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

[Aavco Cup]

Consider the following from the Jets offseason:

The Jets tried to sign Matt Stone/Karl Alzner
The Jets tried to trade for Alexie Emelin
The Jets signed Kulikov
The Jets signed and have been hyping up Hendricks

Whoever the Jets pay for analytics either doesn't exist or is no good or their data is falling on deaf ears.

You have no idea what or how the Jets use data and analytics

 

[Aavco Cup]

I recall the "zone time" talk last year. At the beginning of this season either Chevy or Maurice was saying that they did a lot of work on analytics in the off season trying to determine which metrics were really correlated to winning and which were not. I suspect they don't believe looking for better corsi is the right answer. Aside from the Penguins with their negative corsi winning the cup, the results so far this year would not encourage a team to chase better corsi. Of the top 10 teams (ROW) only 3 have 5 on 5 corsi above 50%. Of the bottom 10 teams 6 do.

I think the field has moved beyond Corsi and the scraped data. Zone entry zone entry denial are one area I wish we had more data for in the public realm

 

[Whileee]

You have no idea what or how the Jets use data and analytics

He also has no idea what the Jets actually tried to do in the off-season with respect to trades or free agent signings. Some HF Jets posters love to evaluate the Jets based on unsubstantiated rumours or fantasy trades that they make up and think the Jets should have made.

 

[Whileee]

I recall the "zone time" talk last year. At the beginning of this season either Chevy or Maurice was saying that they did a lot of work on analytics in the off season trying to determine which metrics were really correlated to winning and which were not. I suspect they don't believe looking for better corsi is the right answer. Aside from the Penguins with their negative corsi winning the cup, the results so far this year would not encourage a team to chase better corsi. Of the top 10 teams (ROW) only 3 have 5 on 5 corsi above 50%. Of the bottom 10 teams 6 do.

As teams adapt and change styles, we shouldn't expect the same metrics to have the same correlations with goals or winning. A lot of teams are starting to try to limit chances against by avoiding low-percentage shots, and many teams are collapsing to the middle of the ice in their own zones to prevent high-danger scoring chances and allow more peripheral shots. It's a shame that so many of the analysts just do proprietary work and there isn't a more robust culture of publishing models and results in the public sphere.

 

[KingBogo]

He also has no idea what the Jets actually tried to do in the off-season with respect to trades or free agent signings. Some HF Jets posters love to evaluate the Jets based on unsubstantiated rumours or fantasy trades that they make up and think the Jets should have made.

That's kind of how it rolls. Hear/read a rumour, think of something you believe Chevy should have done then post it as something real.

 

[KingBogo]

As teams adapt and change styles, we shouldn't expect the same metrics to have the same correlations with goals or winning. A lot of teams are starting to try to limit chances against by avoiding low-percentage shots, and many teams are collapsing to the middle of the ice in their own zones to prevent high-danger scoring chances and allow more peripheral shots. It's a shame that so many of the analysts just do proprietary work and there isn't a more robust culture of publishing models and results in the public sphere.

This is an interesting concept. I'd be really curious to know what teams are doing internally compared to what's available for public consumption. But you can't blame people for going where the money is.

 

[pucka lucka]

Such compelling arguments. Every thread is now infected with this BS.

 

[Aavco Cup]

This is an interesting concept. I'd be really curious to know what teams are doing internally compared to what's available for public consumption. But you can't blame people for going where the money is.

I've said this before. The league morphs over time. And some "metrics" that were shown to have a lower correlation to future results (like the limited stopwatch zone possession tests that were done a few years back) may actually be different today or in the future. That's why it would be nice if this data was readily available in the public sphere. So it could be retested.

 

[Gm0ney]

I recall the "zone time" talk last year. At the beginning of this season either Chevy or Maurice was saying that they did a lot of work on analytics in the off season trying to determine which metrics were really correlated to winning and which were not. I suspect they don't believe looking for better corsi is the right answer. Aside from the Penguins with their negative corsi winning the cup, the results so far this year would not encourage a team to chase better corsi. Of the top 10 teams (ROW) only 3 have 5 on 5 corsi above 50%. Of the bottom 10 teams 6 do.

No one's even played 10 games yet. RIP Corsi.

 

[Gm0ney]

He also has no idea what the Jets actually tried to do in the off-season with respect to trades or free agent signings. Some HF Jets posters love to evaluate the Jets based on unsubstantiated rumours or fantasy trades that they make up and think the Jets should have made.

And some posters like to pretend the Jets' history of lousy defensemen acquisitions never happened. At least they threw some darts at the board, though - unlike their policy with regards to goaltenders and coaching.

 

[Aavco Cup]

And some posters like to pretend the Jets' history of lousy defensemen acquisitions never happened. At least they threw some darts at the board, though - unlike their policy with regards to goaltenders and coaching.

The reports were that at some point in the offseason the Jets decided to explore and use more analytical information.

That has to be a good thing in the long run.

 

[Gm0ney]

The reports were that at some point in the offseason the Jets decided to explore and use more analytical information.

