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

[pucka lucka]

Yesterday Gabriel Desjardins -founder of Behind the Net & Arctic Ice Hockey, and former consultant for multiple NHL teams- commented on that NY Post article. I talked to him after about it.

Essentially his point goes like this:

Hockey stats "revolution" in the NHL is different than Baseball... and it has nothing to do with the style of the game.

Baseball had players that were vastly under appreciated and other vastly over appreciated. Analytics was able to turn the game upside down.

Events have been tracked by teams for decades. Team's have been analyzing events and video for years. There isn't any low hanging fruit 4th line player who is unappreciated and actually a top line player.

With hockey it's not the debate of analysis vs no analysis, as analytics have been going on in hockey for a while. The issue is good analysis vs far better analysis. For the most part team's are extraordinarily internal, and not just in terms of hiding their data. Teams essentially track and test metrics only with themselves and on themselves, and the testing part is very minimal.

This relates to the discussion earlier on this thread between Avco, Whilee, and I. Teams already make good decisions, like Chicago. This doesn't mean that the current movement online that has been vigorously testing things league wide couldn't improve those decisions. They also make bad ones, like the example I gave about situational save percentage they use actually performing worse than regular save percentage.

He adds that you won't be seeing the teams that hired guys turning over and becoming successes instantaneously for that reason.

This sounds like an education issue. A complete lack of understanding of the field of statistics. Probabilities and confidence levels are rocket surgery (they really aren't it just seems that way at first) to most people.

 

[garret9]

Doing some work on defensive pairs for the Jets and discussing on twitter.

Here are a few pics from my tweets:

 

[Joe Hallenback]

Isn't it obvious that the top 2 D men in scoring will have the highest GF% on the team? Or is there evidence that is not the case

 

[garret9]

Isn't it obvious that the top 2 D men in scoring will have the highest GF% on the team? Or is there evidence that is not the case

There is a relationship but not necessarily.

Goal% is just +/- but for 5v5 only... this is just FAR larger samples with multiple seasons, so you lose a lot of the white noise issues with goals.

Guys who score more, on average, are more talented and more talented players will tend to have better goal differentials, but this is not always true.

There are some players who score well that don't promote goal differentials.


For Byfuglien and Enstrom's situation, I don't think it's obvious to all people that if you expand the sample that Tobi-Buff improve the Jets' goal differential greatly. Many say they are not good players and hurt more than they help.

 

[ps241]

garret not sure if anyone told you yet but you got a huge plug from Westy on the Big Show this morning. He was referencing the LPF Corsi numbers from last night.

 

[garret9]

I did not know.
Thanks.

 

[tinyzombies]

Visitor here. Great thread. I come in peace.

How does the breakdown of video play into this. I read on TSN that teams are using a program called PUCKS where the computer analyzes each play and can isolate where the play happened on the ice, and the previous sequences and spit out stats on the spot. Not sure how it identifies players, but that is pretty mindblowing.

So, the value of a player could be measured against various sequences (and tactical variations to each sequence - and also identify what the other team is doing) leading up to scoring chances and compared to tactical play at various links in the sequence (which they will have studied through PUCKS no doubt, and then have an old hockey guy standing there with a bunch of numbers in his hand to verify the robot)?

So they would have stats for wrinkles in every zone?

 

[garret9]

Yes.

There are a few systems out there. SportsVu. PowerScout. PUCKS.

This would be one of the best ways to break down the hows and whys for certain numbers.
You could use this to figure out how to deploy your players in areas that they excel in.
You could use this to figure out what areas you need to concentrate on with teaching.
You could use this to figure out what players should be placed with what players to optimize "chemistry".

Etc.

 

[tinyzombies]

Yes.

There are a few systems out there. SportsVu. PowerScout. PUCKS.

This would be one of the best ways to break down the hows and whys for certain numbers.
You could use this to figure out how to deploy your players in areas that they excel in.
You could use this to figure out what areas you need to concentrate on with teaching.
You could use this to figure out what players should be placed with what players to optimize "chemistry".

Etc.

The application could then feasibly simulate sequences, like a video game. They could actually play out the video game before the game.

So, clearly, each player is microchipped (at least the home team).

We might one day see a penalty for illegal microchipping?

 

[ps241]

I did not know.
Thanks.

Yea he was talking about the data you produced for the game last night then he mentioned you by name and said you were a really smart young guy that has been on H&L and he mentioned they should get you on the morning show. Sounds like he is reading your stuff and drinking the kool aid on advanced stats now. They got into an interesting debate about whether Fenwich or Corsi is better?

 

[Mortimer Snerd]

Yesterday Gabriel Desjardins -founder of Behind the Net & Arctic Ice Hockey, and former consultant for multiple NHL teams- commented on that NY Post article. I talked to him after about it.

Essentially his point goes like this:

Hockey stats "revolution" in the NHL is different than Baseball... and it has nothing to do with the style of the game.

Baseball had players that were vastly under appreciated and other vastly over appreciated. Analytics was able to turn the game upside down.

In hockey, events have been tracked by teams for decades. Teams have been analyzing events and video for years. There isn't any low hanging fruit 4th line player who is unappreciated and actually a top line player.

With hockey it's not the debate of analysis vs no analysis, as analytics have been going on in hockey for a while. The issue is good analysis vs far better analysis. For the most part team's are extraordinarily internal, and not just in terms of hiding their data. Teams essentially track and test metrics only with themselves and on themselves, and the testing part is very minimal.

