Around the NHL: 2019-20 Pt. 3

[Jim Carr's Rug]

Maybe I'm crazy but they kind remind of late 90s Canucks jerseys. Given the proximity to Vancouver, I would have guessed they would have kept their design further away from this:

 

[struckbyaparkedcar]

Maybe I'm crazy but they kind remind of late 90s Canucks jerseys. Given the proximity to Vancouver, I would have guessed they would have kept their design further away from this:

Eh, that organization has two entire color schemes they like more, it’s not like they’re gonna retro those any time soon.

 

[Der Jaeger]

They should put their home numbers and name plates in white. Would stand out more

 

[littletonhockeycoach]

Maybe I'm crazy but they kind remind of late 90s Canucks jerseys. Given the proximity to Vancouver, I would have guessed they would have kept their design further away from this:

That's a good association. My initial reaction was similarity to NYI Fishermans

Similar to..... perhaps the worst NHL jersey ever....

 

[Jim Bob]

OK, now that ruins it for me.

Way to go Bauer.

 

[Myllz]

Yikes, that is horrible.

 

[vcv]

OK, now that ruins it for me.

Way to go Bauer.

Aside from just the overall bad design... why the f*** is there so much red?

 

[Ralonzo]

I was wondering why the font was giving me a gag reflex and then I remembered

 

[Jim Bob]

 

[Jim Bob]

 

[joshjull]

 

[struckbyaparkedcar]

Even if they get better tracking hockey will still have the same issues. The nature of hockey makes it harder to get large sample sizes compared to other sports.

The amount of players that play each game makes it hard to get large sample sizes of any groups of players together relative to similar sports. You also have several game states beside the most common 5v5 (4v4, 5v4, 5v3, 3v5, 4v5, 5v6, 6v5, 3v3 OT) that have an outsized impact on the outcome of a game. But a very small sample of minutes.
Sure.

The NBA in contrast is always 5v5, only has 12 players dress and has starters who play well over half the game together. Giving a pretty sizable sample of their work to look at. By comparison our top line of VO/Jack/Sam only played 18% of our 5v5 minutes together. Thats just 3 players and just at 5v5. If you expand it to 5 players the % drops further. I’d be surprised if any 5 player combo played more than 8-10% of our 5v5 minutes together. A far cry from a starting 5 on an NBA team.

Our 5v5 minutes made up 81% of our total minutes played. Which means the other 19% was split between the other game states that have an outsized impact.

NBA lineups are a bit more fluid than you're giving them credit for. The Warriors "Death Lineup" (Curry, Klay Thompson, KD, Andre Iguodala, & Draymond Green) only played 120 minutes the year they got KD. Their most common 5v5 lineup with Curry (swapping Iggy for Zaza) played ~420 minutes. Even the teams that do full hockey subs with the starters + bench only have that top group together for 18-20 minutes.

Where basketball stats are more predictive, is with individual players and team success, because those elite players will be out there for over half the game (aka why backup goalies drive WAR). Team-level analysis will also be more reflective of top player style-of-play and contributions because of that minute share.

Anyway, the point I was trying to make is how player tracking is able to contextualize expected and actual results in the other two sports. Going back to the Rockets example, their offense transitioned from being a pick-and-roll based scheme to just infinite James Harden isos. Player and team level stats hinted at this change (Hardens shots went up and his assists went down, their pace got slower, etc), but in order to quantify it, you'd have to watch every James Harden possession over the last three years, manually track the actions and outcomes, then have 1-2 other people do the same thing to verify your eyes worked.

With sportVU, that data is just immediately on-hand, and your video review can start with how the Rockets were able to facilitate this change.

Finally getting to hockey, take a player like Dahlin for example. We all saw him nosedive at the beginning of the season in a more conservative system, but we weren't really able to quantify how his responsibilities changed from Housley to Krueger. We also don't know exactly what was behind his end-of-year resurgence, whether it was picking his spots better within the constraints of the system, or the coach relenting and letting a young player get back to his instincts. Player tracking would immediately provide that context.