That has to be a good thing in the long run.

Well that Yost article sort of lays out the idea that some of the black-box private stuff is snake oil ("If it’s not disingenuous, it’s bad faith. And if it’s neither, then it’s unacceptable negligence.").

The offensive zone time stuff Maurice has mentioned a few times is concerning - the point of Corsi isn't that shot-attempts correlate to possession; it's that shot-attempts correlate to winning.

Maurice also said he believes shot quality > quantity - and the Yost article explains that there's no evidence that says that's true.

 

[Gil Fisher]

Maurice also said he believes shot quality > quantity - and the Yost article explains that there's no evidence that says that's true.

I have a suspicion that we're not very good at measuring shot quality.

 

[Aavco Cup]

Well that Yost article sort of lays out the idea that some of the black-box private stuff is snake oil ("If it’s not disingenuous, it’s bad faith. And if it’s neither, then it’s unacceptable negligence.").

The offensive zone time stuff Maurice has mentioned a few times is concerning - the point of Corsi isn't that shot-attempts correlate to possession; it's that shot-attempts correlate to winning.

Maurice also said he believes shot quality > quantity - and the Yost article explains that there's no evidence that says that's true.

I know what Yost is saying. That's why I posted the article. What it doesn't say is that every "black box" is snake oil, only that one should be wary.

You have no idea whether Maurice values quality vs quantity and by how much. You're just imposing your own biases by saying that IMO.

I'm not convinced zone time is irrelevant. I'm not convinced the few studies done in this area have enough sample size and enough depth in them to be so dismissive. Getting data in this area is difficult for the public realm as it is labour intensive and requires a certain level of subjectivity on the observer doing the manual tracking. Did those studies account for in game variables that Yost mentions in his rant?

Zone time is one piece of data I think every NHL team should be willing to pay for. Now how to analyze it and ultimately use that data to make your team better. Isn't that what they should be trying to do?

 

[WPGChief]

Zone time is one piece of data I think every NHL team should be willing to pay for. Now how to analyze it and ultimately use that data to make your team better. Isn't that what they should be trying to do?

Stathletes was one of the first to extensively study zone time possession in the past, and found it correlated less to goals and wins than your other metrics (corsi, expected goals, etc.). Unfortunately this is more what I've heard passed around in the blogging sphere as it isn't public, and old research done by Vic Ferrari/Timothy Barnes has been taken down and made private.

Last I heard, Stathletes were also one of Toronto's major analytics providers.

 

[Board Bard]

That's kind of how it rolls. Hear/read a rumour, think of something you believe Chevy should have done then post it as something real.

That's pretty rich. There are tons of ridiculously rosy posts where people think Chevy is doing this or Maurice is doing that with some grand plan or master strategy, all with the aim of making Jets management seem smarter and better than results indicate they in fact are. The critics might not be perfectly objective but they're head and shoulders more objective than the rosebuds.

 

[Aavco Cup]

Stathletes was one of the first to extensively study zone time possession in the past, and found it correlated less to goals and wins than your other metrics (corsi, expected goals, etc.). Unfortunately this is more what I've heard passed around in the blogging sphere as it isn't public, and old research done by Vic Ferrari/Timothy Barnes has been taken down and made private.

Last I heard, Stathletes were also one of Toronto's major analytics providers.

I've heard it constantly from the blogging community since we started talking about analytics around here. It is almost as if it's an excuse to ignore the whole concept. The game evolves and isn't it normal for science guys to tests and retest and retest again?

 

[Whileee]

Whenever a technical capability or "science" finds a commercial application you end up having reams of folks coming out of the woodwork with their "proprietary" methods and tools, trying to sell it to unsophisticated managers (like NHL GMs). I think that this movement in the NHL to hire in analysts or analytical companies is likely to result in a lot of methods and results being promoted with NHL teams without the kind of peer review and vetting that would make methods better and more robust. That lack of transparency will make it difficult to know whether or how new analytical methods might improve on old ones, and it will constrain the further testing of some of the standard methods. That's unfortunate.

 

[Whileee]

That's pretty rich. There are tons of ridiculously rosy posts where people think Chevy is doing this or Maurice is doing that with some grand plan or master strategy, all with the aim of making Jets management seem smarter and better than results indicate they in fact are. The critics might not be perfectly objective but they're head and shoulders more objective than the rosebuds.

Stanley pick was terrible. So was the Sutter pick.
Setoguchi trade was dumb.
The general strategy of trading picks for rentals was inconsistent with a full "draft and develop" strategy.
Keeping Pavs for so long was bad.
Stuart and Harrison were a travesty.
Wright and Miettinen were mind-bogglingly bad decisions.
Highly skeptical that Maurice will be able to coach a highly skilled team, as the Jets are currently constructed.
Hated the signing of Hendricks.
Demoting Connor was dumb.

That's just a small sample of my views about Chevy and Maurice, expressed in a myriad of critical posts over the years.

