monitoring_string = "358c248ada348a047a4b9bb27a146148"
2019/20 Roster Thread XXIX | Page 37 | HFBoards - NHL Message Board and Forum for National Hockey League
  • Xenforo Cloud upgraded our forum to XenForo version 2.3.4. This update has created styling issues to our current templates, this is just a temporary look. We will continue to work on clearing up these issues for the next few days and restore the site to it's more familiar look, but please report any other issues you may experience so we can look into. Thanks for your patience and understanding.

2019/20 Roster Thread XXIX

Status
Not open for further replies.
JojoTheWhale said:
Different sites use different xGF models. You will find some variation.
Click to expand...
3% swing seems like a lot.

Interesting, every Dman has a higher % on NST vs Corsica......except Hagg

So depending of on what model/site you go by you can build different cases for different people. Maybe the eye test matter too.
 
Beef Invictus said:
We'd have to replace him with someone not in our system. Which brings us back to why it makes no sense to move him now.
Click to expand...

This year yes.
Next year maybe, we'll see how York, Kalynuk, Hogberg, Zamula progress (and if they're signed).

Point is right now the Ghost threshold isn't unreachable, it's higher than the Hagg threshold, but it's not top 4 level.

For you "rel" junkies:
xGF%rel:
Provorov +5.27
Niskanen +5.21
Braun +0.94
Myers +0.32
Sanheim -0.02
Ghost -5.17
Hagg -10.98

xGArel
Provorov -0.026
Braun -0.024
Niskanen -0.017
Myers -0.002
Sanheim +0.016
Ghost +0.018
Hagg +.047

Best xGArel for forwards:
NAK -0.041
JVR -0.033 [bet ya didn't think I could rock n roll]
Pitlick -0.022
Couts -0.016
Voracek -0.015

Worst xGArel
Stewart +0.080
Andreoff +0.056
Frost +0.033 [and this is why he got sheltered minutes]
Twarynski +0.026
Laughton +0.019
Raffl +0.017
Fourth line guys seem to have bad metrics, is it b/c they're bad or they get the worst matchups?
 
Last edited:
Tripod said:
3% swing seems like a lot.

Interesting, every Dman has a higher % on NST vs Corsica......except Hagg

So depending of on what model/site you go by you can build different cases for different people. Maybe the eye test matter too.
Click to expand...

It is a lot!

Everything matters. As the axiom goes, all models are wrong, but some are useful. This is why you need to look at a wide variety of sources of information (numerical, eye, and everything between) and draw your own conclusions.

This is also why I harp on degrees of certainty whenever related topics come up. If you have a player at +2.5% xGFRel and one at +3.2%, it's irresponsible to say that xGF% makes one a clear upgrade. Likewise you should never take out a 40 game sample size as irrefutable proof of anything, especially if we have hundreds of other games of priors on the players in question.

It's fairly rare to look at a player from every angle and be utterly convinced he's tremendous or useless. Hagg does some things well, but there is absolutely no argument he has leveraged those skills well or often enough to be an NHL player. None.
 
deadhead said:
This year yes.
Next year maybe, we'll see how York, Kalynuk, Hogberg, Zamula progress (and if they're signed).

Point is right now the Ghost threshold isn't unreachable, it's higher than the Hagg threshold, but it's not top 4 level.

For you "rel" junkies:
xGFrel:
Provorov +5.27
Niskanen +5.21
Braun +0.94
Myers +0.32
Sanheim -0.02
Ghost -5.17
Hagg -10.98

xGArel
Provorov -0.026
Braun -0.024
Niskanen -0.017
Myers -0.002
Sanheim +0.016
Ghost +0.018
Hagg +.047

Best xGArel for forwards:
NAK -0.041
JVR -0.033 [bet ya didn't think I could rock n roll]
Pitlick -0.022
Couts -0.016
Voracek -0.015

Worst xGArel
Stewart +0.080
Andreoff +0.056
Frost +0.033 [and this is why he got sheltered minutes]
Twarynski +0.026
Laughton +0.019
Raffl +0.017
Fourth line guys seem to have bad metrics, is it b/c they're bad or they get the worst matchups?
Click to expand...

