Friday, February 02, 2007
Flames Game Day
The Flames host the Columbus Blue Jackets tonight in a Classic Trap GameTM (7PM MT, RSN West). Good old TSN Ice Chips has the forward lines from practice:
I'm not going to get too worked up about lines, as Playfair generally uses the same Blender as MacT, but I really question the wisdom of taking Iginla away from Langkow, as to be blunt, they've been GDF'n successful this season. To take a quick look backwards for a sec: in 2001/02, Jarome Iginla won the Art Ross Trophy and Craig Conroy finished 15th in the NHL in scoring. Their points-per-game then were: JI - 1.17, CC - 0.93.
If you jack those rates up to adjust for the fact that there's 11% more scoring in the NHL this season than in '01/02, the pt/Gm rates are: JI - 1.30, CC - 1.03
Here in '06/07, Iginla's and Langkow's rates are: JI - 1.36, DL - 1.06
The difference is small, but what that says is that Iginla/Langkow have been more productive this season than Iginla/Conroy were five years ago, when Iginla led the league in scoring by 6 points (and by 11 goals). Plus of course Conroy is now 35, not 30.
I guess the question for the coach is simply, "What are you hoping to achieve?" If it's to increase the threat from the 2nd line, fine, but you don't want it coming entirely at the expense of the production of the Iginla line, whoever else may be on it with the big man.
A couple of links I wanted to pass on: Allan Muir (who has definitely become one of my favourites) has a thoughtful piece at SI.com; regular readers of this site will recognize the sentiment therein.
Also, Alan Ryder's latest stats piece at the Globe's site is 85% of a solid article (again, regular readers of this site, or a number of other hockey blogs, will be familiar with the subject matter). He talks about evaluating goaltenders, and very ably leads the readers through the drawbacks ("pollution") associated with the more common benchmarks like Wins and Goals Against Average. He even has a very smart but simple line that I think I'll steal in the future:
Indeed. He notes then, by way of building to his grand finale, that "There are also some things you cannot even see with conventional statistics." Good, but here's what follows:
Why? Why?!? Why would you write a piece devoted to explaining (and effectively, to boot), yet finish off with "Here's some mystery numbers from my computer!" Put another way: is there anyone who is very interested in the concept of a Shot Quality-Neutral Save Percentage, but not at all interested in knowing where it comes from or how it's derived? He couldn't have taken two sentences to explain that a shot from the top of the crease is more likely to go in than one from centre ice, and the goalies who disproportionately face the former are going to have a worse SV% than the ones who disproportionately face the latter?
This attitude isn't exactly unique to Ryder. There is a certain segment of stats guys who believe that their work is more interesting or valuable if they appear to have invented a magic formula. They're wrong, because it makes it impossible for them to show both other statheads and the general fan that what they've done has any meaning.
Prediction! It's exceedingly frustrating, but Calgary and Columbus always play close games. The last time they played when the lead was more than 1 going into the final minute was November 1, 2003. But tonight, Iginla returns, so at least one Flame won't be looking ahead to tomorrow night's Canucks game.
Call it Calgary 2 (Iginla, Lombardi) Columbus 1 (Nash on a 2-man PP). Go Flames.
Tanguay - Conroy - Iginla
Huselius - Langkow - Moss
Amonte - Lombardi - Nilson
Friesen - Yelle - Ritchie
Huselius - Langkow - Moss
Amonte - Lombardi - Nilson
Friesen - Yelle - Ritchie
I'm not going to get too worked up about lines, as Playfair generally uses the same Blender as MacT, but I really question the wisdom of taking Iginla away from Langkow, as to be blunt, they've been GDF'n successful this season. To take a quick look backwards for a sec: in 2001/02, Jarome Iginla won the Art Ross Trophy and Craig Conroy finished 15th in the NHL in scoring. Their points-per-game then were: JI - 1.17, CC - 0.93.
If you jack those rates up to adjust for the fact that there's 11% more scoring in the NHL this season than in '01/02, the pt/Gm rates are: JI - 1.30, CC - 1.03
Here in '06/07, Iginla's and Langkow's rates are: JI - 1.36, DL - 1.06
The difference is small, but what that says is that Iginla/Langkow have been more productive this season than Iginla/Conroy were five years ago, when Iginla led the league in scoring by 6 points (and by 11 goals). Plus of course Conroy is now 35, not 30.
I guess the question for the coach is simply, "What are you hoping to achieve?" If it's to increase the threat from the 2nd line, fine, but you don't want it coming entirely at the expense of the production of the Iginla line, whoever else may be on it with the big man.
A couple of links I wanted to pass on: Allan Muir (who has definitely become one of my favourites) has a thoughtful piece at SI.com; regular readers of this site will recognize the sentiment therein.
