All Purpose Analytics and Extended Stats Discussion

[g00n]

Please feel free to ignore that part of the post! There were other player cards posted!

I'm free to not ignore it, too. Just as you are free to not post redundant argumentss that have been debunked already.

 

[twabby]

Plus, again, it ignores the numerous quotes from staff and the player himself talking about the differences and what has changed. We just want to ignore all this because you can’t quantify it and then act like nothing is different, because remember: if the numbers don’t change, nothing can be different.

You’re right, in a way. Statistical analysis does not care about narratives. It’s one of the best parts about it!

 

[twabby]

I'm free to not ignore it, too. Just as you are free to not post redundant argumentss that have been debunked already.

It’s less a redundant argument and more just another data point in favor of Kuznetsov’s on-ice turnaround last year continuing into this year.

It’s also interesting to see how people’s perceptions of Nick Jensen are so much different this year than last despite the stats and player cards being similar. Jensen, like Kuznetsov, has been given more opportunity this year than last and I think that has been the major difference. He’s always been a strong player. He’s been a little bit better this year in terms of offensive results (that I suspect are unsustainable given the disparity between his xGAR and GAR), but he’s always had shutdown D results.

They were quite lucky that Seattle botched the expansion draft and selected Vanecek instead of Jensen!

 

[HTFN]

You’re right, in a way. Statistical analysis does not care about narratives. It’s one of the best parts about it!

You're being deliberately obtuse here and I think you know that. The numbers are only pure until someone decides to use them for their, wait for it, narrative. Everything is a narrative if you want to go that route.

I'm right because you've got the player and organization both talking about changing intangible, non-statistic things. What you feel about it doesn't really matter, people are just trying to help you grow the f*** up.

 

[twabby]

You're being deliberately obtuse here and I think you know that. The numbers are only pure until someone decides to use them for their, wait for it, narrative. Everything is a narrative if you want to go that route.

I'm right because you've got the player and organization both talking about changing intangible, non-statistic things. What you feel about it doesn't really matter, people are just trying to help you grow the f*** up.

The player and the organization have talked about changing the inputs. Training harder, rededicating himself to the game, etc. are all inputs, and the statistics are indifferent to these inputs.

The statistics only describe outputs. In this case, the GAR and xGAR components only describe a player's individual impact on goal differential. Games are won and lost based on goal differential, not on which players and teams train the hardest or other inputs.

The inputs between this year's version and last year's version of Evgeny Kuznetsov could be and probably are completely different. But the outputs, according to several statistical models, appear to be similar.

 

[g00n]

The player and the organization have talked about changing the inputs. Training harder, rededicating himself to the game, etc. are all inputs, and the statistics are indifferent to these inputs.

The statistics only describe outputs. In this case, the GAR and xGAR components only describe a player's individual impact on goal differential. Games are won and lost based on goal differential, not on which players and teams train the hardest or other inputs.

The inputs between this year's version and last year's version of Evgeny Kuznetsov could be and probably are completely different. But the outputs, according to several statistical models, appear to be similar.

No, these are the limited "outputs" you choose to focus on, not the actual output by the player that might contribute to winning.

But you'll never see this because you don't value anything that isn't a fancy stat.

Yes I know this is a fancy stat thread, but expect some disagreement when you say things like that.

 

[twabby]

No, these are the limited "outputs" you choose to focus on, not the actual output by the player that might contribute to winning.

But you'll never see this because you don't value anything that isn't a fancy stat.

Yes I know this is a fancy stat thread, but expect some disagreement when you say things like that.

The outputs I focus on are all related to goal differential, and a player's impact on goal differential.

How else are hockey games judged if not goal differential?

 

[g00n]

The outputs I focus on are all related to goal differential, and a player's impact on goal differential.

How else are hockey games judged if not goal differential?

Oh we're judging hockey games, and not overall player performance when we say X player is no different than last year?

 

[twabby]

Oh we're judging hockey games, and not overall player performance when we say X player is no different than last year?

Yes, his impact on goal differential this year appears to be quite similar to his impact last year.

His impact on our hearts might be quite a bit different!