Don't forget to account for how much Ghost's usage and linemates will be depressing his production. Anyone who gets that usage as his replacement will see a similar drop. Looking at raw numbers and forgetting to fill in the "why" is a way to fall into a trap. Ghost spends most of his time trying to pass to, or be set up by, those worst players. Plugging a new dman into that situation isn't necessarily going to net equal or better results than Ghost.

Don't forget the lesson TK taught you.
 
Striiker said:
For added contextual fun...

Ghost without Hagg: 51.48 xGF%

Hagg without Ghost: 37.36 xGF%

TEAMMATES AND USAGE MATTER A LOT, BOYS AND GIRLS.

Ghost is a hard carry.
Hagg is a deadweight anchor.
Click to expand...
What are EVERY Dman's numbers with and without Hagg.....and with and without Ghost?
 
TLDR:

tumblr_mvn95f60bb1rzoch0o4_r1_250.gif


giphy.gif
 
Beef Invictus said:
Don't forget to account for how much Ghost's usage and linemates will be depressing his production. Anyone who gets that usage as his replacement will see a similar drop. Looking at raw numbers and forgetting to fill in the "why" is a way to fall into a trap. Ghost spends most of his time trying to pass to, or be set up by, those worst players. Plugging a new dman into that situation isn't necessarily going to net equal or better results than Ghost.

Don't forget the lesson TK taught you.
Click to expand...
Ghost, TK, Sanheim, and Sean F***ing Couturier (more than anyone), teach that same lesson.
 
JojoTheWhale said:
It is a lot!

Everything matters. As the axiom goes, all models are wrong, but some are useful. This is why you need to look at a wide variety of sources of information (numerical, eye, and everything between) and draw your own conclusions.

This is also why I harp on degrees of certainty whenever related topics come up. If you have a player at +2.5% xGFRel and one at +3.2%, it's irresponsible to say that xGF% makes one a clear upgrade. Likewise you should never take out a 40 game sample size as irrefutable proof of anything, especially if we have hundreds of other games of priors on the players in question.

It's fairly rare to look at a player from every angle and be utterly convinced he's tremendous or useless. Hagg does some things well, but there is absolutely no argument he has leveraged those skills well or often enough to be an NHL player. None.
Click to expand...
Yeah I admit I have been learning more about the "fancy stats" more this year than before and will always default to the more educated....and of course people I trust.

Appreciate the info!
 
Curufinwe said:
Not sure how Hart's development was sacrificed. If he had struggled in the NHL he would have gone back down again. But he didn't struggle, he thrived, and got plenty of game time.
Click to expand...

Yeah I'm not sure where people are getting this from

For those who think Hart is struggling so bad or has had his development stunted.

Which 21 year old goalie would you rather have

2.83 GAA .905 S% 1 SO

2.61 GAA .905 S% 1 SO

Hint. One is Hart the other is 21 year old Carey Price.
 
EDIT: Ignore this chart. Too much missing information.
I addressed that here:
2019/20 Roster Thread XXIX
_____________________________________________
Tripod said:
What are EVERY Dman's numbers with and without Hagg.....and with and without Ghost?
Click to expand...
Of course this doesn't account for other things like forward teammates or usage but...

***Quick note: it doesn't count games where the player was hurt or healthy scratched. In other words, obviously Provorov has played more than 412:37 at 5v5 away from Hagg this year. It's just that he's played 412:37 when Hagg was dressed but not on the ice. So the "without" numbers are kinda misleading***

-----ALL NUMBERS AT 5v5------
Purple text = small sample (less than 100 minutes)
Green text = positive
Red text = negative

