Player Discussion: Winnipeg Jets Defense

[RustyCat]

So what happens when Fleury returns? Does he take Stanley’s spot?

Yeah makes you wonder as Fleury played pretty well. I wouldnt be surprised to see Chevy move Stanley to VAN for a guy like Deharnais. Play him on the right side and invest in Heinola on the left side, make Fleury the 7D

 

[Buffdog]

So what happens when Fleury returns? Does he take Stanley’s spot?

That's what we're all hoping for. Ultimately think Chevy upgrades that sopt (3LD) at some point

 

[10Ducky10]

I thought Fleury was healthy already?

 

[Whileee]

This post is a good example of why hes not here... the guy clearly explains his models and others and goes in depth as to what the potential flaws or biases are... and then guys who clearly dont read that content make posts about how garrett thinks hes the be all end all.

Also i can tell you if you think peer review is unbiased you are mistaken... as a polar bear guide ive seen various agendas override science in that world.

Nothing is perfect but the guys making these models for hockey do not have a specific agenda except trying to make their models work.... they dont hate stan or pionk or scheif etc...

I like statistical approaches to hockey analysis but I can say unequivocally that the level of "peer review" and expert discourse about hockey-related models doesn't come close to most scientific fields. There isn't nearly as much transparent scrutiny of data quality and methods as you would see in most research fields. Moreover, many of the hockey stats folks make a living from their own proprietary models. I don't object to that at all, but it does tend to create a level of defensiveness and brittleness that can stifle healthy discourse. Add in that some rely on revenue from the sports gambling enterprise, and things get even murkier.

That said, I find most of the public modelers and data scientists in hockey to be very sharp and engaging. They can be overly dismissive, but that's a feature of social media, generally. I would note that there is also a large contingent of hard core "anti-stats" folks that add to the rancour by dismissing useful statistical insights out of hand.

Balance is hard to find, but overall I think that the introduction of statistical insights has been a great addition for the engaged fan.

 

[Jet]

I like statistical approaches to hockey analysis but I can say unequivocally that the level of "peer review" and expert discourse about hockey-related models doesn't come close to most scientific fields. There isn't nearly as much transparent scrutiny of data quality and methods as you would see in most research fields. Moreover, many of the hockey stats folks make a living from their own proprietary models. I don't object to that at all, but it does tend to create a level of defensiveness and brittleness that can stifle healthy discourse. Add in that some rely on revenue from the sports gambling enterprise, and things get even murkier.

That said, I find most of the public modelers and data scientists in hockey to be very sharp and engaging. They can be overly dismissive, but that's a feature of social media, generally. I would note that there is also a large contingent of hard core "anti-stats" folks that add to the rancour by dismissing useful statistical insights out of hand.

Balance is hard to find, but overall I think that the introduction of statistical insights has been a great addition for the engaged fan.

This right here.

 

[TS Quint]

I like statistical approaches to hockey analysis but I can say unequivocally that the level of "peer review" and expert discourse about hockey-related models doesn't come close to most scientific fields. There isn't nearly as much transparent scrutiny of data quality and methods as you would see in most research fields. Moreover, many of the hockey stats folks make a living from their own proprietary models. I don't object to that at all, but it does tend to create a level of defensiveness and brittleness that can stifle healthy discourse. Add in that some rely on revenue from the sports gambling enterprise, and things get even murkier.

That said, I find most of the public modelers and data scientists in hockey to be very sharp and engaging. They can be overly dismissive, but that's a feature of social media, generally. I would note that there is also a large contingent of hard core "anti-stats" folks that add to the rancour by dismissing useful statistical insights out of hand.

Balance is hard to find, but overall I think that the introduction of statistical insights has been a great addition for the engaged fan.

I think that the advanced stats are good….but just a little too broad to discuss in specific cases to be entirely accurate. A lot of time they don’t are into account specific circumstances and focus on average. I don’t believe in averages when you are talking about specific players especially exceptional players to begin with. Exceptional players can be found at either end of the spectrum and of course those are the most talked about topics. Also many of these stats are looked at like each player has the same amount of responsibility on the ice. Never mind the fact you have one constant and 11 variables on the ice. I think it’s obvious to most people that Centres have more potential for impact on the ice than wingers but in many cases people put every player as equals.

Kyle Connor is a great example of this. Every year people who claim to be stats people change their mind. Oh he’s terrible defensively with Scheifele, oh he learned how to play defense with Dubois, oh he’s terrible with Vilardi and Schiefele, oh now he learned how to defend again. Absolutely none of these opinions have anything to do with what he is doing on the ice.

