Speculation: Caps General Discussion (Coaching/FAs/Cap/Lines etc) - 2021-22 Season Part 7: Off-season Edition

[Devil Dancer]

Kalopsia is right, the chart is broken unless we are to believe that players, on average, peak at 18 or younger.

Or the problem could be that the underlying WAR stat is useless trash. Which it certainly is, all hockey WAR stats are made up garbage that don't quantify anything, they just jumble together a bunch of different stats to varying degrees.

But the chart may also be broken.

 

[twabby]

Ok, am I missing something here? This is a "rate of change" graph that's entirely negative. If I'm remembering my precalc right, what this is saying is that players basically decline every year their whole careers. It implies they peak at 18, decline until they briefly level off in their early-mid 20s, and then resume their decline starting at 25 or 26 for D and F respectively. This doesn't mesh with the baseball models they borrowed their technique from. In those, they represent the aging curve as "percentage of peak," rather than year to year change, and they get the results you'd expect - players improve pretty consistently until 26/27, then decline. (Can't insert the pic of the graph for some reason but here's the link) I dunno what's going on with EvolvingWild's aging curve, but he's clearly doing something wrong.

I agree the graph is labeled in a confusing way, so perhaps it's better to look at the table that the graph is derived from:

And the accompanying quote clarifies what the "cumulative difference" column represents:

The simplest way to approach this is by taking each player’s respective change year-over-year in Overall WAR, totaling that change for each “age bucket” (18-19, 25-26, etc.), and then calculating the average for each age bucket. We can convert the average change per age into a cumulative change and adjust the peak to 0 – this is how we visualize the aging curve.

I'd basically just concentrate on the "Avg Change In Overall WAR" column. It suggests that forwards improve in their early 20s and peak from ages ~22-25 on average, and defensemen start at their peak and just steadily decline once they reach 25.

 

[twabby]

Kalopsia is right, the chart is broken unless we are to believe that players, on average, peak at 18 or younger.

Or the problem could be that the underlying WAR stat is useless trash. Which it certainly is, all hockey WAR stats are made up garbage that don't quantify anything, they just jumble together a bunch of different stats to varying degrees.

But the chart may also be broken.

Maybe the underlying WAR stat is useless trash, however I would point out that the author of the WAR stat that is cited in this article is now the Director of Analytics for the Colorado Avalanche.

Barring a rigorous refutation of the WAR stat on your part, I'm going to assume that the Colorado Avalanche (2022 conference finalists and 2021 President's Trophy winners) hiring Dawson Sprigings is a good indicator that the WAR stat and his work in general is not useless trash, but is rather useful trash. Perhaps a nice organic compost?

 

[HTFN]

I'm going to assume Fehervary being a good 1v1 defender is more or less correct (I don't watch the games, I don't know if it's true or not). But it then poses the following questions:

1. How important is 1v1 defense to a player's overall impact? If he was a good 1v1 defender yet his overall defensive impact was so bad, doesn't this suggest that maybe 1v1 defense isn't all that important of a component if he struggles in other parts of the game?

2. How important is in-zone coverage to a player's overall impact? If he was bad in this portion of the game and his overall impact defensive impact was negative, it's consistent with the idea that in-zone coverage may be quite important.

3. What is the expected magnitude of Fehervary's improvements going forward? How much better do defensive reads and in-zone coverage really get with experience? The data seems to suggest that defensemen in aggregate don't really make a ton of defensive improvements over time. Perhaps defensive ability is more of a natural ability than an acquired skill, and if that's the case then is it really reasonable to expect much improvement over time?

It’s not about how important it is, nobody’s going to argue that poor defense is desirable. The point is when you’re looking at a young player like that it’s reassuring to see his faults are the type that can be improved through repetition and experience.

More reps in the NHL is more likely to help your in zone coverage improve, your play recognition, etc. than it is to help you keep up with good skaters 1v1 and not get turnstiled, so if you’re comparing two players you’re probably going to want the one with the tools. Knowing when to switch, who to chase, etc. is not instinctive in a systems game

This is (one of) the reasons defensemen take longer to cook, and frankly if your data is saying that defensemen generally don’t improve in the aggregate I’d suggest you stop looking at the aggregate. Depending on what you’re sampling I think it’s very likely a lot of the improvements we’re talking about here are being lost to the noise of journeymen and serviceable 5/6 D who hold the spot for their athletic primes and don’t project to grow to begin with.

