey... subjectivity. What constitutes a giveaway for one person doesn't for another... if the NHL can't get it right themselves, then it casts doubt on the ability to 'count things' doesn't it?
Quote:
Originally Posted by Mathletic View Post
There have been other advances in quantification of "intangibles". I remember a conference at the MIT Sloan Conference on a group of researchers who have analyzed speech of different players in order to predict future behaviour as well as another team measuring and assessing various leadership skills. For instance, players who touch teammates more than others like Garnett IIR ... I know it doesn't sound good but anyways ... I don't remember the whole thing but anyways, it's being analyzed.
Very cool.
Quote:
Originally Posted by Mathletic View Post
Then again, you have to approach stats analysis in an open way. In general analysts will tell you that hits are vastly overrated unless you create turnovers out of it. Teams that hit a lot, don't have the puck a whole lot. Though 1 smahing hit may make an impact in injuring a player, in general it doesn't do a whole lot to help you win. It only serves the macho in us who wants "displays of power" if I may put it this way.
I think this is an example of where statisticians can go wrong. The numbers show a correlation between two things and then they immediately argue that the numbers prove or show a conclusion... that's not necessarily the case.

I think there's more to a big hit than an empty display of power. If you watch Lindros, Kariya and countless others get crushed by Scott Stevens you're going to be a lot more aware of your surroundings when playing the Devils. It could cause you to fumble the puck or miss an opportunity. The guy doesn't necessarily even have to throw another hit to make an impact. The impact was made when the opposing player was watching footage of Lindros being carted off the ice from a hit two years ago and he doesn't want to be sitting there drooling in his helmet for the rest of his life.
Quote:
Originally Posted by MathMan View Post
Hits, as a stat, are actually negatively correlated to winning, presumably for this very reason. I don't think that being willing to hit actually causes a team to lose more, but you only hit if you don't have the puck... and puck possession is how you win hockey games.
If we could get a decent giveaway stat, I wonder if we might not end up with the entirely counter-intuitive correlation that more giveaways correlate to more winning (because giveaways are a sign of puck possession).
Is this always true? Show us the evidence.

Again, the CORSI stat has not impressed me. I look at the CORSI leaders and they don't correspond to the best players. Brett Hull never had the puck but he was one of the best scorers of all-time.

You make a blanket statement like puck possession wins hockey games... how true is this? What is there to back this up? Why does a player like Gomez who seems to have the puck so much suck so bad?
Quote:
Originally Posted by MathMan View Post
This actually brings to mind another point that, I must admit, occasionally bugs me.

If one can think of an objection to microstats within thirty seconds (what about shot quality? what about teammates? what about those guys who always play against the top lines?), that doesn't give leave to dismiss the whole thing. The odds are very good that the objection has already been thought of by analysts, examined, and either had a metric introduced or was dismissed as not making enough difference to matter on aggregate.
Sure, but if you don't have an answer to the objection it's going to breed a lot more skepticism. We've seen that with CORSI and Gomez. Brett Hull is another example. You come out with a statement like 'puck possession wins hockey games'... well, maybe as a general rule that's true. But it certainly doesn't always seem to be the case. And when I hear somebody defend the hell out of Gomez because of a microstat, it casts doubt on that person's credibility. If you can't come up with a reasonable answer to somebody's objection... you can't blame them for being skeptical. And you can't sit there and argue... well analysts probably have thought this out already... okay, if that's the case what's the response?
Quote:
Originally Posted by Goldthorpe View Post
I agree that this is very annoying, yet a common occurrence in these discussions. People, it's not because you just thought of a zinger 3 minutes ago that it hasn't already been answered 3 years ago.
Sure. But if you can't provide the answer that was developed 3 years ago, then you should expect skepticism.
Quote:
Originally Posted by Goldthorpe View Post
Also, the "you can make statistics say anything" argument.
Well, a lot of people twist stats to come to ridiculous conclusions. We see that all the time on these boards. There's actually some validity to that objection too.