 

[hockeykicker]

MOD NOTE: We recognize that not everyone is into deep statistical analysis. This is a thread for "extended stats" discussions. Stats can still be used in the general roster discussion to make a single point, but if someone wants to debate a player's statistical value ad nauseum, or engage in back and forth about the value of X or Y stat, it must go here. If you have extended analysis that applies in another thread, rather than jam up that thread with large charts and so forth you may want to link to a post here, or use a collapse tag in your original post. Contact a mod if you do not know how to do this.

From the first post in thread. In case anyone has forgotten.

 

[twabby]

Micah just released his blueline traversals research, which estimates players' and coaches' impacts on zone entry offense/defense and zone exit offense/defense, along with players' off-ice impacts. The full analysis and model description is below:

Blueline Traversals

In regards to Washington, here are this year's results for zone entries and exits:

No surprise to see that Orlov does quite well at exiting the defensive zone quickly and entering the offensive zone quickly, but I am kind of surprised to see Kuznetsov so low in terms of zone-entries in particular. Hathaway being this strong is also a little surprising given his role, but he really has had a great season so I'm not too surprised.

Wilson being strong on zone exit defense is a good sign, and indicates that his forechecking truly has been disruptive this year and prevents the opposition from collecting the puck and exiting cleanly. I'm pretty surprised that Schultz has been their strongest zone-entry defender this year, but it does jive with his big uptick in terms of other macrostats such as on-ice xGF%. No surprise to see guys like Ovechkin and Sprong struggle. They aren't particularly strong in the neutral zone preventing zone entries against, and they aren't particularly strong possession players or forecheckers either so zone exit defense isn't really expected to be great.

The off-ice impacts are as follows:

The basic gist of this metric is how do players impact the zone-starts of their teammates? Daniel Sprong appears to do an excellent job leaving his teammates in a better position when his shift is over. I suspect this is because he shoots a lot and perhaps this leads to either goals or offensive zone faceoffs (frozen pucks, pucks deflected out of play, etc.) at higher rates than his teammates, but I'm not sure.

On the other hand, it appears that Carl Hagelin struggles at leaving his team in a better position when he leaves the ice. Again, not too surprising given the season he's had, particularly when it comes to killing offensive zone possessions with poor passing. I'd love to see this data from last year because Hagelin was a much stronger defensive player last year, but unfortunately I do not think Micah computed the off-ice impacts for any year other than this year.

 

[Hivemind]

Will be interesting to compare this to Corey S' zone entry & exit statistics after the season is over (or now, if anyone is a Patreon of Corey S and has access to his current season workbook).

 

[Langway]

Also surprised to see Sheary/Jensen/Wilson so low in that first one.

 

[twabby]

Will be interesting to compare this to Corey S' zone entry & exit statistics after the season is over (or now, if anyone is a Patreon of Corey S and has access to his current season workbook).

Yeah I imagine Corey's data is more accurate since he is tracking entries and exits directly, rather than imputing them from the NHL play by play data. I haven't signed up for his Patreon yet, but I'm thinking about it.

 

[Hivemind]

Yeah I imagine Corey's data is more accurate since he is tracking entries and exits directly, rather than imputing them from the NHL play by play data. I haven't signed up for his Patreon yet, but I'm thinking about it.

Part of me feels like I should sign up for his Patreon given how much I enjoy his work and want to support him, but the rest of win feels that paying money to have more stats for folks on HFBoards to ignore isn't a productive use of my wallet.

 

[twabby]

Stanley Cup contender checklist: How does each playoff team stack up?

Referencing the Cup Checklist, we looked at what each NHL playoff team already has and what it still might need to be considered Cup-worthy.

theathletic.com

It’s good, but missing a few key ingredients​

Washington​

Dom: The Capitals are a deep team, but the roster isn’t exactly overwhelming in any area. Aside from Alex Ovechkin’s miracle season, there isn’t enough gold on the roster to be overly optimistic about this team’s chances of contending. High-end talent has been an issue in every season since Washington won the Cup and it remains one this season.

Ovechkin and John Carlson are capable members of an elite core, but the talent around those two is in decline. Down the middle, neither of Evgeny Kuznetsov or Nicklas Backstrom will cut it as the top center. Though Kuznetsov has nearly played well enough this season, that’s undermined by Backstrom doing the exact opposite behind him. The expectation of a Cup winner’s top two centers is 6.2 wins. Washington’s duo is closer to four.

If you believe those two can find their form come playoff time, that still leaves Washington with a pretty big problem in net. Neither of Vitek Vanecek or Ilya Samsonov look close to good enough.