 

[HTFN]

I agree the graph is labeled in a confusing way, so perhaps it's better to look at the table that the graph is derived from:

View attachment 553243

And the accompanying quote clarifies what the "cumulative difference" column represents:



I'd basically just concentrate on the "Avg Change In Overall WAR" column. It suggests that forwards improve in their early 20s and peak from ages ~22-25 on average, and defensemen start at their peak and just steadily decline once they reach 25.

From any sample? Like… doesn’t matter if you’re a 1st overall pick or 5th round, first line or fourth line? Because I’d think a bunch of mostly average players are going to drag the average down as they crater right on the other side of their athletic prime.

Kind of seems like this graph really just says the league’s oldest talented players can’t drag the WAR up when their draft year starts fading physically, but I’d bet most teams have a top 6/4 player pushing 30+ who defies this chart. We already know the development arc doesn’t track with all players equally so it’s up to you to decide what this data is actually saying.

 

[Carlzner]

I don't. Player development seems boring and I don't have time to worry about that.

But I think I do have a reasonable grasp on the frequency at which players improve from their rookie years, and the magnitude of said improvements.

A New Look at Aging Curves for NHL Skaters (part 1)

How do NHL players age? When do they peak? How quickly do they decline? Questions about player aging in the NHL have been debated for years, and an incredible amount of research has already been do…

hockey-graphs.com

In particular, it appears that defensemen don't really get better as they age, but rather are most impactful as soon as they enter the league. Here's a chart to that effect:

View attachment 553177

And this jives with current results, right? Adam Fox and Cale Makar are already incredibly impactful right now. Erik Karlsson and PK Subban were winning Norris Trophies in their early 20s, now they both aren't very good. John Carlson is seemingly no better or worse now than he was in the 2013-15 period when he was 23-25 years old. Alzner didn't really change much from when he was a rookie. Mike Green didn't really get any better over time.

There are exceptions of course, but the general trend is that defensemen kind of are who they are when they enter the league. Unless there is a specific reason to believe Fehervary will buck this trend, then I think they have to enter next season assuming Fehervary is going to struggle and need to have a contingency plan.

Fehervary contracting COVID is really the only explanation that can explain his poor performance last year, aside from him just not being good. If COVID is why he struggled then I'd be confident in a bounceback season. Otherwise I'm going to be a little pessimistic about his future performance.

On a positive note:, I believe Connor McMichael could be primed for a big improvement next year. If he gets an opportunity.

Yeah that’s a nice chart and all but I more-so meant the fact that player development hinges on the organization that has the hockey player under contract and how they decide to dictate the offseason for said player.

The original comparison of Risto and Fever based on a few charts is completely devoid of any kind of organizational development aspect. The Sabres haven’t developed a good hockey player in over a decade.

 

[twabby]

Yeah that’s a nice chart and all but I more-so meant the fact that player development hinges on the organization that has the hockey player under contract and how they decide to dictate the offseason for said player.

The original comparison of Risto and Fever based on a few charts is completely devoid of any kind of organizational development aspect. The Sabres haven’t developed a good hockey player in over a decade.

I guess that’s a fair point about Buffalo, but in Washington’s case which drafted defensemen have made big improvements from their early years?

I already mentioned it but Carlson, Alzner, Green, Schmidt, Orlov, Siegenthaler, Carrick, Djoos, Jeff Schultz, Sloan, etc. all made very little improvement over their rookie seasons. Why should we expect Fehervary to buck this trend? (Again, unless it’s COVID-related and he fully recovers this offseason.)

 

[twabby]

From any sample? Like… doesn’t matter if you’re a 1st overall pick or 5th round, first line or fourth line? Because I’d think a bunch of mostly average players are going to drag the average down as they crater right on the other side of their athletic prime.

Kind of seems like this graph really just says the league’s oldest talented players can’t drag the WAR up when their draft year starts fading physically, but I’d bet most teams have a top 6/4 player pushing 30+ who defies this chart. We already know the development arc doesn’t track with all players equally so it’s up to you to decide what this data is actually saying.

