HFBoards

Go Back   HFBoards > NHL Western Conference > Pacific Division > Edmonton Oilers
Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
Notices

Oilers analytic/advanced stat thread

Reply
 
Thread Tools
Old
12-06-2016, 05:56 PM
  #76
Aceboogie
Registered User
 
Aceboogie's Avatar
 
Join Date: Aug 2012
Posts: 26,742
vCash: 500
Quote:
Originally Posted by IAGTTAYM View Post
I did the exercise in this video a few months ago: https://youtu.be/_w1IKTPrTNw

It's pretty shocking how much stuff you miss while watching a game.
How is this not bigger. This is awesome!

This sums up my thoughts so much in one video. The game is just so fast and we miss a bunch of stuff. We catch the main just of things but the smaller stuff together can add up to the reason a goal is scored is missed



here is is embedded for others to do

Aceboogie is offline   Reply With Quote
Old
12-06-2016, 06:00 PM
  #77
mkwong268
Registered User
 
Join Date: Dec 2011
Posts: 122
vCash: 50
Quote:
Originally Posted by jumptheshark View Post
Hope I word this correctly.

As of late I have noticed that many people talk about Analytic or advanced stats but I think not all understand them. It is not as simple as 1 + 1 = 2 here. I have an advance degree from London School of Economic(Take five to chuckle and remember--this place is where I come to relax and turn off my brain) and analytics plays a large part of my degree as in business and how numbers are crunched across the board and seven ways from Sundays. (Right now I building a budget based upon my forecast based upon business analytics for how BREXIT will affect all hotels in my chain) For the numbers being tossed out I think people need to explain their interpretation of the numbers and the how and why they got their decision. To explain analytics more and more business are using SWOT or a hockey version of STEEPLE or PEST(these version factor in where they were developed, amount of games they played, international play, amount of games played per year, injured suffered, type of equipment used)

Just tossing out numbers and saying there is the proof--is not how analytics work. Explaining what the numbers mean is the proof and how you got the numbers to make your decision.

For those he keep using analytics do you have advance statistical training or an advanced degree in data collection and analysis method? Most don't(sorry) most have downloaded an excel spread sheet or created one based upon other peoples information and after typing in stats--they get numbers and think is the conclusion and is proof. That is just step 1 of larger process to understand analytics and their use. Data Collection is the easiest part--it is the method used to analyse the data gathered where many people fail to grasp the complex nature of the project.

Not picking on any in particular, I just see people posting small sample sizes and proof of analytics of a certain player or team.

Here are some(small example) of the data collection that is done for D-men
1) Time on ice
2) Partner
3) Shift length(per period and per game)
4) Special teams play(both PP and SH)
5) Play on back to back nights and game per 7 days
6) Quality of competition
7) Time between shifts
8) Time in offensive and defensive zones
9) Quality of line mates beyond that of D-Partner
10) Breakdown of each shift (Shifts are usually broken down into 10 second segments) Break down of period play--each period has 3 ten minute segments-first ten minutes of a period, last ten minutes and 5 minute to 15 minute)
11) Continues play(this is explained best when we have say 5 to 10 minutes without a whistle and a D man may have 3 to 7 shifts during that time)

Just having stats and not explaining the methodology put into is were many people who talk analytics fall down/

Will give three examples.

1) Kris Russel--Corsi sucked--but when you look what lines he played against and who his partner was, situations he played and line mates--it clears the picture a bit
2) Jeff Petry. Petry was and is a 3 to 6 d-man who the oilers kept playing as a top pairing D-man on most nights. His perfect Ice time should have been between 15m to 18 minutes per night, but on many nights he was playing 20+ minutes a night and on back to back to nights. More then that he got worn down. the oilers expected too much too soon from him and his numbers were bad and well it got ugly. He went to Montreal--they changed what TYPE of partner he had, gradually increased his ice time to and limit his shifts
3) Adam Larsson. Lets be honest. We lost a lot posters in the 2 hours after this trade. Posters saw his numbers and went ape ****. Most had never seen the guy play more then two games a year. I saw him play as a 17 year old in Skelleftea and he blew my mind. He is not fancy but he is good and his numbers do not show for the most part how good he has been.