[TABLE="class: brtb_item_table"][TBODY][TR][TD]Provorov with Hagg[/TD][TD]22:19[/TD][TD]65.12 CF%[/TD][TD]73.73 xGF%[/TD][/TR]
[TR][TD]Provorov without Hagg[/TD][TD]412:37[/TD][TD]52.23 CF%[/TD][TD]53.43 xGF%[/TD][/TR]
[TR][TD]Provorov with Ghost[/TD][TD]30:19[/TD][TD]53.85 CF%[/TD][TD]38.60 xGF%[/TD][/TR]
[TR][TD]Provorov without Ghost[/TD][TD]558:57[/TD][TD]51.22 CF%[/TD][TD]53.62 xGF%[/TD][/TR]
[TR][TD][/TD]
[TD][/TD]
[TD][/TD]
[TD][/TD][/TR]
[TR][TD]Sanheim with Hagg[/TD][TD]31:12[/TD][TD]37.70 CF%[/TD][TD]25.31 xGF%[/TD][/TR]
[TR][TD]Sanheim without Hagg[/TD][TD]412:37[/TD][TD]55.73 CF%[/TD][TD]57.16 xGF%[/TD][/TR]
[TR][TD]Sanheim with Ghost[/TD][TD]35:30[/TD][TD]50.00 CF%[/TD][TD]39.01 xGF%[/TD][/TR]
[TR][TD]Sanhieim without Ghost[/TD][TD]553:46[/TD][TD]51.09 CF%[/TD][TD]49.90 xGF%[/TD][/TR]
[TR][TD][/TD]
[TD][/TD]
[TD][/TD]
[TD][/TD][/TR]
[TR][TD]Niskanen with Hagg[/TD][TD]15:34[/TD][TD]40.74 CF%[/TD][TD]57.62 xGF%[/TD][/TR]
[TR][TD]Niskanen without Hagg[/TD][TD]407:10[/TD][TD]51.37 CF%[/TD][TD]54.90 xGF%[/TD][/TR]
[TR][TD]Niskanen with Ghost[/TD][TD]31:00[/TD][TD]60.00 CF%[/TD][TD]67.32 CF%[/TD][/TR]
[TR][TD]Niskanen without Ghost[/TD][TD]545:46[/TD][TD]50.23 CF%[/TD][TD]52.94 xGF%[/TD][/TR]
[TR][TD][/TD]
[TD][/TD]
[TD][/TD]
[TD][/TD][/TR]
[TR][TD]Braun with Hagg[/TD][TD]53:07[/TD][TD]28.57 CF%[/TD][TD]27.76 xGF%[/TD][/TR]
[TR][TD]Braun without Hagg[/TD][TD]297:55[/TD][TD]58.17 CF%[/TD][TD]57.71 xGF%[/TD][/TR]
[TR][TD]Braun with Ghost[/TD][TD]161:06[/TD][TD]47.86 CF%[/TD][TD]46.18 xGF%[/TD][/TR]
[TR][TD]Braun without Ghost[/TD][TD]413:19[/TD][TD]55.02 CF%[/TD][TD]52.71 xGF%[/TD][/TR]
[TR][TD][/TD]
[TD][/TD]
[TD][/TD]
[TD][/TD][/TR]
[TR][TD]Myers with Hagg[/TD][TD]48:19[/TD][TD]44.09 CF%[/TD][TD]49.04 xGF%[/TD][/TR]
[TR][TD]Myers without Hagg[/TD][TD]137:33[/TD][TD]54.49 CF%[/TD][TD]47.31 xGF%[/TD][/TR]
[TR][TD]Myers with Ghost[/TD][TD]110:25[/TD][TD]58.24 CF%[/TD][TD]58.74 xGF%[/TD][/TR]
[TR][TD]Myers without Ghost[/TD][TD]206:46[/TD][TD]45.67 CF%[/TD][TD]43.90 xGF%[/TD][/TR][/TBODY][/TABLE]
 
Last edited:
Beef Invictus said:
Ah, my bad. I thought you used a different stat.

Now did you remember to use a wide variety of stats and information to build a complete picture or are you still doing the thing where you cherry pick a few things without any context at all?

Why have you ignored the isolated impact chart I posted which combines metrics (not just the corsi you claim to despise except when it happens to suit your needs) and presents them in a useful visual format?
Click to expand...
More projecting. You accuse me of misusing stats. You don’t even cite the correct stat I used. You accuse me of ignoring something you post. You repeatedly ignore things I post & simple questions I ask. Not sure why I even bother responding to you in earnest.

Your “isolated impact chart” doesn’t show that Ghost has been a plus defender. “Average” I think is the description.

It doesn’t show that Ghost has been better in his defensive zone than Hagg. And it seems mostly based on CF%. I don’t think anyone would rate CF% as a better judge of defensive play than is xGF% combined with x+/- & ZSR, even with all of such stats being flawed.

You have yet to show stats that show Ghost to be above average defensively while hanging your hat on his defensive impact.