I mean you can go through the work of stratifying the data if you want. It’ll introduce more uncertainty the more you stratify it. Should we restrict Fehervary’s comparables to only second round picks? Should we restrict it to only other 46th overall picks who started on the nominal top pairing? Overstratifying like this almost always yields very little predictive value.

 

[Carlzner]

I guess that’s a fair point about Buffalo, but in Washington’s case which drafted defensemen have made big improvements from their early years?

I already mentioned it but Carlson, Alzner, Green, Schmidt, Orlov, Siegenthaler, Carrick, Djoos, Jeff Schultz, Sloan, etc. all made very little improvement over their rookie seasons. Why should we expect Fehervary to buck this trend? (Again, unless it’s COVID-related and he fully recovers this offseason.)

Huh?

That’s actually quite a solid list of defenseman developed over a ~10 year span. Even Carrick carved out a nice NHL career for an undersized 5th rounder.

Sure there’s no Makar’s or Fox’s on that list… there also aren’t any of those caliber defenseman for like 25 other franchises. And nobody is expecting Fever to be Makar or Fox or Hedman. That doesn’t mean you dump him, or that he won’t develop into a better player.

 

[twabby]

Huh?

That’s actually quite a solid list of defenseman developed over a ~10 year span. Even Carrick carved out a nice NHL career for an undersized 5th rounder.

Sure there’s no Makar’s or Fox’s on that list… there also aren’t any of those caliber defenseman for like 25 other franchises. And nobody is expecting Fever to be Makar or Fox or Hedman. That doesn’t mean you dump him, or that he won’t develop into a better player.

But again those players didn't really improve after they entered the NHL. Fehervary is much closer to Tyler Sloan and Madison Bowey in terms of initial results than he is to just decent players like Karl Alzner and Nate Schmidt. There needs to be vast improvement in Fehervary's game in order for him to really ever become a viable top 4 NHL defenseman, and so far no Washington defensemen in recent memory have really made vast improvements over their rookie seasons.

I'm not saying dump him for nothing. I'm saying that if management is banking on him taking a big step next season without having a good contingency plan in case he falters then they are probably making a big mistake.

 

[Jags]

I already mentioned it but Carlson, Alzner, Green, Schmidt, Orlov, Siegenthaler, Carrick, Djoos, Jeff Schultz, Sloan, etc. all made very little improvement over their rookie seasons.

So you're saying that players like Carlson, Alzner, Orlov, and Green all arrived in the NHL pretty much fully cooked? All those guys took years to round out their games at the NHL level. And Fehervary is comparable to Bowey how, exactly? Bowey was dogshit at the NHL level from beginning to end. Fever looks good. Not great, but it's early. Maturity, experience, continuity, and repetition could easily help him improve and steady his game.

But second and third pair defensemen are pretty easy to get.

But its even easier and cheaper and has more upside to keep the good, homegrown ones. WAAAYYYY more bang for your buck.

 

[HTFN]

I mean you can go through the work of stratifying the data if you want. It’ll introduce more uncertainty the more you stratify it. Should we restrict Fehervary’s comparables to only second round picks? Should we restrict it to only other 46th overall picks who started on the nominal top pairing? Overstratifying like this almost always yields very little predictive value.

Kind of seems like there isn’t any predictive value to begin with if the prediction is “everyone gets worse immediately and decay is king”

Introducing more uncertainty is kind of the point because what we’re “certain” of is a little stupid and doesn’t pass the smell test. If all that data is just noise pulled from all corners and backgrounds it… becomes just as useless to you all over again. It’s like taking all the weather forecasts in the United States and putting together one average temperature, then telling the entire country how it should dress next week.

This chart has too much noise that I can’t see any adjustment for, including bad players and bottom 6 guys who may play 3-7 years total, old men performing into their 30’s, and it also didn’t seem to span a very long time. It seems apparent to me that different starting points and draft classes would have a heavy influence on the beginning of the data set, so charting everyone’s WAR by age across one year is a lot less interesting to me than tracking the career WAR of various players and draft classes.

There’s way too much going on to take anything meaningful from that chart and this one really should have made you pause. Also if I’m reading this correctly it’s measuring the change in WAR, so wouldn’t it almost always decline? It seems like even very good players treading water would not be nearly enough to bring up a league average? Like, every all-star above the age of 23 would have to not just equal their current WAR but exceed it at the rate others are losing theirs just to break even, and since we know there will always be more scrubs than all-stars… everyone who is a pretty good player seems completely lost in this mix.