Also, Alan Ryder's latest stats piece at the Globe's site is 85% of a solid article (again, regular readers of this site, or a number of other hockey blogs, will be familiar with the subject matter). He talks about evaluating goaltenders, and very ably leads the readers through the drawbacks ("pollution") associated with the more common benchmarks like Wins and Goals Against Average. He even has a very smart but simple line that I think I'll steal in the future:
You can spend a great deal of time watching hockey and never be able to objectively identify that kind of skill differential with your eyes alone. In fact we are likely to be distracted by differences in style. This is why we need to measure things.
Indeed. He notes then, by way of building to his grand finale, that "There are also some things you cannot even see with conventional statistics." Good, but here's what follows:
One of these is 'shot quality'. I have developed a way to measure the 'degree of difficulty' of shots faced by a goaltender.
Why? Why?!? Why would you write a piece devoted to explaining (and effectively, to boot), yet finish off with "Here's some mystery numbers from my computer!" Put another way: is there anyone who is very interested in the concept of a Shot Quality-Neutral Save Percentage, but not at all interested in knowing where it comes from or how it's derived? He couldn't have taken two sentences to explain that a shot from the top of the crease is more likely to go in than one from centre ice, and the goalies who disproportionately face the former are going to have a worse SV% than the ones who disproportionately face the latter?
This attitude isn't exactly unique to Ryder. There is a certain segment of stats guys who believe that their work is more interesting or valuable if they appear to have invented a magic formula. They're wrong, because it makes it impossible for them to show both other statheads and the general fan that what they've done has any meaning.
Prediction! It's exceedingly frustrating, but Calgary and Columbus always play close games. The last time they played when the lead was more than 1 going into the final minute was November 1, 2003. But tonight, Iginla returns, so at least one Flame won't be looking ahead to tomorrow night's Canucks game.
Call it Calgary 2 (Iginla, Lombardi) Columbus 1 (Nash on a 2-man PP). Go Flames.
# posted by Matt @ 1:07 pm 
Comments:
There is a certain segment of stats guys who believe that their work is more interesting or valuable if they appear to have invented a magic formula. They're wrong, because it makes it impossible for them to show both other statheads and the general fan that what they've done has any meaning.
Oh boy, I could go on and on on that one. I agree 100%, and would push it even further. I'd say most of the hockey metrics stuff, while good, is being wasted by the inability or unwillingness to explain. I've been sitting on a post about this for months, but way too much of it is inaccessible.
I do like how even when using his own secret metric, he is still willing to gloss over the achievements of C. Mason and J.S. Giguere and jump straight to Marty Brodeur as a guy who should get "stong consideration for both the Vezina and Hart trophies".
Chapter 5: Is it time zone?
It wasn't as widely publicized, but I know Scott Niedermayer won it one month (probably only broadcast on the Ducks telecast) and didn't Crosby win one?
I'd say most of the hockey metrics stuff, while good, is being wasted by the inability or unwillingness to explain.
I don't think this is really the problem at all. It's just that many of the Oilogosphere types that have dabbled in this stuff have been doing so for a couple of years already on the message boards. At least that's why I don't spend much time explaining the analyses up front.
On top of that I think 99% of what I show are just based on GF/GA rates and point scoring rates. Also, correct me if I'm wrong, but the only guy out there with some magic formulas is Ryder and he does post the methodology as .pdf articles.
I should also mention that I have encouraged you in particular to ask questions in the comments of my last couple of data dump posts over at IOF. I'm pretty sure I've tried to explain and answer all of your questions. I operate under the assumption that anyone that wants to know will ask a question.
Basically, expecting to be forcefed this stuff is just wishful thinking. The information is available out there. I think that applies to Ryder, JavaGeek, the IOF gang and MC79's stuff anyway.
This is why we need to measure things.
But you have to control for things too, and unfortunatly for all the money puckers ou there, there's way too much shit going in a hockey game for there to ever be any meaningful statistical analysis beyond what we already have.
So in other words, hockey isn't baseball. Not even close.
Oh hell Rivers, that comment wasn't directed anywhere near you (or Tyler or the other IOF guys).
My frustration, such as it is, is that the few instances where hockey stats done for a bigger audience seem to be the worst and most nonsensical. (See Tyler's periodic comments on the NYTimes' ongoing efforts, for example).
there's way too much shit going in a hockey game for there to ever be any meaningful statistical analysis beyond what we already have.
That, sir, is an appalling lack of imagination. As for "hockey isn't baseball"? Uh, agreed.
As for "hockey isn't baseball"? Uh, agreed.
I mean baseball lends itself perfectly to statistical analysis, whereas hockey doesn't.
Oh hell Rivers, that comment wasn't directed anywhere near you (or Tyler or the other IOF guys).
I think he was talking to me, Matt, and my comment was directed at all those guys.