 

[joshjull]

NBA lineups are a bit more fluid than you're giving them credit for. The Warriors "Death Lineup" (Curry, Klay Thompson, KD, Andre Iguodala, & Draymond Green) only played 120 minutes the year they got KD. Their most common 5v5 lineup with Curry (swapping Iggy for Zaza) played ~420 minutes. Even the teams that do full hockey subs with the starters + bench only have that top group together for 18-20 minutes.

Where basketball stats are more predictive, is with individual players and team success, because those elite players will be out there for over half the game (aka why backup goalies drive WAR). Team-level analysis will also be more reflective of top player style-of-play and contributions because of that minute share.

Anyway, the point I was trying to make is how player tracking is able to contextualize expected and actual results in the other two sports. Going back to the Rockets example, their offense transitioned from being a pick-and-roll based scheme to just infinite James Harden isos. Player and team level stats hinted at this change (Hardens shots went up and his assists went down, their pace got slower, etc), but in order to quantify it, you'd have to watch every James Harden possession over the last three years, manually track the actions and outcomes, then have 1-2 other people do the same thing to verify your eyes worked.

With sportVU, that data is just immediately on-hand, and your video review can start with how the Rockets were able to facilitate this change.

Finally getting to hockey, take a player like Dahlin for example. We all saw him nosedive at the beginning of the season in a more conservative system, but we weren't really able to quantify how his responsibilities changed from Housley to Krueger. We also don't know exactly what was behind his end-of-year resurgence, whether it was picking his spots better within the constraints of the system, or the coach relenting and letting a young player get back to his instincts. Player tracking would immediately provide that context.

A few issues

1) I’m talking about how much value TEAMS can get out of analytics in light of the overall discussion @jc17 kicked off. Its debatable how much NHL teams can get from them vs other sports due to the various reasons listed by myself and others.

But your post is about how YOU or any outside observer of a team can get more understanding about what a team did and how. Which doesn’t jive with what a teams focus would be.

For example, you talk about how YOU need to look at years of stats to know the Rockets transitioned from pick/roll to infinite Harden isos and how it happened. But the Rockets themselves already know this. Since they are the ones who chose to make that transition and any decisions there after to implement it. They don’t need to look at any stats to figure that out.

Which gets to your Dahlin comments. Why would Krueger need to look at any stats to figure out whether or not he decided to change his approach with Dahlin? He would clearly already know that and thus already know the answer to the question your asking about Dahlin’s late season play.

You’re basically arguing for a value to us, the outside observer, that doesn’t exists for the team.

2) Teams already have a lot of ways to use existing stats to contextualize a player’s season. What are their must frequent line combos/d-pairs, how were the deployed and how did they perform in that deployment? How did that performance stack up to Ismail lines around the league. How did it stack up to how the team itself performed, etc. Thats a sampling of where their analysis could go.

Thats how teams should do it and some are clearly better at it than others. But thats not usually how posters on this forum (myself included) or analytics community at large do it. Its usually a much more narrow focus. There aren’t many attempts at deep dives on players to get a fuller picture.

3) There already is player tracking. Teams generally have more data than the public with video coaches breaking every game down. Making that more efficient is a good thing for teams and would improve the information. Making it public knowledge is even better for us. But its still not going to fundamentally change things. Its still getting plugged into the same formulations, etc.

More efficient player tracking in and of itself is not going magically make an individual player’s xGF% much more useful than it is now.

 

[struckbyaparkedcar]

A few issues

1) I’m talking about how much value TEAMS can get out of analytics in light of the overall discussion @jc17 kicked off. Its debatable how much NHL teams can get from them vs other sports due to the various reasons listed by myself and others.

But your post is about how YOU or any outside observer of a team can get more understanding about what a team did and how. Which doesn’t jive with what a teams focus would be.

For example, you talk about how YOU need to look at years of stats to know the Rockets transitioned from pick/roll to infinite Harden isos and how it happened. But the Rockets themselves already know they chose to make that transition and any decisions there after to implement it. They don’t need to look at any stats to figure that out.