Edit: for clarification I see the line graph covers 08-16 but didn’t see that in the tables posted later, so that’s what the “one year” WAR thing refers to

 

[Holtbyisms]

But again those players didn't really improve after they entered the NHL. Fehervary is much closer to Tyler Sloan and Madison Bowey in terms of initial results than he is to just decent players like Karl Alzner and Nate Schmidt. There needs to be vast improvement in Fehervary's game in order for him to really ever become a viable top 4 NHL defenseman, and so far no Washington defensemen in recent memory have really made vast improvements over their rookie seasons.

I'm not saying dump him for nothing. I'm saying that if management is banking on him taking a big step next season without having a good contingency plan in case he falters then they are probably making a big mistake.

I lol'ed

 

[Langway]

Never go full chart autist.

Probably should do something about all of the senior citizens littering the roster first. They must be ECHL quality by now.

 

[twabby]

So you're saying that players like Carlson, Alzner, Orlov, and Green all arrived in the NHL pretty much fully cooked? All those guys took years to round out their games at the NHL level. And Fehervary is comparable to Bowey how, exactly? Bowey was dogshit at the NHL level from beginning to end. Fever looks good. Not great, but it's early. Maturity, experience, continuity, and repetition could easily help him improve and steady his game.



But its even easier and cheaper and has more upside to keep the good, homegrown ones. WAAAYYYY more bang for your buck.

Yes, they were mostly fully cooked by the time they entered the NHL. They were each given more opportunity as they progressed through their careers, but they all showed about the same from their rookie seasons on. Has Carlson really improved over time, or did he just get more ice time as Mike Green faded from the picture? Was Alzner really any better in 2014-15 than he was in 2010-11? Has Orlov really improved that much since his rookie season where he was extremely good, or did the coaches finally just decide to give him more responsibility?

Mike Green did appear to take a big step between his rookie and sophomore seasons based on points, but his "development" from that point on stagnated. However since his rookie year was in 2006-07 there is no WAR data for him (NHL RTSS data started being collected in 2007-08), so I'd be interested to see how good his rookie year was from an impact standpoint rather than just points. It's very possible that Green had a very good rookie year from an impact standpoint but was let down by poor teammates.

Fehervary is similar to Bowey because both were below replacement level players in their rookie years. To be fair, Bowey was more below replacement than Fehervary. But the point still stands that Fehervary's on-ice impact in his rookie season is closer to Bowey's rookie season than Karl Alzner's.

I think once again the main downside of the eye-test has become apparent: the eye-test measures style rather than substance. If a guy can look good yet still provide a negative on-ice impact, what does it matter if he looks good? If Fehervary didn't hit people so much, I kind of doubt whether fans would think so highly of him, for instance.

It's easier and cheaper to just bank on Fehervary going forward, but I don't really know that there's more upside to keeping him. I guess he could break out but after his disappointing rookie season the odds of him breaking out have to be slim at this point. Again, unless COVID is to blame and he makes a full recovery this offseason.

 

[CapitalsCupReality]

Rookie defenseman don’t improve and don’t have upside……you heard it here first. Groundbreaking shit….

 

[HTFN]

This might be your single worst take… ever.

Nobody ever improves, just gets more ice time.

 

[twabby]

Kind of seems like there isn’t any predictive value to begin with if the prediction is “everyone gets worse immediately and decay is king”

Introducing more uncertainty is kind of the point because what we’re “certain” of is a little stupid and doesn’t pass the smell test. If all that data is just noise pulled from all corners and backgrounds it… becomes just as useless to you all over again. It’s like taking all the weather forecasts in the United States and putting together one average temperature, then telling the entire country how it should dress next week.

This chart has too much noise that I can’t see any adjustment for, including bad players and bottom 6 guys who may play 3-7 years total, old men performing into their 30’s, and it also didn’t seem to span a very long time. It seems apparent to me that different starting points and draft classes would have a heavy influence on the beginning of the data set, so charting everyone’s WAR by age across one year is a lot less interesting to me than tracking the career WAR of various players and draft classes.

