HFBoards Top 32 Defensemen - #2

[Bouboumaster]

Hedman

Add Habs #1 dman Matheson plz

(joke)

 

[Phion Keneuf]

Has to be Hedman. Dude is a beast

 

[teravaineSAROS]

And I find the disrespect towards Ekholm and overrating of Josi because points ridiculous. Props to Josi for actually scoring them, but he owes Ekholm and ENORMOUS “thank you” for shouldering the tough defensive load and freeing up Josi to rack up points.


Now look at their shTOI and zone start %s. Ekholm lead his team in short handed minutes and was among the top players in the entire league in PK time, Josi was 6th on the predators in total shTOI and 9th in shTOI per game. Ekholm started in the ozone 36.6% vs 67.2% ozone starts for Josi. For all the accolades Josi gets after a big offensive season, he would not be able to score like he did without Ekholm doing the dirty work for him.

I think the most underrated players in the game are definitely guys like Ekholm.

As far as deployment goes what is Nashville supposed to do? The other defensemen aren't going to score much and a majority of their forwards are known to do everything but produce points.

Nashville is stacked with defense-first players, they owe an enormous thank you to Josi aswell.

Hope you show up in the center polls because I have a feeling some defensively sound forwards will go a lot lower than they should

 

[majormajor]

Hedman has only looked like Hedman about a third of the games in the last couple years. He's been very dinged up much of the time. At some point you might have to ask if he's riding on reputation.

 

[bobholly39]

Fox is going to end up with an average Norris finish the last two years higher than this poll ranks him. Just an observation.

Who exactly would you have him above, and what's the rationale?

Makar - even in the year Fox did win his Norris, there's a good argument to make Makar was better, just with 11 games less. Than Smythe (one of the strongest in the past 20 years), and actual Norris win last year, and also beating Fox year 1 for Calder....easy argument of Makar > Fox in the rankings as of today.

Hedman. Fox may have Norris finishes of 1 and 5 in past 3 years, but Hedman has 6 straight Norris finalists nods. Combine that with his incredibly strong Smythe in 2020 followed by 2 more great playoff runs...again, easy to rank Hedman above too.

Josi - 1 Norris and 1 2nd place in last 3 years, which is better 3 year average than Fox. Mostly - last year's regular season is arguably the strongest regular season by a defenseman in many, many years (despite finishing 2nd to Makar for Norris). So again, easy argument to rank him high too.

For what it's worth - I think Adam Fox 100% belongs in the top 4, and that going into next season this top 4 is a tier above rest of league. And maybe you can argue him above Josi if you really want to - but I personally would slot him 4th.

 

[ReginKarlssonLehner]

Hedman

add chabot

 

[Tufted Titmouse]

Hedman

Add Habs #1 dman Matheson plz

(joke)

Guhle should slot in around 8th or 9th on this list I reckon

 

[North Cole]

Very nice of Mattias Ekholm to eat all the tough minutes in Nashville so that Josi can rack up points and get votes here.

Hedman, and he should’ve gone first IMO.

I see Hedman/Makar has reached that Crosby/McDavid crossover threshold where McDavid was clearly better but we still had some folks holding on because Crosby was the vet.

 

[wintersej]

Who exactly would you have him above, and what's the rationale?

Makar - even in the year Fox did win his Norris, there's a good argument to make Makar was better, just with 11 games less. Than Smythe (one of the strongest in the past 20 years), and actual Norris win last year, and also beating Fox year 1 for Calder....easy argument of Makar > Fox in the rankings as of today.

Hedman. Fox may have Norris finishes of 1 and 5 in past 3 years, but Hedman has 6 straight Norris finalists nods. Combine that with his incredibly strong Smythe in 2020 followed by 2 more great playoff runs...again, easy to rank Hedman above too.

Josi - 1 Norris and 1 2nd place in last 3 years, which is better 3 year average than Fox. Mostly - last year's regular season is arguably the strongest regular season by a defenseman in many, many years (despite finishing 2nd to Makar for Norris). So again, easy argument to rank him high too.

For what it's worth - I think Adam Fox 100% belongs in the top 4, and that going into next season this top 4 is a tier above rest of league. And maybe you can argue him above Josi if you really want to - but I personally would slot him 4th.

I know you are a Habs fan, but McAvoy is certainly in that group, too.

 

[Machinehead]

Who exactly would you have him above, and what's the rationale?