Last edited by Lafleurs Guy: 03-30-2012 at 04:52 PM.
Lafleurs Guy is offline   Reply With Quote
Old
03-30-2012, 04:55 PM
  #37
Mathletic
Registered User
 
Mathletic's Avatar
 
Join Date: Feb 2002
Location: St-Augustin, Québec
Country: Canada
Posts: 11,748
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
Again, the CORSI stat has not impressed me. I look at the CORSI leaders and they don't correspond to the best players. Brett Hull never had the puck but he was one of the best scorers of all-time.

You make a blanket statement like puck possession wins hockey games... how true is this? What is there to back this up?

You come out with a statement like 'puck possession wins hockey games'... well, maybe as a general rule that's true. But it certainly doesn't always seem to be the case.
sorry for the quick edit on your post to get quotes

the main evidence that goes in favor of puck possession is that on average, teams that get hit more are positively correlated with wins, whereas teams that hit more are negatively correlated with wins. Main explanation I see for this is time of possession.

In general I agree with you on CORSI stats. Even though they tell something, I'm not sure it tells a whole lot. The main thing that comes out of CORSI IIR is that they found shots was a better predictor of future goal scoring performances than actual goals are. Meaning shot% is not of great value.

As in anything, stats won't help you predict 100% of the outcome. There will always be randomness in anything that involve humans. Stats will only help you so far. Even though analytics don't have all the answers, the question to answer to answer is does it help make better decisions.

Mathletic is online now   Reply With Quote
Old
03-30-2012, 05:03 PM
  #38
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 20,298
vCash: 500
Quote:
Originally Posted by Cyclones Rock View Post
This would be a very intriguing experiment. If such a staff could create models which were effective, the savings from not signing a bad UFA could more than justify their expense.

Back when I played high school hockey-when dinosaurs roamed the earth-I distinctly remember a hit which altered my (and a few other players) play that couldn't be quantified by any metric of which I'm aware. (The change in play could be quantified but the reason behind it couldn't be. Not in any standard manner.)

In a game early in my senior season, a guy (who will appear in a linked video), charged from the blue line in full stride, left his feet and knocked out about 3 or 4 teeth of one of my teammates. I had a tendency to hold on to the puck a bit more than a should have (I guess I would now call myself a puck possession forward). After seeing this hit, for the rest of the game-and every time I played against this guy thereafter-my style abruptly changed from puck hog (I mean puck possession) to passing teammate (I mean 'dump and skate away'.

Intimidation can't really be measured, but it's very real. A lot of players won't go into the corners as zestfully or as often as they should do to fear. From lower level hockey to the NHL. Quantifying stuff like this is beyond the realm of micro stats, imo.

And here's my former tormentor. R.I.P. He was one scary dude. LOL

http://www.youtube.com/watch?v=68Pmcq-ZA2M
I suspect Gordie Howe (who was skilled) certainly gave himself some extra room with those elbows of his. And when you look at this stat line, it will be excellent for sure. The microstats will tell you that he produces at a high rate... but they won't tell you why.

And that is key.

Somebody can just sit there and say well, he's just really skilled... okay. How much of that though was due to people not wanting to touch him or go near his elbows? The guy gives himself space out there. And if you're going to measure one player against another it's tough to see this on a spreadsheet. It may show up in other stats but it's awfully hard to capture the impact of the fear that Howe instilled in his opponents.

Lafleurs Guy is offline   Reply With Quote
Old
03-30-2012, 05:06 PM
  #39
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 20,298
vCash: 500
Quote:
Originally Posted by airic000 View Post
What I'm saying is that you can't use the Billy Beane method to stack your team with players that will give you a statistical advantage when you don't even know how they will react when playing with one another. Sure there are stats on how Player X may react with his teammates on one team but how does that help me if I want to bring him to my team? He'll have a whole new set of linemates. All I'm saying is that it's completely unpredictable how players will interact with one another before you actually see them playing together.
That's another interesting point. Will two Adam Oates' type guys be better playing with one another or are you better off pairing an Oates with a Hull? If you're building a team how do you decide? If you over emphasize puck possession then you could miss out on the next Hull or Bossy.