Basically, expecting to be forcefed this stuff is just wishful thinking.
You must have meant spoonfed. I get forcefed it all the time.
I mean baseball lends itself perfectly to statistical analysis, whereas hockey doesn't.
I don't understand still. Are you criticizing the fact that statistics are being collected in hockey or the fact that they are being looked at?
Matt: OK, I did assume that Andy's were directed my way though. Apparently correctly.
If we're talking about what has been published in the mainstream in general then I agree. The NY Times stuff was crazy and that study on fighting was as well. I'm not sure about Ryder's shot quality stuff, but he's definitely got a 71 page pdf on his Player Contribution number so he does have some of his methodology out there.
Andy: My email address is probably on IOF somewhere, but anyway it's crowsrmurder@hotmail.com. Go ahead and ask me for any clarifications you want. Like I said, pretty much everything I look at is just based on situational icetimes and GF/GA with little or no manipulation, but I guess you've missed something.
I don't understand still. Are you criticizing the fact that statistics are being collected in hockey or the fact that they are being looked at?
I think he means that baseball can be broken down very easily into a wide variety of discrete situations, each of which can be isolated and controlled for, whereas hockey's more fluid nature makes it a lot more difficult to shake out cause and effect. Reasonable assumptions can be made, mind you, but you don't get as clear a picture as you would from a collection of discrete events like baseball (or football, really).
Right, but to get back to his point:
there's way too much shit going in a hockey game for there to ever be any meaningful statistical analysis beyond what we already have.
I don't understand the criticism here; that we are on a futile exercise to better understand the components of winning so there should be no effort made therein?
I can think of several data elements that would help statistical analysis in hockey, especially shot-neutral-whatever-it-is:
a) was a goaltender screened on any shot?
b) did a shot occur on a breakaway or not?
c) was it an odd-man rush?
d) where on the net was it aimed?
e) was it deflected?
etc., etc.
Sure it's not baseball (and I don't assume anyone said it was), but there are still statistical measures for things more complex than baseball.
I don't understand the criticism here; that we are on a futile exercise to better understand the components of winning so there should be no effort made therein?
No. It's just damn near impossible to make most hockey actions statistically meaningful.
I'd love to elaborate, but unfortunately I'm at work. If you want me to write a long post on why I think trying to fit hockey into statistical models is ridiculous, I'd be more than happy to, but it'll will have to wait.
Heh, I was at work also, but back home now.
Maybe I'm just grumpy today--long business trip and a messy drive to the office.
But still, I'm not fond of the implication "there's nowhere to go, so stop trying."
I know hockey is tougher to model, but doesn't that mean we need to find better indicators / more measures / stronger correlations for things that happen on the ice, rather than just throwing our hands in the air?
I'm interested in hearing what McLea has to say about this - I hereby instruct Loxy and the BCIT people to leave him alone if they see him.
Can't wait to catch this game. Sure not my priority for games tomm. with the Pens taking on the caps but I love the Canadian rivalrys. They are second to none.
Kris
Anyone catch the very beginning of the broadcast, or otherwise know why A-Fence is out of the lineup tonight?
One shouldn't reject the most likely fault: an editor deemed the brief rundown of the origin of the stat as "too numerical" or "too cerebral" or some such thing.
Post a Comment
<< Home
There is a certain segment of stats guys who believe that their work is more interesting or valuable if they appear to have invented a magic formula. They're wrong, because it makes it impossible for them to show both other statheads and the general fan that what they've done has any meaning.
Oh boy, I could go on and on on that one. I agree 100%, and would push it even further. I'd say most of the hockey metrics stuff, while good, is being wasted by the inability or unwillingness to explain. I've been sitting on a post about this for months, but way too much of it is inaccessible.
I do like how even when using his own secret metric, he is still willing to gloss over the achievements of C. Mason and J.S. Giguere and jump straight to Marty Brodeur as a guy who should get "stong consideration for both the Vezina and Hart trophies".
Chapter 5: Is it time zone?
It wasn't as widely publicized, but I know Scott Niedermayer won it one month (probably only broadcast on the Ducks telecast) and didn't Crosby win one?
I'd say most of the hockey metrics stuff, while good, is being wasted by the inability or unwillingness to explain.
I don't think this is really the problem at all. It's just that many of the Oilogosphere types that have dabbled in this stuff have been doing so for a couple of years already on the message boards. At least that's why I don't spend much time explaining the analyses up front.
On top of that I think 99% of what I show are just based on GF/GA rates and point scoring rates. Also, correct me if I'm wrong, but the only guy out there with some magic formulas is Ryder and he does post the methodology as .pdf articles.
I should also mention that I have encouraged you in particular to ask questions in the comments of my last couple of data dump posts over at IOF. I'm pretty sure I've tried to explain and answer all of your questions. I operate under the assumption that anyone that wants to know will ask a question.