This relates to the discussion earlier on this thread between Avco, Whilee, and I. Teams already make good decisions, like Chicago. This doesn't mean that the current movement online that has been vigorously testing things league wide couldn't improve those decisions. They also make bad ones, like the example I gave about situational save percentage they use actually performing worse than regular save percentage.

He adds that you won't be seeing the teams that hired guys turning over and becoming successes instantaneously for that reason.

I didn't take that article as being about the 'advanced stats' themselves or their application. All I took from it was the author bemoaning the absence of more advanced stats and analysis being made available by the league to the public. We should not have needed to rely on a site like extraskater to provide consumer level information. Some of the things that individual teams are doing secretively should be standardized and provided by the league. This response seems to be looking at trees when the article was about the forest. Or am I missing something?

 

[garret9]

I didn't take that article as being about the 'advanced stats' themselves or their application. All I took from it was that the author bemoaning the absence of more advanced stats and analysis being made available by the league to the public. We should not have needed to rely on a site like extraskater to provide consumer level information. Some of the things that individual teams are doing secretively should be standardized and provided by the league. This response seems to be looking at trees when the article was about the forest. Or am I missing something?

Desjardins is "retired" from the analytical community. He mostly went into a rant that slowly veered off topic but was good points.

I thought it was a good thing to grab and discuss his thoughts.
Especially since he's had an inside look on how multiple teams operate.
Also because he founded AIH, sorta, and was basically who I followed into this gig.

 

[Mortimer Snerd]

Desjardins is "retired" from the analytical community. He mostly went into a rant that slowly veered off topic but was good points.

I thought it was a good thing to grab and discuss his thoughts.
Especially since he's had an inside look on how multiple teams operate.
Also because he founded AIH, sorta, and was basically who I followed into this gig.

Fair enough then. I misunderstood.

 

[Gil Fisher]

With ExtraSkater gone, where am I going this year for friendly stats presentation and pretty bubbles?

 

[garret9]

Me

 

[Joe Hallenback]

Having a talk with some guys at work about Corsi in regards to what Doughty said about it.

One of the questions asked was

If I am a player on a line and we have possession in the offensive zone for a minute but we don't register an attempt on net and the opposing gets the puck, carries it out and dumps it in on net and registers a shot. Do we as a line and player get a negative corsi ?

 

[pucka lucka]

Having a talk with some guys at work about Corsi in regards to what Doughty said about it.

One of the questions asked was

If I am a player on a line and we have possession in the offensive zone for a minute but we don't register an attempt on net and the opposing gets the puck, carries it out and dumps it in on net and registers a shot. Do we as a line and player get a negative corsi ?

yes. if a line plays like that a lot, someone might want to tell them they are supposed to try to score.

 

[Gm0ney]

Having a talk with some guys at work about Corsi in regards to what Doughty said about it.

One of the questions asked was

If I am a player on a line and we have possession in the offensive zone for a minute but we don't register an attempt on net and the opposing gets the puck, carries it out and dumps it in on net and registers a shot. Do we as a line and player get a negative corsi ?

If the dump-in is recorded as a "SHOT" in the official NHL RTSS Play-by-Play report, then it would be a -1 Corsi for you. You get zero Corsis for cycling the puck without attempting a shot.

Here's an example RTSS PbP report: http://www.nhl.com/scores/htmlreports/20132014/PL021226.HTM
These are what sites like Extra Skater and Behind the Net "scrape" for their Corsi data, I believe.

 

[pucka lucka]

If the dump-in is recorded as a "SHOT" in the official NHL RTSS Play-by-Play report, then it would be a -1 Corsi for you. You get zero Corsis for cycling the puck without attempting a shot.

Here's an example RTSS PbP report: http://www.nhl.com/scores/htmlreports/20132014/PL021226.HTM
These are what sites like Extra Skater and Behind the Net "scrape" for their Corsi data, I believe.

I scraped from multiple sources. Too busy to build betterthanextraskater though.

 

[garret9]

Having a talk with some guys at work about Corsi in regards to what Doughty said about it.

One of the questions asked was

If I am a player on a line and we have possession in the offensive zone for a minute but we don't register an attempt on net and the opposing gets the puck, carries it out and dumps it in on net and registers a shot. Do we as a line and player get a negative corsi ?

Yes.

This is why 1 game Corsi tends to have very little meaning. At about 5 games you start to reach about 80% accuracy.

 

[Gm0ney]

I scraped from multiple sources. Too busy to build betterthanextraskater though.

What other sources? Isn't everything you need in the Play-by-Play report?

 

[Gm0ney]

Yes.

This is why 1 game Corsi tends to have very little meaning. At about 5 games you start to reach about 80% accuracy.

"Accuracy" meaning correlation to possession? And is that whole-game or individual player?

 

[pucka lucka]

What other sources? Isn't everything you need in the Play-by-Play report?

Need or want? there is a lot of information missing from the html reports

 

[garret9]

"Accuracy" meaning correlation to possession? And is that whole-game or individual player?

Accuracy as in how well the statistic (Corsi%) represents the parameter (puck possession).


The 80% was very approximate... as I was just saying in general. Team level is a bit higher, player level is a bit lower. It also depends on whether you are wanting time with possession of puck or time in attack zone (related and super similar but not quite the same).

 

[Joe Hallenback]

Would Corsi which uses shot attempts then be an inferior way of figuring out possession as oppposed to say zone time?