Makar - even in the year Fox did win his Norris, there's a good argument to make Makar was better, just with 11 games less. Than Smythe (one of the strongest in the past 20 years), and actual Norris win last year, and also beating Fox year 1 for Calder....easy argument of Makar > Fox in the rankings as of today.

Hedman. Fox may have Norris finishes of 1 and 5 in past 3 years, but Hedman has 6 straight Norris finalists nods. Combine that with his incredibly strong Smythe in 2020 followed by 2 more great playoff runs...again, easy to rank Hedman above too.

Josi - 1 Norris and 1 2nd place in last 3 years, which is better 3 year average than Fox. Mostly - last year's regular season is arguably the strongest regular season by a defenseman in many, many years (despite finishing 2nd to Makar for Norris). So again, easy argument to rank him high too.

For what it's worth - I think Adam Fox 100% belongs in the top 4, and that going into next season this top 4 is a tier above rest of league. And maybe you can argue him above Josi if you really want to - but I personally would slot him 4th.

Out of those three, right now, Hedman.

The last Norris nomination Hedman received that wasn't because he's Hedman was probably 2019. I have him more like 4-6 than 1-3.

Which flies in the face of my original point, fair enough.

Over the last three years, there's a close but clear gap and, oddly enough, it's defensively.

 

[stl76]

I think the most underrated players in the game are definitely guys like Ekholm.

As far as deployment goes what is Nashville supposed to do? The other defensemen aren't going to score much and a majority of their forwards are known to do everything but produce points.

Nashville is stacked with defense-first players, they owe an enormous thank you to Josi aswell.

Hope you show up in the center polls because I have a feeling some defensively sound forwards will go a lot lower than they should

I get what you're saying about Nashville having limited options, especially with Ellis leaving (altho Duchene and Forsberg in particular are underrated offensive threats), but it's kind of beside my point. I get WHY they did it, but it still lead to Josi being overrated IMO due to putting up gaudy numbers in very sheltered offensive usage.

I tend to focus less on usage when it comes to forwards TBH. Except for outliers like Bergeron, ROR, etc where their usage is extremely slanted towards difficult defensive minutes.

Out of those three, right now, Hedman.

The last Norris nomination Hedman received that wasn't because he's Hedman was probably 2019. I have him more like 4-6 than 1-3.

Which flies in the face of my original point, fair enough.

View attachment 579737

Over the last three years, there's a close but clear gap and, oddly enough, it's defensively.

What exactly do you believe this chart shows? Please be specific.

 

[Machinehead]

What exactly do you believe this chart shows? Please be specific.

I don't "believe" it. That seems like a disingenuous way of putting it. What's on there is very specific, although you might not recognize it if you're unfamiliar with the abbreviations.

The chart is showing RAPM or Regularized Adjusted Plus Minus. You can break that down into two parts. "Regularized Adjusted" is where your math is. It's a regression model to apply context such as teammates, competition, and different official scoring across venues (more on that in a second).

The only reason that's being done is because folks like to scream "context!!" at stats. Meanwhile, 99% of what's out there has pre-calculated for context since like 2017.

The "plus minus" part is what you think it is. It's goals of course (goals for per 60, or GF/60), scoring chances (expected goals for/against per 60, or xGF/xGA), and shots (corsi for and against, or CF/CA).

Goals against is omitted because no regression model on earth could adjust for goaltending which is the be-all-end-all of goals against. With Vasilevskiy and Shesterkin on their teams, every Lightning and Rangers player is going to be somewhere near the top.

All of these are divided by 60 because that's how long a game is and helps us compare players who don't spend the same amount of time on the ice. It's particularly useful for the daunting task of comparing the overall impact of a forward and a defenseman, because there's a big minutes gap there.

And that's kind of the big misunderstanding. People think analytics measure something abstract. Not really. The abstract part is adjustment and presentation (again, more on that, bear with me). What's being measured is bare bones hockey - goals, shots, scoring chances. And those measurements come from the NHL. All EvolvingHockey does is apply the regression model and make the chart. The numbers are from league official scorers. These official scorers are humans of course with their own patterns, biases, and mistakes, which is why venue adjustment is part of the model.

As far as the chart, "Z-score" might look like a unicorn fart, and there is a bit of algebra to it, but it's just a league average visual. Think of it as a tier system. A regular player is zero. If you're at the top, you're three tiers better than Joe Hockey. This removes the limitation of comparing two guys on two teams and gives more of a league context.