Shayna: Or maybe the Capitals saw that the year we all counted them out was the only one they won. There are just too many “ifs” before even getting to the net.

The nerds don't like Washington's top two centers and goaltending. Backstrom in particular remains the biggest issue on the roster IMO and if they are looking to make noise in the postseason this year they need to look for an upgrade.

 

[twabby]

Here’s some from part 2 that I’m pretty sure is just a smarter version of the problem I was having, suggesting the real peak is farther to the right:

Spoiler: I don't know why

View attachment 554115

And here’s the problem with creating a chimera player (besides Jason) and an explicit reminder that this is not expected growth:

Spoiler: they're all so big

View attachment 554116

Also worth noting: the individual graphs in part 1 suggest players retain their defensive impact for the majority of their career and crater offensively, which seems like the type of thing that bothers you less if you’re a Fehervary type and not, in fact, Erik Karlsson.

I'm not saying there's not uncertainty. There's plenty of uncertainty. This analysis is not conclusive, but it is a data point. As are the other estimates I posted earlier from TopDownHockey and the old WAR on ICE. These data points paint a picture that indicates that there isn't a ton of improvement from ages 22 on, in general. Could Fehervary defy these averages? Yes, of course. Not everything is controlled for.

But unless we've identified a reason why Fehervary specifically should be considered an outlier compared to his peers I'm not willing to project him to do much better than players who have profiled similarly to him in their age 22 season. COVID could be that reason. They need to honestly evaluate how much COVID impacted his season in particular.

Your last paragraph is also quite a bit concerning to me because most of Fehervary's positive impact last year was due to his offensive contributions. He was a tire fire defensively. Indeed, here his is breakdown:

So if he doesn't actually get better defensively, as the "average" player doesn't, then he's going to remain a below-replacement level defensive player for his career, while his modest offensive success tails off. I'm not fully convinced that this will happen exactly, but it is concerning that on average this happens.

 

[twabby]

it’s a shame D never improve right?

Orlov is a perfect example of a player who started out great and has remained consistent, with perhaps a slight decline over time (as the average defenseman does):

The bottom half of the graphic is his WAR/60, and notice how consistent it is (his 12-13 season was only 5 games, which explains the huge spike).

What has changed with Orlov has been his ice time.

 

[CapitalsCupReality]

No he’s gotten better….your math tricks don’t work on me.

 

[twabby]

No he’s gotten better….your math tricks don’t work on me.

Thank you for your contribution to the thread!

 

[HTFN]

I'm not saying there's not uncertainty. There's plenty of uncertainty. This analysis is not conclusive, but it is a data point. As are the other estimates I posted earlier from TopDownHockey and the old WAR on ICE. These data points paint a picture that indicates that there isn't a ton of improvement from ages 22 on, in general. Could Fehervary defy these averages? Yes, of course. Not everything is controlled for.

But unless we've identified a reason why Fehervary specifically should be considered an outlier compared to his peers I'm not willing to project him to do much better than players who have profiled similarly to him in their age 22 season. COVID could be that reason. They need to honestly evaluate how much COVID impacted his season in particular.

Your last paragraph is also quite a bit concerning to me because most of Fehervary's positive impact last year was due to his offensive contributions. He was a tire fire defensively. Indeed, here his is breakdown:

View attachment 554170

So if he doesn't actually get better defensively, as the "average" player doesn't, then he's going to remain a below-replacement level defensive player for his career, while his modest offensive success tails off. I'm not fully convinced that this will happen exactly, but it is concerning that on average this happens.

Got it. So you’re just going to backpedal and suddenly this is just a data point and not one of the dumbest conclusions you’ve ever drawn. You’re talking about a rookie defenseman who played top pairing minutes all season (would still love to see a first half/second half split), got COVID, and hit the rookie wall. I’m not personally worried because I don’t subscribe to this dumb ass theory anyway but you’d probably have to do more than take an average to understand a player’s season. This shit continually doesn’t work the way you think it works and I doubt this is an exception.

Won’t engage in anything about the game that anyone else is saying to you, won’t acknowledge any incendiary mistakes you’re making, you just want to roll in the mud and play with your numbers.

