[Signing] Oilers sign Darnell Nurse to extension beginning in 2022 (8 yrs, $9.25 mil AAV)

[Boreal]

Originally Posted by Oil Stain

Here's the beef I have with XGA/60...

Lets take the last three years of Chris Tanev. I filtered out D-men who played less than 700 ES minutes btw.

Tanev's ranking in xGA/60:

20/21- 1st out of 129 D-men.
19/20- 137th out of 164
18/19- 72nd out 176

In the span of three seasons Tanev went from middle of the pack, to one of the worst, to the best in the league.

How do you take useful information about a player's ability out of something like that?

I mean clearly Tanev didn't play at a seventh defender ability in 19/20, and then become the best shutdown defenceman in the league in 20/21, so what is XGA/60 actually measuring?

Tanev played 19/20 with Quinn Hughes who has the defensive acumen of Tyson Barrie. It is reasonable to assume he was left to do a lot of heavy defensive lifting, which enabled Hughes to thrive.

That being said, the point of analytical information like this is description and somewhat predictive. Like any accurate Bayesian model it will tell you the most at the tails, who is really bad and who is really good. Year after year, we are what we consistently do.

In other words the numbers are an oversimplification to assess a particular part of a players game in context and expose cognitive biases a person may have on a certain part of a certain players game.

Even at 72nd out of 176 defenceman in 18/19 Tanev was statistically a top 4 dman on most teams defensively.

Nurse has not touched being a top 100 dman defensively and he’s being paid like his has, or will. Meanwhile the water carrying Edmonton media is using physicality, face punching and salary inflation as a justification.

If the model is egregiously wrong and he is this hidden stud within the Oiler brass why didn’t they lock him up years ago at a better price point?

Are they stupid or incompetent, or both?

In 3 years you can just add all this nonsense to the long list of absurdity that has made them a laughing stock.

[CorsiHockeyLeague]

The thing is, 19/20 Quinn Hughes appeared to be quite good defensively. He actually fared better defensively away from Tanev than with him. If it was just Tanev handling all the defensive responsibilities and covering for Hughes, you'd expect to see the opposite.

I dunno, the whole Tanev thing really does seem like a bit of a random outlier rather than a good example to refer to.

[memphusk]

Originally Posted by CorsiHockeyLeague

The thing is, 19/20 Quinn Hughes appeared to be quite good defensively. He actually fared better defensively away from Tanev than with him. If it was just Tanev handling all the defensive responsibilities and covering for Hughes, you'd expect to see the opposite.

I dunno, the whole Tanev thing really does seem like a bit of a random outlier rather than a good example to refer to.

He did? Hughes can carry more hard minutes than Tanev? Sounds wrong.

[Samonadreau]

Originally Posted by CorsiHockeyLeague

The thing is, 19/20 Quinn Hughes appeared to be quite good defensively. He actually fared better defensively away from Tanev than with him. If it was just Tanev handling all the defensive responsibilities and covering for Hughes, you'd expect to see the opposite.

I dunno, the whole Tanev thing really does seem like a bit of a random outlier rather than a good example to refer to.

Away from Tanev he probably had easier opponents.

[Jay Random]

Originally Posted by CorsiHockeyLeague

The thing is, 19/20 Quinn Hughes appeared to be quite good defensively. He actually fared better defensively away from Tanev than with him. If it was just Tanev handling all the defensive responsibilities and covering for Hughes, you'd expect to see the opposite.

I'd be interested to see data on Hughes's usage when not paired with Tanev. Could be that Tanev was (a) entrusted with the heavy defensive work when they were paired together, (b) facing the same calibre of opponents when not paired with Hughes (which would mean Hughes was spared the heaviest defensive duties anyway).

Because without Tanev this past year, Hughes seemed to be a defensive tire fire, and his stats, I believe, showed it.

[djsFlames]

Hughes is bad defensively.

Even worse than Darnell. So hey, the Oiler apologists have that to cling to.

Tanev is a unique case. I think we'll get more of an idea after two seasons here with the Flames as to what he is at this point. I'm hoping it's a continuation of last season of course.

[Bingo]

Originally Posted by Rising_Oil_Prices

I don't repute these numbers you have given, mostly because I don't know what they mean... but I do know a quality hockey player when I watch one. Nurse is that.

That's your reply?

Your homer eye test is the way we should all go over stats that prove he's a tire fire in his own zone?

I think Nurse has his qualities, don't get me wrong. He's big and can move. He's pretty good on the transition, gets pucks to the net.

But he's not a complete player as he's terrible defensively, and he's paid to be more than great in all aspects.

[Bingo]

Originally Posted by Oil Stain

Here's the beef I have with XGA/60...

Lets take the last three years of Chris Tanev. I filtered out D-men who played less than 700 ES minutes btw.

Tanev's ranking in xGA/60:

20/21- 1st out of 129 D-men.
19/20- 137th out of 164
18/19- 72nd out 176

In the span of three seasons Tanev went from middle of the pack, to one of the worst, to the best in the league.