Lafleurs Guy is offline   Reply With Quote
Old
03-30-2012, 05:06 PM
  #40
Roke
Registered User
 
Roke's Avatar
 
Join Date: Jul 2003
Location: Winnipeg
Country: Canada
Posts: 1,885
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
I agree. But see my question immediately below:

Problem is that giveaways and takeaways vary hugely from rink to rink. And that leads to another big problem in analytics for a fluid sport such as hockey... subjectivity. What constitutes a giveaway for one person doesn't for another... if the NHL can't get it right themselves, then it casts doubt on the ability to 'count things' doesn't it?
I addressed this slightly up thread (with hits too... I assume you're working your way down) but as they're counted in the NHL hits, takeaways, and giveaways are basically useless when it comes to any kind of evaluation. If the Habs (or anyone) wanted to use them and get something out of it they would have to put come to a definition of each event and track them in-house.

Quote:

Sure, but if you don't have an answer to the objection it's going to breed a lot more skepticism. We've seen that with CORSI and Gomez. Brett Hull is another example.
You come out with a statement like 'puck possession wins hockey games'... well, maybe as a general rule that's true. But it certainly doesn't always seem to be the case. And when I hear somebody defend the hell out of Gomez because of a microstat, it casts doubt on that person's credibility. If you can't come up with a reasonable answer to somebody's objection... you can't blame them for being skeptical. And you can't sit there and argue... well analysts probably have thought this out already... okay, if that's the case what's the response?
I'm cherry-picking the Hull thing and I hope I get this posted before everyone piles on, but Hull led the league in shots on goal on 3 occasions and regularly broke 300 shots from 1988-1997. We don't have the data to be absolutely sure, but he was probably a good possession player because of the huge number of shots he could get off. He may not have carried the puck from end-to-end but when he was on the ice his team likely had the puck more than the oppposition and his work away from the puck and working with linemates probably meant he would be a high Corsi player... probably.

As with most things, it's a matter of actually having the data in order to truly be sure.

Gomez has dropped into an elevator shaft this season and has been terrible. I think his play being different than the shot-metric based expectations MathMan, myself (I picked him for my fantasy team because of the Pacioretty-Gomez-Gionta trio lighting it up last season), and others had are for a number of reasons.

For one, we under-estimated how good Pacioretty is -he's close to being a star winger if not already and Gionta hasn't been a slouch there either. Gionta and Pacioretty were probably driving the play for Gomez last year just like Pacioretty is doing the main driving on the Desharnais line this year. Gomez' getting injured multiple times hasn't helped but his play has dropped off to a point (and his cap-hit) mean that there's no point waiting around any more.

If Molson wants to win Gomez won't be here next year. My expectation was to have him around for this season soaking up minutes (like an old, soft-tossing starting pitcher who just eats up innings) while Eller and Desharnais were given more time to develop. With Gomez' early injuries that didn't happen, his play has taken a considerable hit since his second injury of the season, and Eller and Desharnais are actually doing better than I expected. Had Gomez been able to soak up the minutes and the disastrous December-February Gauthier moves not been made the Habs probably would have been better off. It didn't happen.

Roke is offline   Reply With Quote
Old
03-30-2012, 05:09 PM
  #41
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,891
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
Is this always true? Show us the evidence.
In the long term, pretty much. Look up at the various links I posted for any number of posts that demonstrate this.

Quote:
Originally Posted by Lafleurs Guy View Post
You make a blanket statement like puck possession wins hockey games... how true is this? What is there to back this up? Why does a player like Gomez who seems to have the puck so much suck so bad?
Not to turn this into another pointless Gomez debate, but essentially: because Gomez doesn't suck nearly as much as advertised. He's not worth 7 million. He's struggled with injuries this year. But there was an interesting point made here by someone who has no dog in the fight: http://jetsnation.ca/2012/1/26/is-sc...underrated-yes

(That said, for the sake of completeness, since his last return from injury Gomez has stepped into an elevator shaft and hasn't been driving the play North. There's any number of reasons why that could be, of course -- going from a style that specifically de-emphasizes puck possession to the possibility that he might be playing hurt to the possibility that he basically hit the breaking point right there.)

Interestingly, notions that Gomez sucks have always been revolved around his lack of traditional points. This is basically admitting the notion that total points (which are a metric) are paramount. Not only is this also a statistical argument, it's not a particularly good one, as the whys and wherefores of total points and how they might fluctuate have been heavily studied.

(Other complaints about him tend to be rooted in that, either offered up as explanations for lack of counting stats, or basically projections of flaws into his games that aren't there (such as "taking too many penalties").)