Yeah, I know--will get the Did not read gif for this post. but it is an observation that not everyone who talks about analytics actually understands them, can not put them in context or explain methodology behind how the numbers were concluded
I feel like the biggest problem with hockey analytics is how it's being used in player evaluation by the media. Shot attempts (Corsi) became popular because it was the best (good but not great R^2) at predicting team performance (objectively measurable by points in the standings). People then tried to use it as a measurement of player performance, where it doesn't work quite as well. A player's individual and on ice shot attempt numbers has some correlation to the team's performance but it's only a small piece of the puzzle. There's research constantly being done in trying to better evaluate players, and a lot of models are able to beat the eye test alone, but it's still no where close to being perfect, and I'm speaking from experience of working with NHL teams on hockey analytics projects on and off for the last three years. I wished more people understood that hockey statistics is meant to be used in conjunction with the eye test and not in place of it, at least not until we can perfectly model the game

mkwong268 is offline   Reply With Quote
Old
12-06-2016, 06:02 PM
  #78
Fourier
Registered User
 
Fourier's Avatar
 
Join Date: Dec 2006
Location: Waterloo Ontario
Country: Canada
Posts: 12,475
vCash: 500
Quote:
Originally Posted by Spawn View Post
Here's where I have some issues with some of these advanced stats.

High danger scoring chances suggests that it tracks scoring chances. But it doesn't track that. It tracks shots on net within a certain distance from the net. It makes the assumption that every shot within X distance of the net is "equal" and every shot outside of that distance is "equal." There is a hell of a lot more that goes into a scoring chance than simply where the puck is shot from. How quickly the puck is shot, what kind of traffic is in front of the net, who is shooting, and how well the puck is shot etc. So when someone argues that Nuge is creating scoring chances at a higher rate than any point in his career based on these stats, that isn't a true statement. A true statement would be that Nuge is getting more shots from closer to the net than any other point in his career.

I legitimately find a lot of these stats for interesting to read. But I take issue when they are used to draw such strong conclusions about players one way or the other. These things can certainly be useful, but any stat to be accurate needs to be narrow in its focus. So you can only really draw narrow conclusions from it.
I think you are perfectly correct in questioning the ability of any of these stats to give anywhere close to a complete picture of a players game. But each of the stats can tell you certain things that the eyes may not record accurately. In the case of the SC and HDSC stats there is a positive correlation between these stats and predictability of goals scored, but it is not as high as one might expect given some of the conclusions that are drawn from it.

On the other hand what they can do is provide pretty strong evidence to counter the assertion that Nuge's SH% is a result of shooting muffins from the perimeter. Since the stats are zone based while one cannot actually determine the "quality" of a shot one can get a very good picture of where the majority are coming from. Nuge's stats in this are suggest that he is not simply staying on the perimeter but is in fact getting into areas where success is typically higher or a fairly regular basis.

Fourier is offline   Reply With Quote
Old
12-06-2016, 06:04 PM
  #79
jumptheshark
McDavid Headquarters
 
jumptheshark's Avatar
 
Join Date: Oct 2003
Location: Lord of HFBOARDS
Country: United Nations
Posts: 74,929
vCash: 9240
Quote:
Originally Posted by Aceboogie View Post
How is this not bigger. This is awesome!

This sums up my thoughts so much in one video. The game is just so fast and we miss a bunch of stuff. We catch the main just of things but the smaller stuff together can add up to the reason a goal is scored is missed



here is is embedded for others to do
Every so often I get to sit with scouts here in Europe and the one thing I learned is what scouts look at and for. I sat with two scouts from the Sens( I played with one many many years ago). He was watching the play and the second was looking at what was happening behind the play and benches. He explained to me why was watching the D-men positioning. how closely they watched the play down the ice and adapted to what they saw and whether or not they actually watched the play closely(not as crazy as it sounds) and he was watching the coach and what players he had to coach shift to shift.

__________________
**Avatar approved by the powers that be***

I am the KING of Alternative Facts as any hockey fan is
jumptheshark is offline   Reply With Quote
Old
12-06-2016, 06:07 PM
  #80
Fourier
Registered User
 
Fourier's Avatar
 
Join Date: Dec 2006
Location: Waterloo Ontario
Country: Canada
Posts: 12,475
vCash: 500
Quote:
Originally Posted by mkwong268 View Post
I feel like the biggest problem with hockey analytics is how it's being used in player evaluation by the media. Shot attempts (Corsi) became popular because it was the best (good but not great R^2) at predicting team performance (objectively measurable by points in the standings). People then tried to use it as a measurement of player performance, where it doesn't work quite as well. A player's individual and on ice shot attempt numbers has some correlation to the team's performance but it's only a small piece of the puzzle. There's research constantly being done in trying to better evaluate players, and a lot of models are able to beat the eye test alone, but it's still no where close to being perfect, and I'm speaking from experience of working with NHL teams on hockey analytics projects on and off for the last three years. I wished more people understood that hockey statistics is meant to be used in conjunction with the eye test and not in place of it, at least not until we can perfectly model the game
This is a very solid post. The problem tends to be that the discussion on advanced stats tends to be extremely polarizing. They are neither useless or the perfect revealer of truth. But when they are used badly it really does no service to the field.