How do you take useful information about a player's ability out of something like that?

I mean clearly Tanev didn't play at a seventh defender ability in 19/20, and then become the best shutdown defenceman in the league in 20/21, so what is XGA/60 actually measuring?

You don't have to like xGA ... honestly I think it's a stat that will continue to evolve (definitions, weightings etc changing)

But it's hard to look past basic shot metrics and counting. There isn't anything wrong with a "model" that says you'd like your defenseman not to be at the bottom of the pile for giving up shots on goal (or shot attempts, or scoring chances).

Who you play with is part of it.

I think some felt that Tanev had his wheels come off, but it could be that his numbers were hurt carrying Hughes. I guess this year will help solve some of that.

[Vinny01]

The Flames currently pay the top pair of Hanifin-Tanev a combined cap hit of 9.45M for the next 3 years a mere 200k more than Nurse will earn from his 27-35 aged seasons.

[CorsiHockeyLeague]

Originally Posted by Jay Random

I'd be interested to see data on Hughes's usage when not paired with Tanev. Could be that Tanev was (a) entrusted with the heavy defensive work when they were paired together, (b) facing the same calibre of opponents when not paired with Hughes (which would mean Hughes was spared the heaviest defensive duties anyway).

I can't remember where it was, probably the Athletic. But basically he had much better numbers at both ends of the ice with Tyler Myers, which makes no sense because Tyler Myers sucks at both ends of the ice. It was probably just a sample size issue.

Because without Tanev this past year, Hughes seemed to be a defensive tire fire, and his stats, I believe, showed it.

Oh boy did they ever.

https://twitter.com/user/status/1371957273237553152

... It's going to be really funny when they give him the same contract as Nurse.

[Gordies Elbow]

Originally Posted by Boreal

That being said, the point of analytical information like this is description and somewhat predictive. Like any accurate Bayesian model it will tell you the most at the tails, who is really bad and who is really good. Year after year, we are what we consistently do.

In other words the numbers are an oversimplification to assess a particular part of a players game in context and expose cognitive biases a person may have on a certain part of a certain players game.

The key point you make is around "any accurate Bayesian model" - I'm not sure that these are those.

In mathematics, you need to diagnose your model, evaluating the basic assumptions and comparing model approaches, and prepare the model for a proper defense.

The primary arguments when defending are around assumptions. Some of the assumptions added to these models are subjective (e.g. what constitutes a turnover, what is a proper zone entry.) Some make assumptions about specificity (e.g. shot location makes a shot harder or easier to save.) These may seem obvious, but are they statistically significant given the small numbers of minutes of performance?

In short, models can be descriptive (e.g. relative performance of a machine.) or predictive (e.g. economic forecast.) Descriptive models can be tested against current performance, predictive against future events. In order to be validated, they should be tested. Are any of these?

I (unfortunately) somewhat agree with Brian Burke, where he said “Analytics are like a lamppost to a drunk. They’re good for support, but not for illumination.”

[Bingo]

Originally Posted by Gordies Elbow

The key point you make is around "any accurate Bayesian model" - I'm not sure that these are those.

In mathematics, you need to diagnose your model, evaluating the basic assumptions and comparing model approaches, and prepare the model for a proper defense.

The primary arguments when defending are around assumptions. Some of the assumptions added to these models are subjective (e.g. what constitutes a turnover, what is a proper zone entry.) Some make assumptions about specificity (e.g. shot location makes a shot harder or easier to save.) These may seem obvious, but are they statistically significant given the small numbers of minutes of performance?

In short, models can be descriptive (e.g. relative performance of a machine.) or predictive (e.g. economic forecast.) Descriptive models can be tested against current performance, predictive against future events. In order to be validated, they should be tested. Are any of these?

I (unfortunately) somewhat agree with Brian Burke, where he said “Analytics are like a lamppost to a drunk. They’re good for support, but not for illumination.”

That's a cop out.

You don't have to get into any sort of determinative stats if you don't want to.

If a player is on the ice for some of the highest shot attempt and shot rates in the league can it ever be a good thing?

A guy like Tyson Barrie is paid $5M a season because he's elite at one thing, and not very good at another. Sure looks to me like Nurse is the same, but paid to be complete.

[Gordies Elbow]

Originally Posted by Bingo

That's a cop out.

If a player is on the ice for some of the highest shot attempt and shot rates in the league can it ever be a good thing?

A guy like Tyson Barrie is paid $5M a season because he's elite at one thing, and not very good at another. Sure looks to me like Nurse is the same, but paid to be complete.

It's not a cop out - it's adding mathematical rigor to the process.

As to the shot attempt rates, if the shots aren't converted to goals and by proxy, to wins, does it matter?

As to the complete player, baseball had guys that were focused on "five tool" players, and believed them when they saw them. "Moneyball"/Sabermetrics was about removing bias, specifically drilling down to specific skills (e.g. "he gets on base" from the movie.)

To argue your assumption - are "complete" players better than "incomplete" players in all cases?