That being said, the numbers being calculated into Z-score and the visual are very real. Elsewhere on the website, you can see (for a quick example) that the Rangers gave up 52.73 shot attempts every 60 minutes Fox was on the ice and the Lightning gave up 49.49 shot attempts every 60 minutes Hedman was on the ice over the three seasons Fox is in the league. That's real. That happened. It's NHL data. You could watch every Rangers game, every Lightning game, count it, and divide it by 60. It's a real documented thing that happened on a rink. The data is from the league.

Now that puts Hedman ahead. But he played for a team that was in three straight Finals, winning two. Fox was on a team that... didn't do that. That's where your adjustment comes in.

That's not to say the charts are perfect. Any on-ice stat has four other guys factoring in and the regression models aren't a magic wand. Maybe the official scoring is off, although it is a three-year sample. Maybe you just think the regression model is a problem but I'm afraid I'm going to ask you to show your work if that's your case.

I do think they present some strong evidence that Fox is a better defender, particularly the large gap in expected goals against.

And when we say that the models aren't perfect, and they're not, remember that we're not either. That's not saying that your visual analysis isn't good or that I know hockey better then anybody else, but nobody can watch every game, particularly comparing guys on different teams. I also think, if I'm arguing Fox is better at defense, that there's a massive height bias and Stanley Cup bias making it sound like a hotter take than it is.

Frankly, I don't know why I bothered but I have something to copy and paste next time somebody asks.

 

[stl76]

I don't "believe" it. That seems like a disingenuous way of putting it. What's on there is very specific, although you might not recognize it if you're unfamiliar with the abbreviations.

The chart is showing RAPM or Regularized Adjusted Plus Minus. You can break that down into two parts. "Regularized Adjusted" is where your math is. It's a regression model to apply context such as teammates, competition, and different official scoring across venues (more on that in a second).

The only reason that's being done is because folks like to scream "context!!" at stats. Meanwhile, 99% of what's out there has pre-calculated for context since like 2017.

The "plus minus" part is what you think it is. It's goals of course (goals for per 60, or GF/60), scoring chances (expected goals for/against per 60, or xGF/xGA), and shots (corsi for and against, or CF/CA).

Goals against is omitted because no regression model on earth could adjust for goaltending which is the be-all-end-all of goals against. With Vasilevskiy and Shesterkin on their teams, every Lightning and Rangers player is going to be somewhere near the top.

All of these are divided by 60 because that's how long a game is and helps us compare players who don't spend the same amount of time on the ice. It's particularly useful for the daunting task of comparing the overall impact of a forward and a defenseman, because there's a big minutes gap there.

And that's kind of the big misunderstanding. People think analytics measure something abstract. Not really. The abstract part is adjustment and presentation (again, more on that, bear with me). What's being measured is bare bones hockey - goals, shots, scoring chances. And those measurements come from the NHL. All EvolvingHockey does is apply the regression model and make the chart. The numbers are from league official scorers. These official scorers are humans of course with their own patterns, biases, and mistakes, which is why venue adjustment is part of the model.

As far as the chart, "Z-score" might look like a unicorn fart, and there is a bit of algebra to it, but it's just a league average visual. Think of it as a tier system. A regular player is zero. If you're at the top, you're three tiers better than Joe Hockey. This removes the limitation of comparing two guys on two teams and gives more of a league context.

That being said, the numbers being calculated into Z-score and the visual are very real. Elsewhere on the website, you can see (for a quick example) that the Rangers gave up 52.73 shot attempts every 60 minutes Fox was on the ice and the Lightning gave up 49.49 shot attempts every 60 minutes Hedman was on the ice over the three seasons Fox is in the league. That's real. That happened. It's NHL data. You could watch every Rangers game, every Lightning game, count it, and divide it by 60. It's a real documented thing that happened on a rink. The data is from the league.

Now that puts Hedman ahead. But he played for a team that was in three straight Finals, winning two. Fox was on a team that... didn't do that. That's where your adjustment comes in.

That's not to say the charts are perfect. Any on-ice stat has four other guys factoring in and the regression models aren't a magic wand. Maybe the official scoring is off, although it is a three-year sample. Maybe you just think the regression model is a problem but I'm afraid I'm going to ask you to show your work if that's your case.

I do think they present some strong evidence that Fox is a better defender, particularly the large gap in expected goals against.