Quote:
Originally Posted by Lafleurs Guy View Post
You come out with a statement like 'puck possession wins hockey games'... well, maybe as a general rule that's true. But it certainly doesn't always seem to be the case.
Here's another point that's very important to realize. In hockey, there is not ever going to ever be any such thing as a method that says "if you do X, then you are 100% guaranteed to win", regardless of whether it's via analytics or not. There is actually an important lesson here that studying analytics will tell you, though it's certainly possible to learn this very important lesson by other means:

It's hockey. **** happens.

More formally, Hockey is a game with a lot of variances, what might be called "luck" in a mathematical sense. Analytics will help you maximize your chances of winning games in the future. They will tell you what things will, in the long run, lead to better odds of winning. But better odds is all you are ever going to get.

You can never get to a point where you're guaranteed victory. The better team wins often but often doesn't win. You are going to lose games despite playing better than the other guys. Sometimes this will happen several times in a row. Sometimes you could be consistently playing worse than your opponents and keep winning games, so much so that you end up leading your conference, and then things get back to normal and you end up in the cellar and people wonder what the heck happened to your team, and in reality nothing did except regression to the mean, and you were just a bad team getting results over its head.

In a formal sense, at the NHL level, "puck possession and variance/mathematical luck and a bit of goaltending talent and a teeny tiny bit of shooting talent and some amount of special team talent wins hockey games". But because puck possesion and variance are by far the biggest factors, and teams don't have control over variance (it's why it's called that), "puck possession wins hockey games" is meaningful shorthand.

Also, puck possession wins hockey games, but goaltending can sure as hell lose them.

MathMan is offline   Reply With Quote
Old
03-30-2012, 05:11 PM
  #42
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,891
vCash: 500
On that subject, I am kicking myself for not remembering this awesome blog post, which nicely explains regression to the mean and variance in a way non-mathheads can relate to.

http://blogs.thescore.com/nhl/2012/0...nd-transience/

MathMan is offline   Reply With Quote
Old
03-30-2012, 05:11 PM
  #43
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 20,298
vCash: 500
Quote:
Originally Posted by Mathletic View Post
sorry for the quick edit on your post to get quotes

the main evidence that goes in favor of puck possession is that on average, teams that get hit more are positively correlated with wins, whereas teams that hit more are negatively correlated with wins. Main explanation I see for this is time of possession.
My initial thought would be that puck possession is a good thing. I mean why wouldn't it be? But how important is it? Where does a player like Hull fit into this?
Quote:
Originally Posted by Mathletic View Post
In general I agree with you on CORSI stats. Even though they tell something, I'm not sure it tells a whole lot. The main thing that comes out of CORSI IIR is that they found shots was a better predictor of future goal scoring performances than actual goals are. Meaning shot% is not of great value.

As in anything, stats won't help you predict 100% of the outcome. There will always be randomness in anything that involve humans. Stats will only help you so far. Even though analytics don't have all the answers, the question to answer to answer is does it help make better decisions.
You and I are on the same page. I think it's a useful tool that casts light on areas that you might not have known or thought about. I don't think it provides you with a complete picture though.

I use baseball as a counter example simply because I think that it's such a stat driven sport that you really can almost rely solely on these stats to build a club. I don't think we get nearly as clear a picture in hockey and the purpose of this thread is not to challenge or dismiss these analytics but merely to try to get an understanding or where it stands in relation to importance on building a club and how much weight they should be given.

Lafleurs Guy is offline   Reply With Quote
Old
03-30-2012, 05:15 PM
  #44
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,891
vCash: 500
Quote:
Originally Posted by Mathletic View Post
The main thing that comes out of CORSI IIR is that they found shots was a better predictor of future goal scoring performances than actual goals are. Meaning shot% is not of great value.
Corsi is actually a proxy for puck possession. If you used a stopwatch and counted possession and zone time, you'd come up with numbers that are very close to Corsi.

Fenwick (which is Corsi except without counting blocked shots either way) actually has a better correlation to winning and to scoring chances. Although obviously, it's usually pretty close to Corsi.

Corsi also has a much larger sample size than goals (which are, when you think about it, pretty rare) so it can give you information more rapidly than goals can.