[GioforPM]

Originally Posted by Gordies Elbow

To argue your assumption - are "complete" players better than "incomplete" players in all cases?

Let's put it this way - if you pay a Dman $9M/y you want him to be able to play heavy minutes, match up against the top opposition from the other team and prevent them from scoring, while adding some offence of their own.

If you can't put the guy against the top lines successfully, the minutes drop and you shouldn't pay as much. With Nurse, they seem to have tried putting him out there, but it wasn't very successful. And it's not going to get better without Larsson.

[Bingo]

Originally Posted by Gordies Elbow

It's not a cop out - it's adding mathematical rigor to the process.

As to the shot attempt rates, if the shots aren't converted to goals and by proxy, to wins, does it matter?

As to the complete player, baseball had guys that were focused on "five tool" players, and believed them when they saw them. "Moneyball"/Sabermetrics was about removing bias, specifically drilling down to specific skills (e.g. "he gets on base" from the movie.)

To argue your assumption - are "complete" players better than "incomplete" players in all cases?

Well he's 160th in actual goals against per 60 as well, which is a 5th defenseman.

No model there.

Just bad.

And yeah I'd assume most hockey people would rather make sure the 4th highest paid defenseman in hockey can do more than just skate around in the offensive zone. So probably good assumption that complete is better than incomplete, unless the disparity in pay cheque makes up for what is lacking.

[Gordies Elbow]

Originally Posted by Bingo

Well he's 160th in actual goals against per 60 as well, which is a 5th defenseman.

No model there.

Just bad.

And yeah I'd assume most hockey people would rather make sure the 4th highest paid defenseman in hockey can do more than just skate around in the offensive zone. So probably good assumption that complete is better than incomplete, unless the disparity in pay cheque makes up for what is lacking.

He's 4th in minutes played, 1st in 5v5 goals, and had 25% of voters list him in the top 5 defenders as voted on for the Norris trophy this year.

I get that this is CalgaryPuck, and E=NG here, but from what I've read here, do you really think that Nurse is a 5th defenseman on any team?

[Shazam]

Originally Posted by Gordies Elbow

The key point you make is around "any accurate Bayesian model" - I'm not sure that these are those.

In mathematics, you need to diagnose your model, evaluating the basic assumptions and comparing model approaches, and prepare the model for a proper defense.

The primary arguments when defending are around assumptions. Some of the assumptions added to these models are subjective (e.g. what constitutes a turnover, what is a proper zone entry.) Some make assumptions about specificity (e.g. shot location makes a shot harder or easier to save.) These may seem obvious, but are they statistically significant given the small numbers of minutes of performance?

In short, models can be descriptive (e.g. relative performance of a machine.) or predictive (e.g. economic forecast.) Descriptive models can be tested against current performance, predictive against future events. In order to be validated, they should be tested. Are any of these?

I (unfortunately) somewhat agree with Brian Burke, where he said “Analytics are like a lamppost to a drunk. They’re good for support, but not for illumination.”

These advanced stats aren’t models. There’s no predictive analysis. People could even take past data and model it forward and match it against past real data but I’ve never seen that done.

I’ve seen some terribly naive analysis done (mostly by the Nation) where they attempt to find things out like when players start to decline but it was done all wrong.

[GioforPM]

Originally Posted by Gordies Elbow

He's 4th in minutes played, 1st in 5v5 goals, and had 25% of voters list him in the top 5 defenders as voted on for the Norris trophy this year.

I get that this is CalgaryPuck, and E=NG here, but from what I've read here, do you really think that Nurse is a 5th defenseman on any team?

Minutes played really depends on what team you're on. We all know he had goals. And Norris voters are a pretty mixed bag who overemphasize points because they can't or don't look at anything else. One dude put him ahead of Makar, which is silly. Another was Eric Francis, who put him second (LOL), another was Mark Spector, etc.

Nurse got about the same amount of people voting for him as Mackenzie Weegar, who just got a contract the year before. Not $9M though LOL.

[edslunch]

Just a note for the advanced stat gurus here - you lose people when you jump straight to xGA/60 etc when rating a player. The only thing that matters on ice is actual results, like GA/60, so start there. That’s the fact. Now dig deeper with advanced stats to evaluate the actual number - do they suggest that performance is representative and sustainable or not.

Remember that stats are useful for supporting the drunk but the drunk is the only thing that is real.

[Bingo]

Originally Posted by Gordies Elbow

He's 4th in minutes played, 1st in 5v5 goals, and had 25% of voters list him in the top 5 defenders as voted on for the Norris trophy this year.

I get that this is CalgaryPuck, and E=NG here, but from what I've read here, do you really think that Nurse is a 5th defenseman on any team?

That's not what I said or am saying.

He's not a good defenseman in his own zone (5th defenseman level). He's elite offensively.

that as a mix gets you a solid hockey player, but not the guy you pay to be one of the best defensemen in the league.

Plenty of non Calgarypuck sources out there pointing out the same thing.