Fourier is offline   Reply With Quote
Old
12-06-2016, 06:09 PM
  #81
Aceboogie
Registered User
 
Aceboogie's Avatar
 
Join Date: Aug 2012
Posts: 26,742
vCash: 500
Quote:
Originally Posted by jumptheshark View Post
Every so often I get to sit with scouts here in Europe and the one thing I learned is what scouts look at and for. I sat with two scouts from the Sens( I played with one many many years ago). He was watching the play and the second was looking at what was happening behind the play and benches. He explained to me why was watching the D-men positioning. how closely they watched the play down the ice and adapted to what they saw and whether or not they actually watched the play closely(not as crazy as it sounds) and he was watching the coach and what players he had to coach shift to shift.
Thats interesting! But that adds even more questions about fans scouting reports. I dont think a lot of them are doing this type of stuff. Unless you tape and rewatch an entire game its hard to watch a D away from the puck, then also watch the puck and vice versa. You get glimpses of both but never both in full detail

Aceboogie is offline   Reply With Quote
Old
12-06-2016, 06:12 PM
  #82
Aceboogie
Registered User
 
Aceboogie's Avatar
 
Join Date: Aug 2012
Posts: 26,742
vCash: 500
Quote:
Originally Posted by mkwong268 View Post
I feel like the biggest problem with hockey analytics is how it's being used in player evaluation by the media. Shot attempts (Corsi) became popular because it was the best (good but not great R^2) at predicting team performance (objectively measurable by points in the standings). People then tried to use it as a measurement of player performance, where it doesn't work quite as well. A player's individual and on ice shot attempt numbers has some correlation to the team's performance but it's only a small piece of the puzzle. There's research constantly being done in trying to better evaluate players, and a lot of models are able to beat the eye test alone, but it's still no where close to being perfect, and I'm speaking from experience of working with NHL teams on hockey analytics projects on and off for the last three years. I wished more people understood that hockey statistics is meant to be used in conjunction with the eye test and not in place of it, at least not until we can perfectly model the game
Absolutely this always the best approach unquestionable. I still find a discouraging lack of use of analytics in player discussions though, or the complete brush off of them if they are used. I realize the use is incorrect at times but so is pure scouting reports from eye test. Neither are perfect but I find only one is questioned

Aceboogie is offline   Reply With Quote
Old
12-06-2016, 06:19 PM
  #83
LaGu
Registered User
 
Join Date: Jan 2011
Location: Luxembourg
Posts: 3,375
vCash: 500
Quote:
Originally Posted by jumptheshark View Post
Hope I word this correctly.

As of late I have noticed that many people talk about Analytic or advanced stats but I think not all understand them. It is not as simple as 1 + 1 = 2 here. I have an advance degree from London School of Economic(Take five to chuckle and remember--this place is where I come to relax and turn off my brain) and analytics plays a large part of my degree as in business and how numbers are crunched across the board and seven ways from Sundays. (Right now I building a budget based upon my forecast based upon business analytics for how BREXIT will affect all hotels in my chain) For the numbers being tossed out I think people need to explain their interpretation of the numbers and the how and why they got their decision. To explain analytics more and more business are using SWOT or a hockey version of STEEPLE or PEST(these version factor in where they were developed, amount of games they played, international play, amount of games played per year, injured suffered, type of equipment used)

Just tossing out numbers and saying there is the proof--is not how analytics work. Explaining what the numbers mean is the proof and how you got the numbers to make your decision.

For those he keep using analytics do you have advance statistical training or an advanced degree in data collection and analysis method? Most don't(sorry) most have downloaded an excel spread sheet or created one based upon other peoples information and after typing in stats--they get numbers and think is the conclusion and is proof. That is just step 1 of larger process to understand analytics and their use. Data Collection is the easiest part--it is the method used to analyse the data gathered where many people fail to grasp the complex nature of the project.