Basically, expecting to be forcefed this stuff is just wishful thinking. The information is available out there. I think that applies to Ryder, JavaGeek, the IOF gang and MC79's stuff anyway.
This is why we need to measure things.
But you have to control for things too, and unfortunatly for all the money puckers ou there, there's way too much shit going in a hockey game for there to ever be any meaningful statistical analysis beyond what we already have.
So in other words, hockey isn't baseball. Not even close.
Oh hell Rivers, that comment wasn't directed anywhere near you (or Tyler or the other IOF guys).
My frustration, such as it is, is that the few instances where hockey stats done for a bigger audience seem to be the worst and most nonsensical. (See Tyler's periodic comments on the NYTimes' ongoing efforts, for example).
there's way too much shit going in a hockey game for there to ever be any meaningful statistical analysis beyond what we already have.
That, sir, is an appalling lack of imagination. As for "hockey isn't baseball"? Uh, agreed.
As for "hockey isn't baseball"? Uh, agreed.
I mean baseball lends itself perfectly to statistical analysis, whereas hockey doesn't.
Oh hell Rivers, that comment wasn't directed anywhere near you (or Tyler or the other IOF guys).
I think he was talking to me, Matt, and my comment was directed at all those guys.
Basically, expecting to be forcefed this stuff is just wishful thinking.
You must have meant spoonfed. I get forcefed it all the time.
I mean baseball lends itself perfectly to statistical analysis, whereas hockey doesn't.
I don't understand still. Are you criticizing the fact that statistics are being collected in hockey or the fact that they are being looked at?
Matt: OK, I did assume that Andy's were directed my way though. Apparently correctly.
If we're talking about what has been published in the mainstream in general then I agree. The NY Times stuff was crazy and that study on fighting was as well. I'm not sure about Ryder's shot quality stuff, but he's definitely got a 71 page pdf on his Player Contribution number so he does have some of his methodology out there.
Andy: My email address is probably on IOF somewhere, but anyway it's crowsrmurder@hotmail.com. Go ahead and ask me for any clarifications you want. Like I said, pretty much everything I look at is just based on situational icetimes and GF/GA with little or no manipulation, but I guess you've missed something.
I don't understand still. Are you criticizing the fact that statistics are being collected in hockey or the fact that they are being looked at?
I think he means that baseball can be broken down very easily into a wide variety of discrete situations, each of which can be isolated and controlled for, whereas hockey's more fluid nature makes it a lot more difficult to shake out cause and effect. Reasonable assumptions can be made, mind you, but you don't get as clear a picture as you would from a collection of discrete events like baseball (or football, really).
Right, but to get back to his point:
there's way too much shit going in a hockey game for there to ever be any meaningful statistical analysis beyond what we already have.
I don't understand the criticism here; that we are on a futile exercise to better understand the components of winning so there should be no effort made therein?
I can think of several data elements that would help statistical analysis in hockey, especially shot-neutral-whatever-it-is:
a) was a goaltender screened on any shot?
b) did a shot occur on a breakaway or not?
c) was it an odd-man rush?
d) where on the net was it aimed?
e) was it deflected?
etc., etc.
Sure it's not baseball (and I don't assume anyone said it was), but there are still statistical measures for things more complex than baseball.
I don't understand the criticism here; that we are on a futile exercise to better understand the components of winning so there should be no effort made therein?
No. It's just damn near impossible to make most hockey actions statistically meaningful.
I'd love to elaborate, but unfortunately I'm at work. If you want me to write a long post on why I think trying to fit hockey into statistical models is ridiculous, I'd be more than happy to, but it'll will have to wait.
Heh, I was at work also, but back home now.
Maybe I'm just grumpy today--long business trip and a messy drive to the office.
But still, I'm not fond of the implication "there's nowhere to go, so stop trying."
I know hockey is tougher to model, but doesn't that mean we need to find better indicators / more measures / stronger correlations for things that happen on the ice, rather than just throwing our hands in the air?
I'm interested in hearing what McLea has to say about this - I hereby instruct Loxy and the BCIT people to leave him alone if they see him.
# posted by 
: February 02, 2007 5:04 pm
: February 02, 2007 5:04 pm Can't wait to catch this game. Sure not my priority for games tomm. with the Pens taking on the caps but I love the Canadian rivalrys. They are second to none.
Kris
# posted by : February 02, 2007 8:07 pm
: February 02, 2007 8:07 pm Anyone catch the very beginning of the broadcast, or otherwise know why A-Fence is out of the lineup tonight?
One shouldn't reject the most likely fault: an editor deemed the brief rundown of the origin of the stat as "too numerical" or "too cerebral" or some such thing.
Post a Comment
<< Home
