Toggle sidebar

2017-18 stats and underlying metrics thread [Mod: updated season]

Status
Not open for further replies.
Aavco Cup

Aavco Cup

"I can make you cry in this room"
Sep 5, 2013
37,630
10,440
http://www.extraskater.com

Offline
Sorry, Extra Skater is offline.
Click to expand...

Been off line since yesterday at least. I wonder if this is related at all to the NHL's change? Or technical trouble? Or as Friedman speculates.

@FriedgeHNIC: Have had a few questions about @ExtraSkater today. It was not an NHL-mandated shutdown. Maybe he was "Dellowed."
 
Last edited:
garret9

garret9

AKA#VitoCorrelationi
Mar 31, 2012
21,738
4,380
Vancouver
www.hockey-graphs.com
Darryl's a coder not an analyzer, so if he's been picked up it is to join a team's analytical team in support of analysts.

Should note that Darryl stopped replying to my emails about the same time he stopped tweeting...

There are multiple other options though of possible situations.
 
Last edited:

Whileee

Registered User
May 29, 2010
46,073
33,127
garret9 said:
I wrote on prospects, expectations, NHLEs, and probabilities:
http://morethan140charactershockey.blogspot.ca/2014/08/copps-and-prospects.html?spref=tw
Click to expand...

Interesting article, and I fully understand the concept of probabilities. But I also think it's important to consider more than a single variable to make projections. The statistical relationship for a complex outcome is rarely robust enough to rely on a single independent variable (like points). Have you done any assessments of the model fit? Is it even appropriate to use a linear regression model? Is a "bell curve" the best representation of the probability distribution?

There are a number of other variables that I think would be relevant, if quantifiable. These include:

- career trajectory (points, goals).
- scoring relative to team-mates (doesn't penalize players on low-scoring teams).
- size (Copp is reportedly 6'1" and over 200 pounds - Albert is 5'10").
- goal / assist ratio (Albert scored 6 goals as a Junior, with most of his points coming from assists).

Maybe these issues have already been considered and analyzed and the NHLE ppg has been shown to be robust, and that the model specification is appropriate.

Again, this isn't to denigrate statistical analysis, because I think it is likely to add a lot. My concern is that statistical analysis should go further than a lot of the more simplistic analyses and conclusions if we want it to add much to what scouts and fans already consider.
 
garret9

garret9

AKA#VitoCorrelationi
Mar 31, 2012
21,738
4,380
Vancouver
www.hockey-graphs.com
Whileee said:
Interesting article, and I fully understand the concept of probabilities. But I also think it's important to consider more than a single variable to make projections. The statistical relationship for a complex outcome is rarely robust enough to rely on a single independent variable (like points). Have you done any assessments of the model fit? Is it even appropriate to use a linear regression model? Is a "bell curve" the best representation of the probability distribution?

There are a number of other variables that I think would be relevant, if quantifiable. These include:

- career trajectory (points, goals).
- scoring relative to team-mates (doesn't penalize players on low-scoring teams).
- size (Copp is reportedly 6'1" and over 200 pounds - Albert is 5'10").
- goal / assist ratio (Albert scored 6 goals as a Junior, with most of his points coming from assists).

Maybe these issues have already been considered and analyzed and the NHLE ppg has been shown to be robust, and that the model specification is appropriate.

Again, this isn't to denigrate statistical analysis, because I think it is likely to add a lot. My concern is that statistical analysis should go further than a lot of the more simplistic analyses and conclusions if we want it to add much to what scouts and fans already consider.
Click to expand...

There is only one variable being discussed when we're discussing that variable. This is not a discussion on creating a highly-predictive model, but rather discussing trends in how higher scoring players tend to succeed more...

Why do people keep on acting like I am singularly focused on scoring? Am I not communicating well? (Totally possible... I'm a very meh writer)

Scoring isn't everything and Copp's game isn't reliant on scoring alone - This is a strawman argument, as that is not the point with using player's scoring rates and discussing prospect potential for success.
Click to expand...

and

I think his (edit: Copp's) defensive game and "intangibles" give him an edge on some of the players that "failed" on the above graph.
Click to expand...