There’s way too much going on to take anything meaningful from that chart and this one really should have made you pause. Also if I’m reading this correctly it’s measuring the change in WAR, so wouldn’t it almost always decline? It seems like even very good players treading water would not be nearly enough to bring up a league average? Like, every all-star above the age of 23 would have to not just equal their current WAR but exceed it at the rate others are losing theirs just to break even, and since we know there will always be more scrubs than all-stars… everyone who is a pretty good player seems completely lost in this mix.

Edit: for clarification I see the line graph covers 08-16 but didn’t see that in the tables posted later, so that’s what the “one year” WAR thing refers to

If the conclusion were really "everyone gets worse immediately and decay is king" there would be a ton of predictive value. That's not exactly the conclusion. It appears that defensemen on average stagnate from ages 20 to 24 before seeing their play gradually drop off.

The important thing for me is that it appears that the on-ice impact for a rookie defenseman, especially a 22 year old, is quite predictive. Namely, if a rookie has a below replacement level impact, as Fehervary did, then the odds are stacked against him putting together an impactful career.

This might be your single worst take… ever.

Nobody ever improves, just gets more ice time.

I never said nobody improves. I said on average defensemen don't improve much after their rookie years.

Given Fehervary's low baseline performance to start, he would not only need to improve but he'd need to improve a lot in order to become an effective top 4 defenseman. The odds seem stacked against that happening.

 

[Holtbyisms]

This might be your single worst take… ever.

Nobody ever improves, just gets more ice time.

This is a borderline banished to hack level for all of HFCaps history type of take.

 

[g00n]

I hate the offseason.

 

[Kalopsia]

I mean you can go through the work of stratifying the data if you want. It’ll introduce more uncertainty the more you stratify it. Should we restrict Fehervary’s comparables to only second round picks? Should we restrict it to only other 46th overall picks who started on the nominal top pairing? Overstratifying like this almost always yields very little predictive value.

Well, we can't because they didn't make the underlying data available and the WAR data it's based on is in an invite-only google doc, but setting that aside...

There's always some degree of restriction in any study. Read any medical study and check out the selection criteria - they make the population as homogenous as they possibly can while maintaining a big enough sample size for their desired statistical power. If you were diagnosed with cancer and asked your doctor what your chances of surviving were, would you want him to give you the numbers for everyone with your cancer, or would you want the numbers for people at your stage, at roughly the same age and with roughly the same physical health? Over-stratifying lowers predictive value, but under-stratifying risks leaving confounding factors that invalidate the conclusions.

In this case, lumping together the top 210 defensemen in TOI each year means you're getting a ton of scrubs and players who missed time with injuries added into the mix, which makes using it to judge the development of a top pairing caliber defenseman highly dubious. I think they would've been better off limiting it to the top 120 and doing a similar cutoff for forwards.

Yes, they were mostly fully cooked by the time they entered the NHL. They were each given more opportunity as they progressed through their careers, but they all showed about the same from their rookie seasons on. Has Carlson really improved over time, or did he just get more ice time as Mike Green faded from the picture? Was Alzner really any better in 2014-15 than he was in 2010-11? Has Orlov really improved that much since his rookie season where he was extremely good, or did the coaches finally just decide to give him more responsibility?

Mike Green did appear to take a big step between his rookie and sophomore seasons based on points, but his "development" from that point on stagnated. However since his rookie year was in 2006-07 there is no WAR data for him (NHL RTSS data started being collected in 2007-08), so I'd be interested to see how good his rookie year was from an impact standpoint rather than just points. It's very possible that Green had a very good rookie year from an impact standpoint but was let down by poor teammates.

Fehervary is similar to Bowey because both were below replacement level players in their rookie years. To be fair, Bowey was more below replacement than Fehervary. But the point still stands that Fehervary's on-ice impact in his rookie season is closer to Bowey's rookie season than Karl Alzner's.

I think once again the main downside of the eye-test has become apparent: the eye-test measures style rather than substance. If a guy can look good yet still provide a negative on-ice impact, what does it matter if he looks good? If Fehervary didn't hit people so much, I kind of doubt whether fans would think so highly of him, for instance.

It's easier and cheaper to just bank on Fehervary going forward, but I don't really know that there's more upside to keeping him. I guess he could break out but after his disappointing rookie season the odds of him breaking out have to be slim at this point. Again, unless COVID is to blame and he makes a full recovery this offseason.