Not picking on any in particular, I just see people posting small sample sizes and proof of analytics of a certain player or team.

Here are some(small example) of the data collection that is done for D-men
1) Time on ice
2) Partner
3) Shift length(per period and per game)
4) Special teams play(both PP and SH)
5) Play on back to back nights and game per 7 days
6) Quality of competition
7) Time between shifts
8) Time in offensive and defensive zones
9) Quality of line mates beyond that of D-Partner
10) Breakdown of each shift (Shifts are usually broken down into 10 second segments) Break down of period play--each period has 3 ten minute segments-first ten minutes of a period, last ten minutes and 5 minute to 15 minute)
11) Continues play(this is explained best when we have say 5 to 10 minutes without a whistle and a D man may have 3 to 7 shifts during that time)

Just having stats and not explaining the methodology put into is were many people who talk analytics fall down/

Will give three examples.

1) Kris Russel--Corsi sucked--but when you look what lines he played against and who his partner was, situations he played and line mates--it clears the picture a bit
2) Jeff Petry. Petry was and is a 3 to 6 d-man who the oilers kept playing as a top pairing D-man on most nights. His perfect Ice time should have been between 15m to 18 minutes per night, but on many nights he was playing 20+ minutes a night and on back to back to nights. More then that he got worn down. the oilers expected too much too soon from him and his numbers were bad and well it got ugly. He went to Montreal--they changed what TYPE of partner he had, gradually increased his ice time to and limit his shifts
3) Adam Larsson. Lets be honest. We lost a lot posters in the 2 hours after this trade. Posters saw his numbers and went ape ****. Most had never seen the guy play more then two games a year. I saw him play as a 17 year old in Skelleftea and he blew my mind. He is not fancy but he is good and his numbers do not show for the most part how good he has been.

Yeah, I know--will get the Did not read gif for this post. but it is an observation that not everyone who talks about analytics actually understands them, can not put them in context or explain methodology behind how the numbers were concluded
I actually do have university and professional background in statistics and of course you are right in that conclusions cannot be drawn without completely understanding the question and approach to data collection. But that is why it's a discussion. If you see someone drawing in your mind wrong conclusions you can engage them and reason, I have many times, and (to no ones surprise) I have also realised that I have been wrong on occasion(s)...

I enjoy hockey stats because it kind of takes me back to basics, and to be honest you can do a lot with an excel sheet. I find that breaking down stats to the "smallest" pieces and then put it back together again is the best, and most enjoying approach. Don't trust, verify. But I am on the dark side compared to you, physics... I keep a lot of my own stat sheets (yes mostly in excel) since on many occasions I do think the stats is flawed, or rather the way it is represented is not complete with regards to the question asked. For example Qualcomp which I generally don't like, but I won't get into that now, because there is a game on and I am awake (in europe)!!

LaGu is offline   Reply With Quote
Old
12-06-2016, 06:20 PM
  #84
Aceboogie
Registered User
 
Aceboogie's Avatar
 
Join Date: Aug 2012
Posts: 26,742
vCash: 500
Quote:
Originally Posted by Spawn View Post
Here's where I have some issues with some of these advanced stats.

High danger scoring chances suggests that it tracks scoring chances. But it doesn't track that. It tracks shots on net within a certain distance from the net. It makes the assumption that every shot within X distance of the net is "equal" and every shot outside of that distance is "equal." There is a hell of a lot more that goes into a scoring chance than simply where the puck is shot from. How quickly the puck is shot, what kind of traffic is in front of the net, who is shooting, and how well the puck is shot etc. So when someone argues that Nuge is creating scoring chances at a higher rate than any point in his career based on these stats, that isn't a true statement. A true statement would be that Nuge is getting more shots from closer to the net than any other point in his career.