MathMan is offline   Reply With Quote
Old
03-30-2012, 05:18 PM
  #45
Roke
Registered User
 
Roke's Avatar
 
Join Date: Jul 2003
Location: Winnipeg
Country: Canada
Posts: 1,885
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
I use baseball as a counter example simply because I think that it's such a stat driven sport that you really can almost rely solely on these stats to build a club. I don't think we get nearly as clear a picture in hockey and the purpose of this thread is not to challenge or dismiss these analytics but merely to try to get an understanding or where it stands in relation to importance on building a club and how much weight they should be given.
I don't agree with the baseball assertion. The Blue Jays under JP Riccardi de-emphasized scouting and ended up not doing well - though it didn't help that their GM couldn't multi-task, gave out horrible contracts, and hamstrung their drafting and development by only drafting College guys and basically not signing international free agents.

In baseball, the best organizations (Boston, present-day Toronto, Tampa Bay come to mind) use a combination of both. Scouting tends to be focused a lot on skill and mechanics while you compliment that with the statistical analysis of a player's performance. If the two don't agree on a player or team that's an impetus to look closer and try to figure out why. There's a reason the Toronto Blue Jays at least doubled their scouting staff when Alex Anthopoulos took over as GM - scouting still has value.

Roke is offline   Reply With Quote
Old
03-30-2012, 05:26 PM
  #46
MathMan
Registered User
 
Join Date: Jan 2006
Posts: 16,891
vCash: 500
On the Hull question: modern hockey analytics mostly concern themselves with the modern NHL. There's a limited amount of work done on pre-2005 players, because the game was different and the data pretty much isn't there. Applying this reasoning to other eras of hockey is therefore fraught with peril -- the parity wasn't the same, goaltending was different, systems weren't as well-developped... (It goes without saying that it's also risky to apply these same analytics to non-NHL leagues without caution.)

That said, Brett Hull was a shot machine. A shot machine over ten years, which included several that didn't include Adam Oates. And I don't think the game has changed to the extent that one could achieve 300+ shots in a season without one's team owning the puck a fair deal.

So either Hull had a positive impact on puck possession (while he was remembered for his one-timers, that doesn't mean it was the only thing he did), either he played with players who were adept at obtaining the puck and dishing it (had a positive impact on puck possession), or played in a system that heavily favored it in an era where systems weren't quite so developped.

But in any case -- the dude had four shots a game. Some years he had almost five. That just doesn't happen if your team doesn't have the puck a lot. Someone on that team, be it Brett Hull or someone else, had to generate enough puck possession for him to get so many shots.

MathMan is offline   Reply With Quote
Old
03-30-2012, 05:30 PM
  #47
Lafleurs Guy
Registered User
 
Lafleurs Guy's Avatar
 
Join Date: Jul 2007
Posts: 20,298
vCash: 500
Quote:
Originally Posted by Roke View Post
I addressed this slightly up thread (with hits too... I assume you're working your way down) but as they're counted in the NHL hits, takeaways, and giveaways are basically useless when it comes to any kind of evaluation. If the Habs (or anyone) wanted to use them and get something out of it they would have to put come to a definition of each event and track them in-house.
Okay. That works for pro scouting and teams that play against the Habs. How do you do it for minor leaguers that you want to build your team with?

I know... something is better than nothing. Simply saying that it's not all that easy to do and requires a tremendous amount of effort.
Quote:
Originally Posted by Roke View Post
I'm cherry-picking the Hull thing and I hope I get this posted before everyone piles on, but Hull led the league in shots on goal on 3 occasions and regularly broke 300 shots from 1988-1997. We don't have the data to be absolutely sure, but he was probably a good possession player because of the huge number of shots he could get off. He may not have carried the puck from end-to-end but when he was on the ice his team likely had the puck more than the oppposition and his work away from the puck and working with linemates probably meant he would be a high Corsi player... probably.
Truth is... I don't know. I think he's an interesting example in theory because so often he would just sit in the slot and Oates would feather a pass to him and he'd score. Hull wasn't a particularly good skater and didn't deke out players the way Denis Savard did but he was a hell of a scorer.