Plus multiple other comments with the key words like "trend".

I know these other variables exist and matter. I've never once took the stance that these numbers are saying everything. I'm just adding one more layer of information.

This all came about when I said: Hey, Slater scored at a higher rate and was a highly valued and projected prospect (1st rounder). He was valued highly by scouts. Was a leader on a team. And was successful in carving out a long career as a NHL player but was never a plus player. So, be careful in getting overly excited on players. **** happens.
That's it. Caveats added on doesn't change that point.



To answer some of your other questions, NHLEs seem to point out that the model is mostly linear regression is most appropriate, although Volman has found that there is a limit and a tail off. The probability curve is likely skewed. Anyways, I linked in my article to a good article though that discusses this further.

Edit
As a little add on I think this is a neat aspect to look at it:
nhlchart_medium.JPG

http://assets.sbnation.com/assets/666667/nhlchart_medium.JPG
 
Last edited:

Whileee

Registered User
May 29, 2010
46,073
33,127
garret9 said:
There is only one variable being discussed when we're discussing that variable. This is not a discussion on creating a highly-predictive model, but rather discussing trends in how higher scoring players tend to succeed more...

Why do people keep on acting like I am singularly focused on scoring? Am I not communicating well? (Totally possible... I'm a very meh writer)



and



Plus multiple other comments with the key words like "trend".

I know these other variables exist and matter. I've never once took the stance that these numbers are saying everything. I'm just adding one more layer of information.

This all came about when I said: Hey, Slater scored at a higher rate and was a highly valued and projected prospect (1st rounder). He was valued highly by scouts. Was a leader on a team. And was successful in carving out a long career as a NHL player but was never a plus player. So, be careful in getting overly excited on players. **** happens.
That's it. Caveats added on doesn't change that point.

To answer some of your other questions, NHLEs seem to point out that the model is mostly linear regression is most appropriate, although Volman has found that there is a limit and a tail off. The probability curve is likely skewed. Anyways, I linked in my article to a good article though that discusses this further.

Edit
As a little add on I think this is a neat aspect to look at it:
nhlchart_medium.JPG

http://assets.sbnation.com/assets/666667/nhlchart_medium.JPG
Click to expand...

Thanks for the further notes.

I know that you take a more nuanced view to these issues, but I'm not sure that many others do. I also think that quantitative folks like you who understand the underlying methods are in the best position to examine the structure of associations and to come up with better models to associate the independent variables with important outcomes (I don't actually like the notion of a "predictive" model since we are always talking about a range of possible outcomes across a probability distribution. If one wants to talk about "prediction", then I'd like to see measures of statistical error such as confidence intervals).

Getting back to Copp, I agree that we should be cautious in our enthusiasm. But there are a number of factors that strike me as being favourable in terms of other variables. In particular, he is still fairly early in his development and has made very big strides in the past two years, based on all accounts. Last year he led his team in goal-scoring, and was basically tied in PPG as a sophomore, after having a very strong second-half of his Freshman year. Personally, I really liked his play at the WJHC. His size is another asset.

If he is able to develop into a player like Slater, I'll be very happy. After all, he was a #104 pick overall, so he would be beating the odds.
 

jetkarma*

Guest
Whileee said:
Thanks for the further notes.

I know that you take a more nuanced view to these issues, but I'm not sure that many others do. I also think that quantitative folks like you who understand the underlying methods are in the best position to examine the structure of associations and to come up with better models to associate the independent variables with important outcomes (I don't actually like the notion of a "predictive" model since we are always talking about a range of possible outcomes across a probability distribution. If one wants to talk about "prediction", then I'd like to see measures of statistical error such as confidence intervals).

Getting back to Copp, I agree that we should be cautious in our enthusiasm. But there are a number of factors that strike me as being favourable in terms of other variables. In particular, he is still fairly early in his development and has made very big strides in the past two years, based on all accounts. Last year he led his team in goal-scoring, and was basically tied in PPG as a sophomore, after having a very strong second-half of his Freshman year. Personally, I really liked his play at the WJHC. His size is another asset.