I legitimately find a lot of these stats for interesting to read. But I take issue when they are used to draw such strong conclusions about players one way or the other. These things can certainly be useful, but any stat to be accurate needs to be narrow in its focus. So you can only really draw narrow conclusions from it.
Ok then give me a scouting report on all of Nuges shots and the danger of each one. Go by shot velocity, who was screening goalie and who tipped the puck. Why take issue it from it when used for analytics purposes but then fail to recognise that tracking that stuff yourself is next to impossible. Regarding RNH if you have a bias against him you might only remember the poor shots. I recall a time of times his shot handcuffed the goalie or nearly beat the goalie. Its a matter of opinion but if you want to prove that wrong youll have go back and go over every single shot

If you want to make the distinction between high danger chances and "shots from in close" thatd be a hell of a time waster because statistically, shots from in close are high danger. Not EVERY single time, but on average it is

Aceboogie is offline   Reply With Quote
Old
12-06-2016, 06:32 PM
  #85
jumptheshark
McDavid Headquarters
 
jumptheshark's Avatar
 
Join Date: Oct 2003
Location: Lord of HFBOARDS
Country: United Nations
Posts: 74,929
vCash: 9240
Quote:
Originally Posted by LaGu View Post
I actually do have university and professional background in statistics and of course you are right in that conclusions cannot be drawn without completely understanding the question and approach to data collection. But that is why it's a discussion. If you see someone drawing in your mind wrong conclusions you can engage them and reason, I have many times, and (to no ones surprise) I have also realised that I have been wrong on occasion(s)...

I enjoy hockey stats because it kind of takes me back to basics, and to be honest you can do a lot with an excel sheet. I find that breaking down stats to the "smallest" pieces and then put it back together again is the best, and most enjoying approach. Don't trust, verify. But I am on the dark side compared to you, physics... I keep a lot of my own stat sheets (yes mostly in excel) since on many occasions I do think the stats is flawed, or rather the way it is represented is not complete with regards to the question asked. For example Qualcomp which I generally don't like, but I won't get into that now, because there is a game on and I am awake (in europe)!!
Hi-lighted the a key point--how often do we actual discuss the "Number" in front to see if it stands up when looked at differently and method how the number was gotten to. More often then not, those who push Analytics give the result and expect it to stand without question

jumptheshark is offline   Reply With Quote
Old
12-06-2016, 06:38 PM
  #86
LaGu
Registered User
 
Join Date: Jan 2011
Location: Luxembourg
Posts: 3,375
vCash: 500
Quote:
Originally Posted by jumptheshark View Post
Hi-lighted the a key point--how often do we actual discuss the "Number" in front to see if it stands up when looked at differently and method how the number was gotten to. More often then not, those who push Analytics give the result and expect it to stand without question
That is true, but it does not stop you from questioning it.

Anyway, it's not that important for me to tell you or anyone what/how to post. I just think that there is a middle ground which we should seek when discussing the numbers, but it's either one extreme or the other. And right now we have really moved away from the point of the thread since we're discussing posters use of it and its merits instead of the numbers themselves.

LaGu is offline   Reply With Quote
Old
12-06-2016, 06:52 PM
  #87
Paralyzer008
Registered User
 
Paralyzer008's Avatar
 
Join Date: Jan 2008
Country: Canada
Posts: 10,058
vCash: 500
http://www.sportsnet.ca/hockey/nhl/k...ytics-critics/

Many quotes from Chiarelli regarding advanced stats and how they relate to Russell.

I'm...not a huge fan of the article. A few things seem like they're coming from a guy that isn't digging deep enough.

For example:

"Edmonton goalies have a .949 saves percentage while Russell is on the ice at even strength — 17th among NHL defencemen who have played 300 minutes this season. That number is tops on Edmonton, meaning the quality of shots Russell has surrendered this season are inferior to virtually every other Oilers defenceman. (The team’s save percentage overall is .915.)"

Hard to completely justify that argument when Russell had a .903 save percentage last season - why the big variance when he basically looks the same as last year's Russell in Calgary? I don't think Spector is accounting for the element of luck - didn't really point out any inferences from Russell's current PDO.

If you wanted to see what kind of quality of shots Russell's leaving on the ice, you'd need to look at high-danger chances for/against as a more accurate stat than his goalie's save percentage. Or, also, average shot distance against.


Last edited by Paralyzer008: 12-06-2016 at 06:59 PM.
Paralyzer008 is offline   Reply With Quote
Old
12-06-2016, 06:57 PM
  #88
Ol' Jase
No mas, Adam. No mas
 
Ol' Jase's Avatar
 
Join Date: Jul 2005
Location: Edmonton
Country: Canada
Posts: 10,273
vCash: 894
Quote:
Originally Posted by jumptheshark View Post
with Larsson many and I do mean many people said he was a product of the devils system and had Schneider behind him--I tried to warn people he was better then they knew
Yes, you did and you were absolutely right.