It doesn’t even sound like you read any of what I quoted you from your own source. It’s not just uncertainty, they have reason to believe the “true” prime is to the right of the established graph. Every single mention of the early prime age is tempered with “that doesn’t seem entirely right and we’re going to adjust for it later”. Explicit mentions that this is not expected growth for a single individual, but hey, you justify using it as a “data point” and making wild conclusions if you want.

And then only having data for 8 f***ing years and deciding you know how all players will grow…

Play away, it doesn’t matter.

 

[CapitalsCupReality]

Thank you for your contribution to the thread!

In my defense….you decided to move it here!

Please continue….

 

[twabby]

Got it. So you’re just going to backpedal and suddenly this is just a data point and not one of the dumbest conclusions you’ve ever drawn. You’re talking about a rookie defenseman who played top pairing minutes all season (would still love to see a first half/second half split), got COVID, and hit the rookie wall. I’m not personally worried because I don’t subscribe to this dumb ass theory anyway but you’d probably have to do more than take an average to understand a player’s season. This shit continually doesn’t work the way you think it works and I doubt this is an exception.

Won’t engage in anything about the game that anyone else is saying to you, won’t acknowledge any incendiary mistakes you’re making, you just want to roll in the mud and play with your numbers.

It doesn’t even sound like you read any of what I quoted you from your own source. It’s not just uncertainty, they have reason to believe the “true” prime is to the right of the established graph. Every single mention of the early prime age is tempered with “that doesn’t seem entirely right and we’re going to adjust for it later”. Explicit mentions that this is not expected growth for a single individual, but hey, you justify using it as a “data point” and making wild conclusions if you want.

And then only having data for 8 f***ing years and deciding you know how all players will grow…

Play away, it doesn’t matter.

The only conclusions I have drawn relate to what has happened in the past with Fehervary’s peers, and how players on average progress through their careers.

I have not concluded anything on Fehervary’s career arc. I am merely concerned that because he had a poor rookie season that this could portend a disappointing career based on what his peers have done in the past on average. He could be very good next year and it would not be a remarkable result. Flipping heads three times in a row isn’t anything special, for instance. I wouldn’t bet on it happening, but it’s not a rare outcome.

I feel like the line of argument you are taking is to assume players must develop and better better over time, because that’s what everyone has always believed. Under this assumption of course the analysis presented must be wrong.

But perhaps the assumption you’re making is wrong. Are you so out of touch? No, it’s the children who are wrong.

 

[Kalopsia]

Sorry for the delay.

I went ahead and did the following: I took every player-season from every defenseman from 2007-08 until 2021-22. For each season, I took the top 120 TOI player-seasons. This gave 1800 player-seasons.

Of these 1800 player-seasons, I further culled them by taking the first instance of a player under the age of 23. For instance, Aaron Ekblad has been in the top 120 TOI seasons since he entered the league, so I only considered his age 18 season (2014-15) to start.

This gave me about 109 players under the age of 23 who both got "top 4" minutes but also have more records in the database. For instance, Martin Fehervary has the former (an age 22 season with top 4 minutes) but does not yet have a subsequent season to compare to. So he's not a part of this dataset, which makes sense because we are trying to forecast his future based on last season. I took the WAR/60 of the first instance of a player under 23 getting top 4 minutes.

I then took the cumulative WAR/60 of these players following their initial top 4 season as a comparison. For instance, I took Aaron Ekblad's 2014-15 WAR/60 and plotted it against his 2015-16 thru 2021-22 cumulative WAR/60. I did this for each of the 109 players. Here is the scatterplot:

View attachment 553804

A few notes:

1. The correlation coefficient of this dataset is about r = .44, and r^2 = 0.19. This is actually a somewhat stronger correlation than I was expecting. The trendline is also shown in blue.

2. Martin Fehervary's WAR/60 in his first Top 4 year (e.g. this past year) was -0.016.

3. In the plot above, the red line represents the median top 4 player's subsequent WAR/60 (about 0.04).

Overall, I think this kind of further supports what I've been saying. Most players under the age of 23 who played a lot of minutes and had initial results similar to Fehervary never really progressed to becoming above average top 4 players. That's not to say they flame out of the league (though many did), but it's more to say most of them become expendable.

I think Washington needs to consider that Fehervary may not progress into much more than a third pairing talent for his career, especially if he doesn't improve next year. Next year will be key, and I think by the time they reach the TDL they should know much more about his projected future.