The positive impact of Getzlaf or Carter would more than make up for the difference between Ghost & Friedman. IMO.
 
Ghosts Beer said:
More projecting. You accuse me of misusing stats. You don’t even cite the correct stat I used. You accuse me of ignoring something you post. You repeatedly ignore things I post & simple questions I ask. Not sure why I even bother responding to you in earnest.

Your “isolated impact chart” doesn’t show that Ghost has been a plus defender. “Average” I think is the description.

It doesn’t show that Ghost has been better in his defensive zone than Hagg. And it seems mostly based on CF%. I don’t think anyone would rate CF% as a better judge of defensive play than is xGF% combined with x+/- & ZSR, even with all of such stats being flawed.

You have yet to show stats that show Ghost to be above average defensively while hanging your hat on his defensive impact.

The positive impact of Getzlaf or Carter would more than make up for the difference between Ghost & Friedman. IMO.
Click to expand...

What about Ghost and Hagg?
 
Beef Invictus said:
Don't forget to account for how much Ghost's usage and linemates will be depressing his production. Anyone who gets that usage as his replacement will see a similar drop. Looking at raw numbers and forgetting to fill in the "why" is a way to fall into a trap. Ghost spends most of his time trying to pass to, or be set up by, those worst players. Plugging a new dman into that situation isn't necessarily going to net equal or better results than Ghost.

Don't forget the lesson TK taught you.
Click to expand...

"rel" numbers partially take that into account, that's why they're a check on absolute metrics.
 
Ghosts Beer said:
More projecting. You accuse me of misusing stats. You don’t even cite the correct stat I used. You accuse me of ignoring something you post. You repeatedly ignore things I post & simple questions I ask. Not sure why I even bother responding to you in earnest.

Your “isolated impact chart” doesn’t show that Ghost has been a plus defender. “Average” I think is the description.

It doesn’t show that Ghost has been better in his defensive zone than Hagg. And it seems mostly based on CF%. I don’t think anyone would rate CF% as a better judge of defensive play than is xGF% combined with x+/- & ZSR, even with all of such stats being flawed.

You have yet to show stats that show Ghost to be above average defensively while hanging your hat on his defensive impact.

The positive impact of Getzlaf or Carter would more than make up for the difference between Ghost & Friedman. IMO.
Click to expand...

I didn't say he was positive. I was refuting your claim that he is a net negative.

Nothing changes the fact that you are intently, zealously yelling about a single statistic and another flawed metric while ignoring all other context. All this vitriol you're pointing at me for observing you're using flawed methodology doesn't change that.
 
JojoTheWhale said:
It is a lot!

Everything matters. As the axiom goes, all models are wrong, but some are useful.
Click to expand...

That's a paraphrase from George E.P. Box, one of the leading time series scholars.

All models are wrong - Wikipedia

It's one of my favorite aphorisms, which I used in my last book.

In both economics and statistics, you're working with incomplete data, and left out variable error (which is widely ignored b/c it's so inconvenient) due to the inability to specify and measure all pertinent variables, not to mention for economists, theory which is based on empirically invalidated assumptions.

Most people who use the phrase these days fail to credit Box.

Which is a shame, because his point is vital, too many researchers suffer from "significance" hubris, my model is significant > 95%, therefore I've accurately modeled said phenomena. Once you understand the model is flawed, then you understand that significance tests are also flawed, and should be seen as a guide, not a measure. Robustness and replication are needed to have confidence in a model (not to mention that the pressure to publish has encouraged wide spread data mining - Leamer, "Let's Take the Con Out of Econometrics" American Economic Review 1983, is still highly relevant).
 
Beef Invictus said:
I didn't say he was positive. I was refuting your claim that he is a net negative.

Nothing changes the fact that you are intently, zealously yelling about a single statistic and another flawed metric while ignoring all other context. All this vitriol you're pointing at me for observing you're using flawed methodology doesn't change that.
Click to expand...

A visual representation of his argument...

source.gif


He's gotta be close to China at this point...
 
Two third pair defensemen. One is getting $4.5M per and one is getting $1.15. With similar stats, which one is the best bargain for the organization?
 
Status
Not open for further replies.

Latest posts

Ad

Upcoming events

Ad

Top
-->->