Ol' Jase is offline   Reply With Quote
Old
12-06-2016, 11:05 PM
  #89
MessierII
Registered User
 
Join Date: Aug 2011
Posts: 11,493
vCash: 500
Quote:
Originally Posted by Aceboogie View Post
No he uses it the same methodology. He justs collects his own data (which is exactly what NHL collects), calls it something different and then uses that

Sure the stats might be wrong in some occurances but you dont throw the baby out with the bathwater because the counter might be off a few shots. I have yet to see anyone on HF ever say "Yeah bullshi** shots were 30, I counted 20 thats it". If the shot counts were so inaccurate its weird noone has ever voiced this concern. I can see hits and takeaways but both are open to judgement as to whats a hit or whats a takeaway
Actually his primary evaluation tool is the equivalent of a scoring chance +\- he probably uses shot differentials to a degree but it's not the primary stat like it is in the analytics community.

MessierII is offline   Reply With Quote
Old
12-06-2016, 11:13 PM
  #90
MessierII
Registered User
 
Join Date: Aug 2011
Posts: 11,493
vCash: 500
Quote:
Originally Posted by Aceboogie View Post
Could you give an example of how 90% of people use it incorrectly? Quality of comp and quality of teammate and zone starts are ignored by some people yes but its far lower than 90%
Everyone cherry picks stats to skew them to their view. It's like the Seguin vs Crosby example I used earlier. Even with a multitude of stats you can skew them to re enforce whatever you want to say. It's what nearly every blogger in the analytics community does. I'll use the stats myself upon occasion but there's not a single blogger I enjoy reading. It's a circle jerk of smugness I can't deal with.

MessierII is offline   Reply With Quote
Old
12-07-2016, 11:35 AM
  #91
Paralyzer008
Registered User
 
Paralyzer008's Avatar
 
Join Date: Jan 2008
Country: Canada
Posts: 10,058
vCash: 500
http://www.hockeybuzz.com/blog/Matt-...nned/191/81191

Interesting counter-article to Spector's.

Paralyzer008 is offline   Reply With Quote
Old
12-07-2016, 12:40 PM
  #92
stoff
Registered User
 
stoff's Avatar
 
Join Date: Apr 2007
Country: Canada
Posts: 1,101
vCash: 500
Quote:
Originally Posted by Paralyzer008 View Post
http://www.sportsnet.ca/hockey/nhl/k...ytics-critics/

Many quotes from Chiarelli regarding advanced stats and how they relate to Russell.

I'm...not a huge fan of the article. A few things seem like they're coming from a guy that isn't digging deep enough.

For example:

"Edmonton goalies have a .949 saves percentage while Russell is on the ice at even strength — 17th among NHL defencemen who have played 300 minutes this season. That number is tops on Edmonton, meaning the quality of shots Russell has surrendered this season are inferior to virtually every other Oilers defenceman. (The team’s save percentage overall is .915.)"

Hard to completely justify that argument when Russell had a .903 save percentage last season - why the big variance when he basically looks the same as last year's Russell in Calgary? I don't think Spector is accounting for the element of luck - didn't really point out any inferences from Russell's current PDO.

If you wanted to see what kind of quality of shots Russell's leaving on the ice, you'd need to look at high-danger chances for/against as a more accurate stat than his goalie's save percentage. Or, also, average shot distance against.
I agree the article is not well done. I applaud Spector for attempting to dig into this stuff because he is squarely on the other side of the debate as a proponent of the eye test with very little time for advance statistics or analytics or whatever you want to call them. But he got alot of stuff in the article wrong.

First, like you said, he concludes that Russell must be giving up less quality shots. But then leaves it at that with no evidence to support it. It seems like irresponsible journalism. In fact there is publicly available research that contradicts his hypothesis completely. There is no demonstrable ability for defenceman to impact goaltender save percentage.

https://hockey-graphs.com/2016/03/09...visited-again/

Also he brings up the PDO of Russel at 104.1 and says this is 10th best in the NHL and again leaves it at that with no explanation. Again this is irresponsible. A high PDO infers a high degree of luck and that his results are likely unsustainable. But he doesn't touch on it.

That Matt Henderson piece that others have mentioned goes into it even further and i agree with it's premise. The stats that Chiarelli is leaning on don't stand up to scrutiny. If Russel is one of the best in the league at zone entries then why are the Oilers shot attempts, shots, scoring chances, etc all demonstrably lower when he's on the ice? This does not flow logically.