When I look at Gomez, his CORSI numbers are among the best in the league, but he never produces. I'm sorry but I don't buy the "luck" explanation, it's simply not good enough. Year in and year out there will be players with lower CORSI who will produce more than that guy does. So how useful is that stat to begin with?

Quote:
Originally Posted by Roke View Post
As with most things, it's a matter of actually having the data in order to truly be sure.

Gomez has dropped into an elevator shaft this season and has been terrible. I think his play being different than the shot-metric based expectations MathMan, myself (I picked him for my fantasy team because of the Pacioretty-Gomez-Gionta trio lighting it up last season), and others had are for a number of reasons.

For one, we under-estimated how good Pacioretty is -he's close to being a star winger if not already and Gionta hasn't been a slouch there either. Gionta and Pacioretty were probably driving the play for Gomez last year just like Pacioretty is doing the main driving on the Desharnais line this year. Gomez' getting injured multiple times hasn't helped but his play has dropped off to a point (and his cap-hit) mean that there's no point waiting around any more.
But see... I could've told you that he was the best player on that line. I said long ago, Pac was carrying Gomez. That's because I would watch the games. Pac produces with anyone because he shoots the puck and can actually get it in the net.
Quote:
Originally Posted by Roke View Post
If Molson wants to win Gomez won't be here next year. My expectation was to have him around for this season soaking up minutes (like an old, soft-tossing starting pitcher who just eats up innings) while Eller and Desharnais were given more time to develop. With Gomez' early injuries that didn't happen, his play has taken a considerable hit since his second injury of the season, and Eller and Desharnais are actually doing better than I expected. Had Gomez been able to soak up the minutes and the disastrous December-February Gauthier moves not been made the Habs probably would have been better off. It didn't happen.
His play sucked before the injury though. You have CORSI to back you up but I've got the point totals. You say it's luck but I say he sucks. At the end of the day if a guy isn't producing and youv'e cast him in an offensive role, you've got to rid yourself of him. I don't care what the microstats say, you can't let the tail wag the dog. 1 year without a goal just isn't acceptable.

Lafleurs Guy is offline   Reply With Quote
Old
03-30-2012, 05:40 PM
  #48
Mathletic
Registered User
 
Mathletic's Avatar
 
Join Date: Feb 2002
Location: St-Augustin, Québec
Country: Canada
Posts: 11,748
vCash: 500
Quote:
Originally Posted by MathMan View Post
Corsi is actually a proxy for puck possession. If you used a stopwatch and counted possession and zone time, you'd come up with numbers that are very close to Corsi.

Fenwick (which is Corsi except without counting blocked shots either way) actually has a better correlation to winning and to scoring chances. Although obviously, it's usually pretty close to Corsi.

Corsi also has a much larger sample size than goals (which are, when you think about it, pretty rare) so it can give you information more rapidly than goals can.
thanks for the precision, I'm no expert on those stats

Mathletic is online now   Reply With Quote
Old
03-30-2012, 05:56 PM
  #49
Talks to Goalposts
Registered User
 
Join Date: Apr 2011
Location: Edmonton
Country: Canada
Posts: 3,560
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post

I think there's more to a big hit than an empty display of power. If you watch Lindros, Kariya and countless others get crushed by Scott Stevens you're going to be a lot more aware of your surroundings when playing the Devils. It could cause you to fumble the puck or miss an opportunity. The guy doesn't necessarily even have to throw another hit to make an impact. The impact was made when the opposing player was watching footage of Lindros being carted off the ice from a hit two years ago and he doesn't want to be sitting there drooling in his helmet for the rest of his life.
I can see the case for intimdation being a factor but I think the premise that if a player has a talent for that it won't show up quantitatively. If your measuring overall play and your intimidation factor causes opposing player to play poorly it will show up in terms of you playing well. So like skating or stick handling if its an ability that affects the game in your favour it will show up in a measure of overall play. If it doesn't actually cause the opposition to play measurably worse then it wasn't actually a factor in the game.

And if you are good a intimidation but don't get good overall results then the rest of your game is probably so poor that intimidation is pointless for the team because your giving it away elsewhere.

I think there is a real danger of double counting attributes when you want to count things like intimidation over and above ability to measurably effect the game.