 

[Gm0ney]

I think that the advanced stats are good….but just a little too broad to discuss in specific cases to be entirely accurate. A lot of time they don’t are into account specific circumstances and focus on average. I don’t believe in averages when you are talking about specific players especially exceptional players to begin with. Exceptional players can be found at either end of the spectrum and of course those are the most talked about topics. Also many of these stats are looked at like each player has the same amount of responsibility on the ice. Never mind the fact you have one constant and 11 variables on the ice. I think it’s obvious to most people that Centres have more potential for impact on the ice than wingers but in many cases people put every player as equals.

Kyle Connor is a great example of this. Every year people who claim to be stats people change their mind. Oh he’s terrible defensively with Scheifele, oh he learned how to play defense with Dubois, oh he’s terrible with Vilardi and Schiefele, oh now he learned how to defend again. Absolutely none of these opinions have anything to do with what he is doing on the ice.

You're saying that the general opinion of "stats people" has vacillated on Connor's defense over the years? That's not how I remember it. But human memory is notoriously fallible. That's one of the reasons why stats come in handy.

I think the general stats people consensus was that the combo of Connor + Scheifele was poor defensively - and that's pretty much been the case in every season since 2017-18, until this year.

Pointing out that the on-ice results were that the Jets gave up fewer chances and significantly outscored their opponents when he was with Dubois isn't flip flopping.

Also, I think Connor's blame for defensive ineptitude is based a lot on the eye test. He was often controller-disconnected in the D-zone in previous years, but he's way more engaged and looks like he's trying this year.

 

[DRW204]

You're saying that the general opinion of "stats people" has vacillated on Connor's defense over the years? That's not how I remember it. But human memory is notoriously fallible. That's one of the reasons why stats come in handy.

I think the general stats people consensus was that the combo of Connor + Scheifele was poor defensively - and that's pretty much been the case in every season since 2017-18, until this year.

Pointing out that the on-ice results were that the Jets gave up fewer chances and significantly outscored their opponents when he was with Dubois isn't flip flopping.

Also, I think Connor's blame for defensive ineptitude is based a lot on the eye test. He was often controller-disconnected in the D-zone in previous years, but he's way more engaged and looks like he's trying this year.

not sure why folks on here have such a hard time accepting more than 1 thing being true.
from '18-19 to '23-24 : connor/top line was poor defensively and usually around even or worse in terms of net-goals, and this year they've been fantastic.

 

[BobTheSolarsystem]

I like statistical approaches to hockey analysis but I can say unequivocally that the level of "peer review" and expert discourse about hockey-related models doesn't come close to most scientific fields. There isn't nearly as much transparent scrutiny of data quality and methods as you would see in most research fields. Moreover, many of the hockey stats folks make a living from their own proprietary models. I don't object to that at all, but it does tend to create a level of defensiveness and brittleness that can stifle healthy discourse. Add in that some rely on revenue from the sports gambling enterprise, and things get even murkier.

That said, I find most of the public modelers and data scientists in hockey to be very sharp and engaging. They can be overly dismissive, but that's a feature of social media, generally. I would note that there is also a large contingent of hard core "anti-stats" folks that add to the rancour by dismissing useful statistical insights out of hand.

Balance is hard to find, but overall I think that the introduction of statistical insights has been a great addition for the engaged fan.

Interesting anecdote about peer review... I've been a polar bear guide for 25 years now and in the George W years there was a real race between science and politics, especially for ANWR rights.

Well, our top three polar bear scientists came together and built a model based on climate projections... that served the purpose to stop oil and gas expansion but was soon proved wrong - because we had based it on projections from an extreme el nino/la nina phase before we really understoof the natural cycles.

Regardless, the top three had staked their reps on it and for 10-15 years - any papers that disagreed with this model, that was rapidly being proven wrong in real life - were essentially denied peer review... until nature provided so much data that the model was quietly 'forgotten'...

There are egos and agendas in all facets of life and in data - but honestly I find the hockey adv stats to be pretty open with their own perceived flaws - given they are all trying to get more contracts with NHL teams... you don't get paid to push your own opinion - they already have scouts for that.

 

[Buffdog]

not sure why folks on here have such a hard time accepting more than 1 thing being true.
from '18-19 to '23-24 : connor/top line was poor defensively and usually around even or worse, and this year they've been fantastic.

Last year, CSV gave up 1.21 shot attempts per minute. This year it's 1.05 shot attempts per minute. That's about 2 extra shot attempts per game more last year. Not shots. Shot attempts (all at 5v5)

They're basically exactly the same defensively as they were last year

 

[Whileee]

I think that the advanced stats are good….but just a little too broad to discuss in specific cases to be entirely accurate. A lot of time they don’t are into account specific circumstances and focus on average. I don’t believe in averages when you are talking about specific players especially exceptional players to begin with. Exceptional players can be found at either end of the spectrum and of course those are the most talked about topics. Also many of these stats are looked at like each player has the same amount of responsibility on the ice. Never mind the fact you have one constant and 11 variables on the ice. I think it’s obvious to most people that Centres have more potential for impact on the ice than wingers but in many cases people put every player as equals.