Appreciate the time and effort that went into this. Gotta concede that it's not bad players throwing the curve off.

I still think the other possible flaw I mentioned - that the WAR stat isn't accurately assessing defensive WAR because it's not adjusting enough for deployment - is a much more likely explanation for the weird defenseman aging curve than hockey defensemen entering the NHL as finished products. As HFTN mentioned, I think the separate aging curves for offensive and defensive WAR in the original article you posted support this. The offensive one is a pretty standard aging curve for both forwards and defensemen:

While the defensive one is this weird plateau from about 20 to 32:

Either defense in hockey is some bizarre innate talent that lasts long after age-related decline dulls other skills, both in hockey and in every other sport I've seen aging curves for, or maybe hockey analytics just aren't at the point yet where they can properly adjust for differences in difficulty of deployment, and coaches naturally give players harder roles as they improve and easier roles as they decline, leading to a smoothing of the aging curve? The hockey analytics community borrows a lot of concepts from baseball analytics, as this article demonstrates, but this is one of the things where there's really no precedent in baseball. Every first baseman on every team in baseball is asked to make the same plays in the field and face the same pitchers, there's no way for teams to shelter a guy if he's bad or give him tougher assignments if he's good, and the only "systems" in place are shifts or tailored pitching approaches that are easy to quantify. As a result, comparing a player to the replacement level at their position is pretty straightforward (although even in baseball, quantifying defense is the still a matter of serious debate and even stat nerds take defensive metrics with a huge grain of salt). In hockey there's a massive difference between what the 1LD and the 3LD on the same team are expected to do and who they face, and there can be a massive difference between the roles of 1LD's on different teams, and their roles might change within a single game based on the game situation or what the opposing team's doing. I'm sure Dawson Sprigings is brilliant, but he can be brilliant and still not get quantifying defensive value right on the first try. There's a lot of really smart guys working on the much easier task of quantifying defensive value in baseball, and they haven't gotten it nailed down yet, so it's probably gonna be a while before we nail it down for hockey.

No, the bad ones are not going to affect the curve unless you can show bad players get worse at different rates than good players get worse.

I also disagree that there is more volatility in bad vs. good players in general. Good players have higher highs and higher lows, but there is still the same volatility from season to season, in general.

The statistics vs. individual argument is kind of silly. Yes, of course every individual is different and unique. But trends exist. You can make informed predictions about how players will perform based on how players before them have done in similar situations.

Unless you have specific knowledge that indicates Fehervary is going to improve more than other players in similar situations have done in the past, then it just seems like wishful thinking rather than an educated guess to believe he is going to take a big step forward next season.

This paragraph frustrates me, because a lot of people have been saying since the start of this debate "he had covid and a massive increase in workload which made him fall off in the second half." He shouldn't be treated like he played at the same level all season. Look at the Smoothed 5v5 xG/60 in this chart:

A lot of black pre-covid, a lot of red afterwards. I know you hate narratives, but I think there's a pretty clear one here of a player who was coming into his own, then got covid, and the lingering effects of that plus conditioning issues stemming from being limited by injuries to only 24 games the year before caused him to lose a step. If post-covid Fehervary was all we saw no one would be questioning your assessment, but it seems like you're totally writing off what he did before covid and ignoring some pretty obvious explanations for his drop off in the second half.

 

[HTFN]

The only conclusions I have drawn relate to what has happened in the past with Fehervary’s peers, and how players on average progress through their careers.

I have not concluded anything on Fehervary’s career arc. I am merely concerned that because he had a poor rookie season that this could portend a disappointing career based on what his peers have done in the past on average. He could be very good next year and it would not be a remarkable result. Flipping heads three times in a row isn’t anything special, for instance. I wouldn’t bet on it happening, but it’s not a rare outcome.

I feel like the line of argument you are taking is to assume players must develop and better better over time, because that’s what everyone has always believed. Under this assumption of course the analysis presented must be wrong.

But perhaps the assumption you’re making is wrong. Are you so out of touch? No, it’s the children who are wrong.

Pretty haughty stance for someone who asserts NHL defensemen are fully formed on debut.

You’re wrong. Every single person here has told you you’re wrong. This isn’t even a debate, it’s just you plugging your ears while everyone else tries to help you, and just like every other time you’ve been wrong it’s going absolutely nowhere, and with the utmost arrogance