Anyways i appreciated that he took the time to attempt to show both sides of the debate but he did so in an entirely irresponsible manner. There is a ton of publicly available information, articles, studies, that digs into this stuff but he just attempted to make his own conclusions based on nothing.


Last edited by stoff: 12-07-2016 at 12:50 PM.
stoff is offline   Reply With Quote
Old
12-07-2016, 12:55 PM
  #93
belair
Registered User
 
belair's Avatar
 
Join Date: Apr 2010
Location: Canada
Country: Canada
Posts: 11,617
vCash: 68
Quote:
Originally Posted by stoff View Post
I agree the article is not well done. I applaud Spector for attempting to dig into this stuff because he is squarely on the other side of the debate as a proponent of the eye test with very little time for advance statistics or analytics or whatever you want to call them. But he got alot of stuff in the article wrong.

First, like you said, he concludes that Russell must be giving up less quality shots. But then leaves it at that with no evidence to support it. It seems like irresponsible journalism. In fact there is publicly available research that contradicts his hypothesis completely. There is no demonstrable ability for defenceman to impact goaltender save percentage.

https://hockey-graphs.com/2016/03/09...visited-again/

Also he brings up the PDO of Russel at 104.1 and says this is 10th best in the NHL and again leaves it at that with no explanation. Again this is irresponsible. A high PDO infers a high degree of luck and that his results are likely unsustainable. But he doesn't touch on it.

That Matt Henderson piece that others have mentioned goes into it even further and i agree with it's premise. The stats that Chiarelli is leaning on don't stand up to scrutiny. If Russel is one of the best in the league at zone entries then why are the Oilers shot attempts, shots, scoring chances, etc all demonstrably lower when he's on the ice? This does not flow logically.

Anyways i appreciated that he took the time to attempt to show both sides of the debate but he did so in an entirely irresponsible manner. There is a ton of publicly available information, articles, studies, that digs into this stuff but he just attempted to make his own conclusions based on nothing.
Because Kris Russell is the Edmonton Oilers top PKer and has literally zero PP time.

belair is offline   Reply With Quote
Old
12-07-2016, 01:22 PM
  #94
stoff
Registered User
 
stoff's Avatar
 
Join Date: Apr 2007
Country: Canada
Posts: 1,101
vCash: 500
Quote:
Originally Posted by belair View Post
Because Kris Russell is the Edmonton Oilers top PKer and has literally zero PP time.
That has no effect. Shot/scoring chance/corsi/whatever rates are based on 5 v 5 play. For example he is worst among oiler dmen in on ice Shots For per 60 at 5 v 5 and on ice shot attempts for per 60 according to stats.hockeyanalysis.com. So if he was good at generating zone entries this should lead to more shots. But the opposite occurs.

stoff is offline   Reply With Quote
Old
12-07-2016, 01:23 PM
  #95
Boom Boom Aho
Sooo Good
 
Boom Boom Aho's Avatar
 
Join Date: Sep 2006
Posts: 22,768
vCash: 500
Quote:
Originally Posted by Aceboogie View Post
Matt Benning

But Benning right now is one of the best possession D in the NHL. Here is where he ranks amongst regular D in different areas:

CF%: 2nd in NHL
FF%: 6th in NHL
SF%: 14th in NHL
GF%: 9th in NHL
SCF%: 1st in NHL
HDCF%: 5th in NHL
Playing really easy minutes by analytics are Muzzin-esque
Great thread BTW! Question, where do you get the numbers from? particularly the HDCF%?

Boom Boom Aho is offline   Reply With Quote
Old
12-07-2016, 02:42 PM
  #96
Aceboogie
Registered User
 
Aceboogie's Avatar
 
Join Date: Aug 2012
Posts: 26,742
vCash: 500
Quote:
Originally Posted by Boom Boom Aho View Post
Great thread BTW! Question, where do you get the numbers from? particularly the HDCF%?
http://www.naturalstattrick.com/playerteams.php

(then filters are defenseman and over 50 mins TOI)

Nautral stat trick is great. Although no per 60 which kinda sucks

Aceboogie is offline   Reply With Quote
Old
12-07-2016, 02:45 PM
  #97
Boom Boom Aho
Sooo Good
 
Boom Boom Aho's Avatar
 
Join Date: Sep 2006
Posts: 22,768
vCash: 500
Quote:
Originally Posted by Aceboogie View Post
http://www.naturalstattrick.com/playerteams.php

(then filters are defenseman and over 50 mins TOI)

Nautral stat trick is great. Although no per 60 which kinda sucks
Thanks! Much appreciated.