And when we say that the models aren't perfect, and they're not, remember that we're not either. That's not saying that your visual analysis isn't good or that I know hockey better then anybody else, but nobody can watch every game, particularly comparing guys on different teams. I also think, if I'm arguing Fox is better at defense, that there's a massive height bias and Stanley Cup bias making it sound like a hotter take than it is.

Frankly, I don't know why I bothered but I have something to copy and paste next time somebody asks.

What are you, paid by the word?



Sincerely appreciate the time it took you to write all that but you’ve completely misunderstood my question. What is your analysis of the data? People lazily post these charts ALL THE TIME as if data is an argument. It’s not. It’s data, and I believe, often flawed models that poorly capture defensive play in the NHL.

Sounds like your thesis is that Fox is better defensively because xGF and xGA?

If you actually believe regression current modes can fully account for usage and deployment, especially for defenseman, then we are just going to disagree. For instance, how can a model account for the minutes that Josi didn’t have to play because Ekholm did the heavy lifting? The data doesn’t capture he benefit an offensive dman gets from not having to play tough defensive minutes. Instead people too often post show colorful charts showing team based shooting metrics and then poorly interpret what the data is actually capturing and often inappropriately extrapolate it to defensive ability.

EDIT: and just for the record, I love Fox. I’d have him #3 ahead of Josi. Just a fantastic dman!

 

[Machinehead]

What are you, paid by the word?



Sincerely appreciate the time it took you to write all that but you’ve completely misunderstood my question. What is your analysis of the data? People lazily post these charts ALL THE TIME as if data is an argument. It’s not. It’s data, and I believe, often flawed models that poorly capture defensive play in the NHL.

Sounds like your thesis is that Fox is better defensively because xGF and xGA?

If you actually believe regression current modes can fully account for usage and deployment, especially for defenseman, then we are just going to disagree. For instance, how can a model account for the minutes that Josi didn’t have to play because Ekholm did the heavy lifting? The data doesn’t capture he benefit an offensive dman gets from not having to play tough defensive minutes. Instead people too often post show colorful charts showing team based shooting metrics and then poorly interpret what the data is actually capturing and often inappropriately extrapolate it to defensive ability.

Partially, yes, I think Fox is better defensively because the numbers show he's better at preventing xGA. I understand your point that the data doesn't necessarily capture the full breadth of defensive play but at a certain point, I kind of...don't care? If you're preventing chances that much better than the next guy over a three-season sample, I'm not really asking how you got there.

It also passes the eye test to be completely honest, if you ignore reputation, which is what most people mean by "eye test."

I don't know about fully accounting. No tool can do that which is why we have different opinions. It would be boring if we just knew who was better. That said, usage and deployment is another reason. Fox actually plays insane minutes. The year he won the Norris, he had top 5 hardest competition in the league. K'Andre Miller got a bigger role this past year, but the top defensive assignments are still split at best. I think it's just assumed he has offensive deployment because he scores. That, and he does start in the offensive zone quite a bit. Zone starts are meaningless but that's another whole long rant.

He has harder deployment than Hedman in terms of competition. The McDonagh pair was actually the one that did the most shit-shoveling on Tampa's Cup teams. Which is fine. On the other side of that coin, I think some folks die at the alter of usage. Hedman puts up points and should be allowed to, and I certainly think he could play hard minutes if asked to, which he might be now that McDonagh is gone. You shouldn't have to play defensive minutes if you're a star player and your team is better if you don't.

However, we're kind of splitting hairs in the big four, and Fox also puts up the points while being the only one of the four who's really on his team's shutdown pair.

 

[Lavar Ball]

What are you, paid by the word?



Sincerely appreciate the time it took you to write all that but you’ve completely misunderstood my question. What is your analysis of the data? People lazily post these charts ALL THE TIME as if data is an argument. It’s not. It’s data, and I believe, often flawed models that poorly capture defensive play in the NHL.

Sounds like your thesis is that Fox is better defensively because xGF and xGA?

If you actually believe regression current modes can fully account for usage and deployment, especially for defenseman, then we are just going to disagree. For instance, how can a model account for the minutes that Josi didn’t have to play because Ekholm did the heavy lifting? The data doesn’t capture he benefit an offensive dman gets from not having to play tough defensive minutes. Instead people too often post show colorful charts showing team based shooting metrics and then poorly interpret what the data is actually capturing and often inappropriately extrapolate it to defensive ability.

EDIT: and just for the record, I love Fox. I’d have him #3 ahead of Josi. Just a fantastic dman!

You: Please be specific

*Is Specific*

You: What are you, paid by the word?