Kyle Connor is a great example of this. Every year people who claim to be stats people change their mind. Oh he’s terrible defensively with Scheifele, oh he learned how to play defense with Dubois, oh he’s terrible with Vilardi and Schiefele, oh now he learned how to defend again. Absolutely none of these opinions have anything to do with what he is doing on the ice.

You make some really good points about the limitations of many statistical summaries.

However, my own view is that Connor has struggled in the past defensively when playing with Scheifele. My view isn't just based on stats, but also watching and noting how often they failed in zone exits and turned the puck over in the neutral zone and didn't maintain a "high" F in the offensive zone when warranted.

My view is that you have to interpret the stats with caution, but when they are really consistent and persistent, it's useful to try to understand what they are indicating by watching the play more closely, and not just the skilled plays.

 

[Whileee]

Last year, CSV gave up 1.21 shot attempts per minute. This year it's 1.05 shot attempts per minute. That's about 2 extra shot attempts per game more last year. Not shots. Shot attempts (all at 5v5)

They're basically exactly the same defensively as they were last year

But the types of shots they are giving up are very different, and their overall expected goals against were 25% higher last season. I think it's two things. One, they are much more efficient in zone exits and we aren't seeing nearly as many of those awful shifts where they are chasing opponents around in their own zone until the D are exhausted and they give up prime scoring chances. Two, they are much better at avoiding odd-man breaks against. This is by being more sure with the puck in the neutral zone and being more structured in the offensive zone so that they are getting very good back pressure from Connor and Scheifele.

Look at these two shot against charts for Connor. Even if shot volumes aren't so different, the location of shots against is much different this season.

 

[DRW204]

But the types of shots they are giving up are very different, and their overall expected goals against were 25% higher last season. I think it's two things. One, they are much more efficient in zone exits and we aren't seeing nearly as many of those awful shifts where they are chasing opponents around in their own zone until the D are exhausted and they give up prime scoring chances. Two, they are much better at avoiding odd-man breaks against. This is by being more sure with the puck in the neutral zone and being more structured in the offensive zone so that they are getting very good back pressure from Connor and Scheifele.

Look at these two shot against charts for Connor. Even if shot volumes aren't so different, the location of shots against is much different this season.
View attachment 964303View attachment 964305

Looking at the trio's xGA or HDCA this year vs last is really telling. Maybe they're relatively closer in shot attempts but as the heat map suggests, they appear vastly different in terms of quality. I'll take 10 long range shot attempts vs 10 high danger ones. Counts as the same # of shot attempts (or Corsi-against), which is why IMO pairing a quality and volume measure is beneficial.

Their improvement in effort defensively (+ Helle of course) has translated to Connor nearing the 17-18 season's level in net-goals.

I made a post last week giving KC specifically his props bc I think his level of engagement this year has been great compared to recent years.

 

[RustyCat]

I like statistical approaches to hockey analysis but I can say unequivocally that the level of "peer review" and expert discourse about hockey-related models doesn't come close to most scientific fields. There isn't nearly as much transparent scrutiny of data quality and methods as you would see in most research fields. Moreover, many of the hockey stats folks make a living from their own proprietary models. I don't object to that at all, but it does tend to create a level of defensiveness and brittleness that can stifle healthy discourse. Add in that some rely on revenue from the sports gambling enterprise, and things get even murkier.

That said, I find most of the public modelers and data scientists in hockey to be very sharp and engaging. They can be overly dismissive, but that's a feature of social media, generally. I would note that there is also a large contingent of hard core "anti-stats" folks that add to the rancour by dismissing useful statistical insights out of hand.

Balance is hard to find, but overall I think that the introduction of statistical insights has been a great addition for the engaged fan.

Liked how you framed this. Thanks for that insight

 

[BobTheSolarsystem]

Last year, CSV gave up 1.21 shot attempts per minute. This year it's 1.05 shot attempts per minute. That's about 2 extra shot attempts per game more last year. Not shots. Shot attempts (all at 5v5)

They're basically exactly the same defensively as they were last year

You realize this amounts to 12-15 goals less... per season

Ah i see shot attempts - still thats prob 10 goals over the season which is a legit difference

 

[Buffdog]

But the types of shots they are giving up are very different, and their overall expected goals against were 25% higher last season. I think it's two things. One, they are much more efficient in zone exits and we aren't seeing nearly as many of those awful shifts where they are chasing opponents around in their own zone until the D are exhausted and they give up prime scoring chances. Two, they are much better at avoiding odd-man breaks against. This is by being more sure with the puck in the neutral zone and being more structured in the offensive zone so that they are getting very good back pressure from Connor and Scheifele.