See if you can cite 5 positive posts you've made about Chevy or Maurice in the past few seasons.

 

[WPGChief]

I've heard it constantly from the blogging community since we started talking about analytics around here. It is almost as if it's an excuse to ignore the whole concept. The game evolves and isn't it normal for science guys to tests and retest and retest again?

Then I encourage you to go track zone time for ~1200+ games per season for however long you like. For companies like Stathletes and Sportslogiq and HockeyData, they are fortunate enough to be able to conduct proprietary research and analysis, and get paid for it, unlike the public work we are fortunate enough to see around the Twittersphere or blogosphere. If based on their previous work or others' previous work they are confident enough to move on from zone time, like many others in the public space as well, you probably should be as well.

Also, if you believe that the game has drastically evolved since ~5 years ago so that some metrics have drastically changed their correlations and importance, I have some magic beans to sell you.

 

[Aavco Cup]

Then I encourage you to go track zone time for ~1200+ games per season for however long you like. For companies like Stathletes and Sportslogiq and HockeyData, they are fortunate enough to be able to conduct proprietary research and analysis, and get paid for it, unlike the public work we are fortunate enough to see around the Twittersphere or blogosphere. If based on their previous work or others' previous work they are confident enough to move on from zone time, like many others in the public space as well, you probably should be as well.

Also, if you believe that the game has drastically evolved since ~5 years ago so that some metrics have drastically changed their correlations and importance, I have some magic beans to sell you.

You misunderstand me. The NHL needs to do more. The NBA keeps pushing the envelope. Now a puck moves faster and is much smaller than a basketball so I realize there are challenges.

My only fear is that the NHL will eventually do more but withold data from the public

 

[Whileee]

Then I encourage you to go track zone time for ~1200+ games per season for however long you like. For companies like Stathletes and Sportslogiq and HockeyData, they are fortunate enough to be able to conduct proprietary research and analysis, and get paid for it, unlike the public work we are fortunate enough to see around the Twittersphere or blogosphere. If based on their previous work or others' previous work they are confident enough to move on from zone time, like many others in the public space as well, you probably should be as well.

Also, if you believe that the game has drastically evolved since ~5 years ago so that some metrics have drastically changed their correlations and importance, I have some magic beans to sell you.

Well, there's a thriving business in magic beans these days when it comes to the NHL and analytics....

Games and strategies evolve. How drastically is a different question. If the basic correlations and relative importance (vis-à-vis other metrics) aren't changing, it seems that many teams are buying a lot of magic beans that they probably don't need.

 

[Aavco Cup]

Well, there's a thriving business in magic beans these days when it comes to the NHL and analytics....

Games and strategies evolve. How drastically is a different question. If the basic correlations and relative importance (vis-à-vis other metrics) aren't changing, it seems that many teams are buying a lot of magic beans that they probably don't need.

Perhaps but just like at the entry draft perhaps you select a few different options and see what develops. I hope nobody is putting all their eggs in one magic bean basket

 

[Whileee]

Stathletes was one of the first to extensively study zone time possession in the past, and found it correlated less to goals and wins than your other metrics (corsi, expected goals, etc.). Unfortunately this is more what I've heard passed around in the blogging sphere as it isn't public, and old research done by Vic Ferrari/Timothy Barnes has been taken down and made private.

Last I heard, Stathletes were also one of Toronto's major analytics providers.

Stathletes is the company that John Chayka set up. As I understand it, they focus a lot on tracking individual players and their various capabilities (such as ability to "make plays under pressure"). I would imagine that the Yotes are using that system extensively. http://edmontonjournal.com/sports/hockey/nhl/cult-of-hockey/john-chayka-arizona-coyotes
https://www.stathletes.com/

As you've noted, there should be a value proposition to move to proprietary analytical methods and tools (which cost more), and without much of the methods or testing being published for wider peer review, it's hard to know how much snake oil different teams are buying.

I hope the Jets are getting and using good advice in this area, but it's hard to know.

Meanwhile, since the Stathletes methodology is meant to not only identify effective players and systems, but also to provide instruction to individual players to improve their own performance, I wonder how well Tocchet is able to incorporate this into his coaching. So far, the results haven't been very impressive.

 

[Mortimer Snerd]

Then I encourage you to go track zone time for ~1200+ games per season for however long you like. For companies like Stathletes and Sportslogiq and HockeyData, they are fortunate enough to be able to conduct proprietary research and analysis, and get paid for it, unlike the public work we are fortunate enough to see around the Twittersphere or blogosphere. If based on their previous work or others' previous work they are confident enough to move on from zone time, like many others in the public space as well, you probably should be as well.

Also, if you believe that the game has drastically evolved since ~5 years ago so that some metrics have drastically changed their correlations and importance, I have some magic beans to sell you.

I would be surprised if things changed drastically but it seems logical that some of these effects would tilt a little, one way or another as people become more aware of the data and the analysis. If, or by how much, or in what direction are all unknown if the assumptions aren't being regularly re-tested.