Boom Boom Aho is offline   Reply With Quote
Old
12-08-2016, 01:38 AM
  #98
McOkMcgoMcoil
Registered User
 
McOkMcgoMcoil's Avatar
 
Join Date: Aug 2005
Location: Edmonton, AB
Posts: 14,357
vCash: 500
In a pure statistical sense the problem with advanced stats in hockey is there is so little independent variation. Meaning players stats are the results of so many co-variations with other players plays. It is difficult if not impossible to tease out these co-variances. plays are just so intertwined.

I still think at the end of the days the best stat would be to just break down every moment of every game. Have a panel of the same hockey analysts all ranks how good of a play the player made. If it was a defence man trying to defend a 2 on 1, how well did they do on a scale of 1 to 10. Do this and average the results. You would get the best unbiased measure of how the player actually performs. This is pretty much impossible though, it would just take way to much time. So we are back with points and a bunch of stuff that isn't measured well and can't account for the interdependence of players.

Oh well, not all sports are like baseball.

McOkMcgoMcoil is offline   Reply With Quote
Old
12-08-2016, 09:52 AM
  #99
Aceboogie
Registered User
 
Aceboogie's Avatar
 
Join Date: Aug 2012
Posts: 26,742
vCash: 500
Quote:
Originally Posted by McOkMcgoMcoil View Post
In a pure statistical sense the problem with advanced stats in hockey is there is so little independent variation. Meaning players stats are the results of so many co-variations with other players plays. It is difficult if not impossible to tease out these co-variances. plays are just so intertwined.

I still think at the end of the days the best stat would be to just break down every moment of every game. Have a panel of the same hockey analysts all ranks how good of a play the player made. If it was a defence man trying to defend a 2 on 1, how well did they do on a scale of 1 to 10. Do this and average the results. You would get the best unbiased measure of how the player actually performs. This is pretty much impossible though, it would just take way to much time. So we are back with points and a bunch of stuff that isn't measured well and can't account for the interdependence of players.

Oh well, not all sports are like baseball.
Thatd be awesome but like completely impossible to do. Who would pay for this- I doubt any "expert" is going to spend 3/4 hours a day doing this for free. And those who would do it for free probably arent experts. You would also still get complications with the "stats"- experts are humans and value different things/ Maybe 1 expert is really versed at offense and values that way higher than he would for D zone play. Expert 2 is the opposite- so the stats would still have bias and still be hard to say are uniform over every team

Aceboogie is offline   Reply With Quote
Old
12-08-2016, 09:54 AM
  #100
phaedrusDH
Registered User
 
Join Date: Jul 2009
Posts: 1,281
vCash: 500
Quote:
Originally Posted by stoff
Also he brings up the PDO of Russel at 104.1 and says this is 10th best in the NHL and again leaves it at that with no explanation. Again this is irresponsible. A high PDO infers a high degree of luck and that his results are likely unsustainable. But he doesn't touch on it.
You said it yourself it is an inferrence. PDO does not necessarily mean a player is lucky, as can be affected be a lot of variables. you know that old say you gotta be good to be lucky? in Russell's case, a lot of shots against come from outside or long distances, so they should be easy to save, shouldn't they? he doesn't give up a whole lot in tight.

If you remove all skill effects and look at PDO for an individual player, it is a measure of past performance against league-wide expectations (i.e. sh% + s% = 100%). it does not imply that he is bound to regress. besides, all things being future probabilities are not based on past results.

Quote:
Originally Posted by stoff View Post
That has no effect. Shot/scoring chance/corsi/whatever rates are based on 5 v 5 play. For example he is worst among oiler dmen in on ice Shots For per 60 at 5 v 5 and on ice shot attempts for per 60 according to stats.hockeyanalysis.com. So if he was good at generating zone entries this should lead to more shots. But the opposite occurs.
i'm sorry, but does the whistle blow at the end of a PP?

once again, Russell's style results in a lot of shots and shot attempts against. he also gives up a lot of zone entries. but that doesn't mean he doesn't contribute on the other end of the ice. he's a give and take type player.


advanced stats are useful, but you have to look at them in the context of the game.

phaedrusDH 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 11:09 PM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"

vBulletin Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. @2017 All Rights Reserved.