If he is able to develop into a player like Slater, I'll be very happy. After all, he was a #104 pick overall, so he would be beating the odds.
Click to expand...

Here's a debate or an issue and I would like a yes or a no from posters here . While it isn't meant to be a definitive analysis of two players , or that it must be predictive of a future out come , I think as a piece it should be included. That it is a legitimate part of a comparison for two players , or something to look at and see if at the same stage , results were similar or not.

Both players played for the same team in their 20th year, although 6 or 7 years apart. They also both played for the same countries WJC team. One player totaled 39 points in 43 total games , the other player had 34 points in 38 games.

Not a perfect test or so large as to be a pretty defacto indicator , but I think it is more than valid to use as part of an overall evaluation. Sure strength of teams come into play , league scoring etc. but interested to have posters opinions on whether they think that year, the last available year for one of the players, has merit as part of an overall evaluation and measurable.

Not ideal or or perfect but worthwhile to look at and include , yes or no.
 
garret9

garret9

AKA#VitoCorrelationi
Mar 31, 2012
21,738
4,380
Vancouver
www.hockey-graphs.com
Whileee said:
Thanks for the further notes.

I know that you take a more nuanced view to these issues, but I'm not sure that many others do. I also think that quantitative folks like you who understand the underlying methods are in the best position to examine the structure of associations and to come up with better models to associate the independent variables with important outcomes (I don't actually like the notion of a "predictive" model since we are always talking about a range of possible outcomes across a probability distribution. If one wants to talk about "prediction", then I'd like to see measures of statistical error such as confidence intervals).

Getting back to Copp, I agree that we should be cautious in our enthusiasm. But there are a number of factors that strike me as being favourable in terms of other variables. In particular, he is still fairly early in his development and has made very big strides in the past two years, based on all accounts. Last year he led his team in goal-scoring, and was basically tied in PPG as a sophomore, after having a very strong second-half of his Freshman year. Personally, I really liked his play at the WJHC. His size is another asset.

If he is able to develop into a player like Slater, I'll be very happy. After all, he was a #104 pick overall, so he would be beating the odds.
Click to expand...
Yup. Agreed on all fronts.
jetkarma said:
Here's a debate or an issue and I would like a yes or a no from posters here . While it isn't meant to be a definitive analysis of two players , or that it must be predictive of a future out come , I think as a piece it should be included. That it is a legitimate part of a comparison for two players , or something to look at and see if at the same stage , results were similar or not.

Both players played for the same team in their 20th year, although 6 or 7 years apart. They also both played for the same countries WJC team. One player totaled 39 points in 43 total games , the other player had 34 points in 38 games.

Not a perfect test or so large as to be a pretty defacto indicator , but I think it is more than valid to use as part of an overall evaluation. Sure strength of teams come into play , league scoring etc. but interested to have posters opinions on whether they think that year, the last available year for one of the players, has merit as part of an overall evaluation and measurable.

Not ideal or or perfect but worthwhile to look at and include , yes or no.
Click to expand...
Not ideal and not perfect. Worthwhile depends on what you are doing. Limiting your comparison (and combining their NCAA and WJC points together) is only meaningful in an anecdotal fashion, not analytical fashion, because the one player is not representative of the population nor the probabilistic outcomes. It is a possibility but one player does not tell you the probability.

Either way, I still struggle to see how Pacioretty's dominant and over point per game as a freshmen year where everyone knew he was beyond college level, is comparable to Copp's sophomore year. Especially given it is a full year relative to draft earlier, although not far in age and the season prior he was over point per game in USHL.

Either way, a point doesn't make a trend just like Copp's worst comparable doesn't necessitate he will fail either (like guys who outscored Copp in NCAA and WJC but busted).
 

jetkarma*

Guest
Same age , same teams , same overall production.

One team was much better than the other UM team , one player had very unique circumstances leading to his 20th year season .

You think it has no bearing , I absolutely do . You think Slater is a worthwhile comparable , I think there are better ones . Like the much more recent UM grads I have mentioned.