Quote:
Originally Posted by Lafleurs Guy View Post
Again, the CORSI stat has not impressed me. I look at the CORSI leaders and they don't correspond to the best players. Brett Hull never had the puck but he was one of the best scorers of all-time.
Corsi isn't supposed to be used as a straight measurement of how good a player is, that would be absurd for a number of reasons. Its used as a measure of dominance within circumstances. For example Subban and Weber had similar corsi/fenwick etc. last season but Subban's circumstances were by far and away more difficult.

Its also a matter of thinking less about individual brilliance and more defining a player's worth in the context of how well his team does with him on the ice and working backwards from that. Its a matter of the idea that team's that are good at this are good teams for whom we can predict that they will do well in the future, lines that are good at this are good lines that we expect good things from and that the value of a player is in how much their presence makes teams and lines good.


Your Hull example actually a good one for illustrating some of the concepts inherent to what we're are discussing here.

1. When we're discussing possession in this context its really a shorthand for a bigger concept than just having the puck on your stick. Its more of a complex of possession and territorial play involved in controlling who has the puck and where. With the overall goal of your team having the puck in scoring position more often than your opponent. So a forward that's excellent at getting in scoring position like Hull is going to have "possession" value even if he doesn't handle the puck much himself. On the other side you can have defensive defensemen that have "possession" value even if they don't handle the puck much if they prevent opportunities for the opposition.

2. Amount of Goals/Points doesn't necessarily have a direct correlation with value. There are plenty of NHL players that score a lot at the expense of defense. And if you give up more than you create then you're hurting your team even if you're one of the top scorers. A player's value has to correlate with how much they help you outscore the opposition rather than just how much they help you score. For example, Kessel might be one of the top even strength scorers in the league this year but I don't think he's near the top for even strength contribution to winning. This of course ignores special teams but that's its own situation and should be considered somewhat independently from even strength play.

3. "Finishing" talent (likelihood of scoring with a given chance) does exist at the NHL level but a) its really hard to say for sure who actually has it and who is just "hot" or "cold." b) there doesn't seem to be as much of a spread on it, if you're good enough to make the NHL chances are you're part of the very large mass of players that aren't noticeably better or worse at it.

So for one thing its hard to say for sure if this sort of thing would undervalue a Brett Hull without having any real information about how it actually values Brett Hull. And if it did rate a top goal scorer like Hull low that should also lead to some consideration on how much value he does provide.

Talks to Goalposts is offline   Reply With Quote
Old
03-30-2012, 06:08 PM
  #50
Talks to Goalposts
Registered User
 
Join Date: Apr 2011
Location: Edmonton
Country: Canada
Posts: 3,560
vCash: 500
Quote:
Originally Posted by Lafleurs Guy View Post
I suspect Gordie Howe (who was skilled) certainly gave himself some extra room with those elbows of his. And when you look at this stat line, it will be excellent for sure. The microstats will tell you that he produces at a high rate... but they won't tell you why.

And that is key.

Somebody can just sit there and say well, he's just really skilled... okay. How much of that though was due to people not wanting to touch him or go near his elbows? The guy gives himself space out there. And if you're going to measure one player against another it's tough to see this on a spreadsheet. It may show up in other stats but it's awfully hard to capture the impact of the fear that Howe instilled in his opponents.
Its important to note that for these kinds of analytical discussions "skill" and "talent" refer to any abilities a player has that helps to repeatably produce the results desired. So ability to make space with elbows, to the extent it helps drive results, counts as a skill.

So in that sense one could say for example that Gorges is currently a more skilled player than say Kaberle despite that Kaberle may have more "skill" in the colloquial sense.

This leads to a certain democratic meritocracy in terms of evaluating players, in the particulars of method mean less than the ability to achieve the results desired.



Also, going back to the comparison to baseball, stuff like counting goals/points is a lot better in a game like baseball where offense and defense are distinct. Hockey isn't like that its a game of simultaneous offense and defense. This is exactly why evaluations need to be based off an idea of differentials rather than absolute numbers because that's the only hope you have of modelling the complexities of a fluid game.


Last edited by Talks to Goalposts: 03-30-2012 at 06:13 PM.
Talks to Goalposts is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 04:26 AM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. ©2014 All Rights Reserved.