I generally think this board is too hard on you, but this take is just waaaaay out there man. Seriously, which do you think is more likely? That you've discovered a massive flaw in our understanding of hockey development that's gone unnoticed by more than a century of players, coaches, and fans, or that there's a flaw in the analysis (under-stratifying like I said above) or in the WAR stat (failing to fully account for the difference in difficulty between sheltered and un-sheltered usage)?

 

[Jags]

I hate the offseason.

I like it, seeing all the moving pieces settle, hoping we maneuver better than the other guy. It's one of my favorite times of year, but it's like a beach vacation and sometimes twabby is the rain.

I dig all the data you bring to the table, twabby, but your interpretation gets pretty wonky sometimes.

 

[Jags]

If you were diagnosed with cancer and asked your doctor what your chances of surviving were, would you want him to give you the numbers for everyone with your cancer, or would you want the numbers for people at your stage, at roughly the same age and with roughly the same physical health? Over-stratifying lowers predictive value, but under-stratifying risks leaving confounding factors that invalidate the conclusions.

Really hoping you stick around and post more. Really enjoying your contributions lately.

I generally think this board is too hard on you

We are. And it's not always fair. He adds a lot to the conversation, and we all have our flaws that irk others. One of his -- and it takes time to see the patterns -- is never really being wrong. You'll have discussions with him that go on for pages and days, and you'll think you've made some kind of progress, but then a week or a month or a year later he goes back to saying the same exact things.

And clearly one of mine is being the jackass that can't let shit go.

 

[twabby]

Well, we can't because they didn't make the underlying data available and the WAR data it's based on is in an invite-only google doc, but setting that aside...

There's always some degree of restriction in any study. Read any medical study and check out the selection criteria - they make the population as homogenous as they possibly can while maintaining a big enough sample size for their desired statistical power. If you were diagnosed with cancer and asked your doctor what your chances of surviving were, would you want him to give you the numbers for everyone with your cancer, or would you want the numbers for people at your stage, at roughly the same age and with roughly the same physical health? Over-stratifying lowers predictive value, but under-stratifying risks leaving confounding factors that invalidate the conclusions.

In this case, lumping together the top 210 defensemen in TOI each year means you're getting a ton of scrubs and players who missed time with injuries added into the mix, which makes using it to judge the development of a top pairing caliber defenseman highly dubious. I think they would've been better off limiting it to the top 120 and doing a similar cutoff for forwards.



I generally think this board is too hard on you, but this take is just waaaaay out there man. Seriously, which do you think is more likely? That you've discovered a massive flaw in our understanding of hockey development that's gone unnoticed by more than a century of players, coaches, and fans, or that there's a flaw in the analysis (under-stratifying like I said above) or in the WAR stat (failing to fully account for the difference in difficulty between sheltered and un-sheltered usage)?

Medical studies often involve thousands of participants such that it's possible to make reasonable conclusions not only based on the population as a whole, but in these stratified subsamples which can also contain hundreds of participants. We don't have that luxury in hockey. There are only in the hundreds of players that we can draw conclusions on in total, so stratifying is going to be subject to incredible amounts of noise.

Give me a stratification to work with that includes Fehervary, and I'll try to parse the WAR data. Do you want me to see how top 2 rounders perform after their rookie years? Do you want to see how Slovakian players do after their first year? Do you want to see how players who get more than X minutes in their rookie year perform in subsequent years? I'll try my best to parse through this data and post results. It'll likely have to use Evolving Hockey's WAR stat since I have access to their data, but it'll probably be close enough to Sprigings's WAR stat.

We shouldn't pretend that hockey is a solved problem. It wasn't that long ago where defensemen were still drafted because of size above most else, for instance. Fourth lines up until about 10 years ago consisted mainly of just dudes who could punch. Hockey is way different now than it was even 5-6 years ago.

Is it really unreasonable to think that maybe our ideas about player development timelines might be out of whack? Is it really that crazy to believe that teams give their young players who underperform, especially draftees, way more rope than they would otherwise deserve? Is it insane to think that those rookies who have shown well are not given the same opportunities as established veterans who are not as good?

I'm not saying that players don't develop. But in the case of defensemen, I suspect that almost all of that development is done before they even reach the NHL, on average.

 

[Raikkonen]

First I banned him. Now its not enough!

Ffs, ban me someone. I cant handle reading the same things for next 4 months.