Not a perfect measure as I have always maintained but a very legitimate comparable to use. You've mentioned similar situations for players playing on the same JR team years apart, not sure why one holds water and one doesn't in your opinion.
 
garret9

garret9

AKA#VitoCorrelationi
Mar 31, 2012
21,738
4,380
Vancouver
www.hockey-graphs.com
jetkarma said:
Same age , same teams , same overall production.

One team was much better than the other UM team , one player had very unique circumstances leading to his 20th year season .

You think it has no bearing , I absolutely do . You think Slater is a worthwhile comparable , I think there are better ones . Like the much more recent UM grads I have mentioned.

Not a perfect measure as I have always maintained but a very legitimate comparable to use. You've mentioned similar situations for players playing on the same JR team years apart, not sure why one holds water and one doesn't in your opinion.
Click to expand...

Sigh. Reading my friend.

Not that it has no comparison but:
Me: a point doesn't make a trend
Click to expand...
In other words, one comparable (Slater, Pacioretty, or the tons of other guys that I have in the original link) are meaningless, all of them, 100% meaningless, on their own. Not just an imperfect measure, but meaningless.
It's the general trend when you combine them all that has some meaning.

And again, I did not use Slater as a worthwhile comparable:
Me: Hey, Slater scored at a higher rate and was a highly valued and projected prospect (1st rounder). He was valued highly by scouts. Was a leader on a team. And was successful in carving out a long career as a NHL player but was never a plus player. So, be careful in getting overly excited on players. **** happens.
That's it. Caveats added on doesn't change that point.
Click to expand...

Also, what?
You: You've mentioned similar situations for players playing on the same JR team years apart, not sure why one holds water and one doesn't in your opinion.
Click to expand...
Probabilities: as many players as possible.
Possibilities: one player.
When I do one player, it's usually to show how a player has scored similarly/differently at different levels; a way to give added context. It is not to say that they are equal or better at any means. It's not a comparable. A one-to-one player is only significant in saying that one has outscored the other.
When I want to talk about probabilities, I use numbers like NHLEs as then you are comparing them against THOUSANDS of players.

I'm not sure what's holding you back from understanding this?
 
Mortimer Snerd

Mortimer Snerd

You kids get off my lawn!
Sponsor
Jun 10, 2014
57,344
29,108
Joe Hallenback

Joe Hallenback

Moderator
Mar 4, 2005
15,384
21,544
A lot of teams use different groups. I expect that some of these teams are late to the party

I know the Oilers use Darkhorse Analytics to provide them with data. I expect a lot of teams use companies or people who work in a similar field
 
Aavco Cup

Aavco Cup

"I can make you cry in this room"
Sep 5, 2013
37,630
10,440
Greg Wyshynski (puck daddy) just stated on Sportsnet that he believes one reason the leafs hired Daryl Metcalf was to keep the intellectual property out of the hands of their competitors. Unfortunately it keeps it out of our (fans) hands as well. There has to be a race on now to fill the void.
 
YWGinYYZ

YWGinYYZ

Registered User
Jul 3, 2011
28,480
7,117
Toronto
Aavco Cup said:
Greg Wyshynski (puck daddy) just stated on Sportsnet that he believes one reason the leafs hired Daryl Metcalf was to keep the intellectual property out of the hands of their competitors. Unfortunately it keeps it out of our (fans) hands as well. There has to be a race on now to fill the void.
Click to expand...

Yup, there probably is. ;)
 
Aavco Cup

Aavco Cup

"I can make you cry in this room"
Sep 5, 2013
37,630
10,440
YWGinYYZ said:
Yup, there probably is. ;)
Click to expand...

http://www.russianmachineneverbreaks.com/2014/08/19/extraskater-is-dead-lets-build-a-new-stat-site/

I expect ExtraSkater.com to be gone for good. Metcalf likely won’t have time to devote to it, and it might represent a conflict of interest to work on it. With that hiring, the Maple Leafs have taken away the single best resource for hockey fans and geeks and writers and coaches. Plus the other 29 GMs.

We can’t allow that. We need to replace Extra Skater.
Click to expand...
 
Status
Not open for further replies.

Ad

Latest posts

Upcoming events

Ad

Ad

Top Bottom