Look at these two shot against charts for Connor. Even if shot volumes aren't so different, the location of shots against is much different this season.
View attachment 964303View attachment 964305

How does that difference compare to the team as a whole this year vs last year? Has the whole team given up less quality chances? If so, is it the players that have improved, or the system that they're playing in?

I think you're right with both the zone exits and odd man rushes, and i think it's in large part due to the tweaks that arniel has made in the systems/structure in the offensive zone and neutral zone. Bones pushed a relentless puck pursuit scheme to try to force turnovers whereas Arniel has them backing out and forming a 1-2-2 in the neutral zone that forces teams to dump, puts enough bodies between the retrieving D and the opposition that they have an extra second to make a better first pass and puts the forwards in closer support for a breakout

It all goes back to the use of stats to judge players though. Any conclusions drawn by someone just looking at numbers is bound to miss the nuances that we're talking about here. For example.. "Connor and Schief are bad together defensively"... well, the numbers say that... but can you figure out WHY?

In baseball, a .275 hitter is a .275 hitter. Move him from one line up to another, or bat him somewhere else in the batting order. That's largely what he will be

People try to do that with hockey players too, but this example shows that it's a fools errand. The same player (or combination of players) can have better results in different schemes and systems

 

[Mortimer Snerd]

In statistics, data gathering and cleanliness is the single most important thing, so it's not an out of line question to ask.

And? Criminals often know what they are doing is a crime and it doesn't prevent them from doing it.

I like Garret - I miss him on these boards, but that doesn't mean his methodology is scientifically sound or peer reviewed (peer being an actual data scientist).

Who crowned him as the 'professional statistician'? Is it simply the product of him working and earning money in that role? If I called myself a doctor and started treating patients because I watched youtube videos, would that make me a doctor?

Hockey advanced stats are still in their infancy. There are a lot of people pushing a lot of 'their' metrics, some gimmicky and some not. There are tons of questions with these metrics:

• How are they compiled?
• How are they interpreted?
• How are they weighted or used in calculations?
• What is the background of the people doing this work?
• Are they subjective or objective?
• How likely is there to be bias?

I keep bringing this stuff up but those who use statistics as the bible or foundation of their understanding of hockey and players ignore the questions, it's almost like a cult.

I think what is happening in a lot of cases is that the NHL and teams use their own metrics and advanced stats to make determinations, and couple that with what they are seeing and hard outcomes. Those metrics are usually not what we see, and they aren't shared with the public.

That's why you see a lot of team and coaching moves that don't align with the advanced stats Garret, JFresh, whomever else are promoting.

I'm not saying that these stats are dereft of accuracy or value, but there are a TON of question marks, and when the stats don't match an experienced eye, it can call it in to question. Of course, eye test is very prone to bias as well.

In my profession, I see statistics compiled, bent, and manipulated to tell a story all the time. That's why you need a scientific method and peer review if you want accuracy.


I don't think that's true - at least not from what I've read (please someone correct me if I'm wrong). I don't think disclosure is necessary, though it would be written in the contract.

I think that if you follow Garrett at all you will find that he addresses all of your questions. For example, he explains the strengths and weaknesses of the methods used by other statisticians whose work he quotes. You may or may not like the way he addresses them. That is up to you.

 

[Whileee]

How does that difference compare to the team as a whole this year vs last year? Has the whole team given up less quality chances? If so, is it the players that have improved, or the system that they're playing in?

I think you're right with both the zone exits and odd man rushes, and i think it's in large part due to the tweaks that arniel has made in the systems/structure in the offensive zone and neutral zone. Bones pushed a relentless puck pursuit scheme to try to force turnovers whereas Arniel has them backing out and forming a 1-2-2 in the neutral zone that forces teams to dump, puts enough bodies between the retrieving D and the opposition that they have an extra second to make a better first pass and puts the forwards in closer support for a breakout

It all goes back to the use of stats to judge players though. Any conclusions drawn by someone just looking at numbers is bound to miss the nuances that we're talking about here. For example.. "Connor and Schief are bad together defensively"... well, the numbers say that... but can you figure out WHY?

In baseball, a .275 hitter is a .275 hitter. Move him from one line up to another, or bat him somewhere else in the batting order. That's largely what he will be

People try to do that with hockey players too, but this example shows that it's a fools errand. The same player (or combination of players) can have better results in different schemes and system

The data indicate that Connor's defensive shot metrics have improved much more than the team as a whole.

I think Arniel has been able to persuade that line to play his system. They certainly seem to be making more effort to make smart and lower risk plays when they don't have options. Connor has also been way more active and diligent in back pressure, in my view. Maybe also Vilardi being healthy and that line finding their groove.