Which gets to your Dahlin comments. Why would Krueger need to look at any stats to figure out whether or not he decided to change his approach with Dahlin? He would clearly already know that and thus already know the answer to the question your asking about Dahlin’s late season play.

You’re basically arguing for a value to us, the outside observer, that doesn’t exists for the team.

2) Teams already have a lot of ways to use existing stats to contextualize a player’s season. What are their must frequent line combos/d-pairs, how were the deployed and how did they perform in that deployment? How did that performance stack up to Ismail lines around the league. How did it stack up to how the team itself performed, etc.

Thats how teams should do it and some are clearly better at it than others. But thats not usually how posters on this forum (myself included) or analytics community at large do it. Its usually a much more narrow focus. There aren’t many attempts at deep dives on players to get a fuller picture.

3) There already is player tracking. Teams generally have more data than the public with video coaches breaking every game down. Making that more efficient is a good thing for teams and would improve the information. Making it public knowledge is even better for us. But its still not going to fundamentally change things. Its still getting plugged into the same formulations, etc.

More efficient player tracking in and of itself is not going magically make an individual player’s xGF% much more useful than it is now. I think too many feel a holy grail stat that tells all can be developed.

While my examples might have missed the context of the original discussion by being more publicly-focused, both our posts detail a very important professional value of aggregated player tracking. NHL teams are already spending hundreds of hours processing tape in order to produce results that are generated instantly in other sports.

Also, these anecdotes can quickly be retrofitted to the team level. While D'Antoni would have already had:

• The knowledge he was running more isolation
• The overall data that his offensive efficiency hadn't dropped

He would not have been able to determine the exact % change in his offense or the efficiency by play type without spending a lot of time on the film. He also wouldn't be able to calculate things like expected points per shot unless his video team had their own scoring system for things like "how far away defenders were."

The same goes for Dahlin and the parallel universe Krueger who would actually use tracking data (although, he had access to that in the Premier league, so who knows?). While he'd be generally aware of how Rasmus responded to role changes, it'd take time with the tape to determine exactly what circumstances are driving that success. In tracking land, the video review would start with an understanding of play type, detailed shot quality, etc.

This gets us into the intersection of player tracking and xG. While new tech would improve the application of current xG models by providing context to role, it would also allow for the creation of better models by instantly factoring in things like passing data.

 

[joshjull]

Ok then





"John Chayka has quit as the General Manager and President of Hockey Operations of the Arizona Coyotes.

"The Club is disappointed in his actions and his timing as the Coyotes prepare to enter the NHL's hub city of Edmonton, where the team will begin post-season play for the first time since 2012. Chayka has chosen to quit on a strong and competitive team, a dedicated staff, and the Arizona Coyotes fans, the greatest fans in the NHL.



Yikes

 

[joshjull]

 

[Fjordy]

Pegulas earlier wanted to see Chayka in Buffalo

 

[joshjull]

 

[Fjordy]

It seems to me that Chayka was a solid GM, it seems his main problem is that he held on to the head coach, under which many attacking players degraded.

 

[flashsabre]

No matter what goes on, trashing an employee in the media is a real bad look. You can be pissed off at the person all you want but you need to put on a professional face with the media.

Interesting to hear from Chayka in the coming weeks.

 

[Ace]

I just know somehow these dummies are going to cost this team a first round pick for doing something dumb

there’s stuff about it being an opportunity above just another nhl job. And this was as they decided to fire Botterill and also reorganize PSE. And they were reportedly after him 3 years ago when they hired Botterill.

 

[Fjordy]

I just know somehow these dummies are going to cost this team a first round pick for doing something dumb

How?

 

[CrazyPsycho]

Odd post. Why would we be after Chayka?

 

[Gabrielor]

Glad we aren’t in on Chayka. I don’t exactly look enviously at the Coyotes.

I’m hoping for better from Pegula 4.0.

 

[VaporTrail]

How long before we have a team as good as the Kraken ?