Thoughts on Crosby's completely cherry-picked best stretch of hockey

[WarriorofTime]

but it really about recognizing Crosby's level of play over 4/5seasons is similar to McDavid's peak

I'm not really sure how this proves that. If we want to look at adjusted scoring, we can look at adjusted scoring, which is probably the easiest way to judge a thing like that across era. For McDavid to have the same % lead in PPG over whatever stretch of games, that's just more dependent on the Ovechkin and Malkin of his era having really down seasons in the midst of what happens to be McDavid's best stretch, which hasn't happened. The longer stretch due to Crosby's missed games in his own personal peak era leads to normal problems as well when you use a larger sample favorable to a specific player where you have older players from the beginning of that stretch aging out and younger players from the end of the stretch that weren't there yet at the start.

But while we can contextualize a player's performance relative to the League scoring, I don't see extra utility that comes from looking at how the 2nd/3rd best scorers of the era had down seasons (even relevant to era) as a way to boost up someone comparing across eras. I'm having a tough time seeing the value that comes ahead of just normalizing scoring rates for era.

 

[MadLuke]

that's just more dependent on the Ovechkin and Malkin of his era having really down seasons

Not sure how do we judge Malkin-Stamkos having a down time during Crosby dominance, Malkin won that impressive Ross-Hart in that stretch, stamkos had his 60 goal season, etc... if I do not mistake the time lines.

And 1 player going from 1.2 to 1.3 ppg would not move the average of the top 9 that much.

 

[daver]

I'm not really sure how this proves that. If we want to look at adjusted scoring, we can look at adjusted scoring, which is probably the easiest way to judge a thing like that across era. For McDavid to have the same % lead in PPG over whatever stretch of games, that's just more dependent on the Ovechkin and Malkin of his era having really down seasons in the midst of what happens to be McDavid's best stretch, which hasn't happened. The longer stretch due to Crosby's missed games in his own personal peak era leads to normal problems as well when you use a larger sample favorable to a specific player where you have older players from the beginning of that stretch aging out and younger players from the end of the stretch that weren't there yet at the start.

But while we can contextualize a player's performance relative to the League scoring, I don't see extra utility that comes from looking at how the 2nd/3rd best scorers of the era had down seasons (even relevant to era) as a way to boost up someone comparing across eras. I'm having a tough time seeing the value that comes ahead of just normalizing scoring rates for era.

The larger sample size show that one or two players having peak seasons or off seasons is statistically insignificant. What this is supposed to show is a reasonable reflection of each player's dominance vs. their direct peers given Crosby's lack of full seasons at his peak. Crosby was as dominant at his peak, relative to the league, so it shouldn't be a surprise that the numbers show a very similar per game domination.

Are you surprised by this?

 

[WarriorofTime]

What this is supposed to show is a reasonable reflection of each player's dominance vs. their direct peers given Crosby's lack of full seasons at his peak.

Why not just look at league adjusted scoring to make that point better than try to get there in a roundabout way?

Anyways, Crosby, Ovechkin and Malkin are all generally contemporaries, players born between September 17, 1985 and August 7, 1987, a 1 year, 10 month, 21 day window.

You could also take each of their best stretches of career PPG over whatever cherrypicked sample works best for the particular player (which is a bit harder for Malkin as he has a career 37 in 43 season in between his best seasons), do a small bit of league era adjusting, and get a more "clear" picture of Crosby's dominance over his most clear peers (and do the same for whatever other players from the Crosby era have been up there if trying to add a bigger group to compare).

If you're using very specific time periods to make your point, because it's good for the player in question but a period that is not so good for his peers, that is further deliberately skewing an already skewed perspective. I think it's always been clear and acknowledged that Crosby 2010-11 through 2013-14 had a clear and massive lead on the field. The fact that he is only 9th in total points during that peak era is a big shame given the injuries from that time (it's also basically the same as the mega cherrypicked sample, it just moves Crosby from 1.50 to 1.47 and Malkin from 1.19 to 1.20, which only slightly narrows the gap)

 

[MadLuke]

Why not just look at league adjusted scoring to make that point better than try to get there in a roundabout way?

This has been talked a lot if you mean the hockey reference ways, comparing to elite peer production instead of league wide scoring take scoring distribution into account.

Would it be minutes of powerplays, introducing 3v3 overtimes, there factor that can change how much top first liner score points other than league wide scoring average.

 

[WarriorofTime]

Would it be minutes of powerplays, introducing 3v3 overtimes, there factor that can change how much top first liner score points other than league wide scoring average.

Those all get factored into league-wide production, it's pure hand-wringing at that point to judge the effects of how it benefits top line vs. everyone else (even harder given how top lines get shifted over a course of a season and harder still over longer stretches where players role changes throughout a career), and you shouldn't let perfect get in the way of... very close to perfect.

HYPER increased powerplay seasons where 5 on 5 doesn't increase that much is probably the most problematic in this regard if 5 on 5 scoring isn't moving much, but this is still very hand-wringing as the seasons with huge jumps in PPO do get reflected in league scoring average, and it'd require a bit more tools to try and really isolate out the effects there and aren't worth messing up the data to try and be more perfect and end up creating a negative effect on the intended analysis.

 

[MadLuke]

Those all get factored into league-wide production,

Scoring distribution is not at all factored into league-wide production. The talk being all about scoring distribution here.

it's pure hand-wringing at that point to judge the effects of how it benefits top line vs. everyone else

That would turn this thread into Howiez Morenz having the best offensive season of all time debate all over again :

Did Howie Morenz have the best offensive single season ever?

Adjusted Points 1. Howie Morenz* 190 1927-28 2. Wayne Gretzky* 170 1985-86 3. Cooney Weiland* 168 1929-30 4. Wayne Gretzky* 166 1984-85 5. Mario Lemieux* 165 1988-89 6. Wayne Gretzky* 163 1983-84 7. Frank Boucher* 160 1929-30 8. Wayne Gretzky* 159 1982-83 What do you think?

forums.hfboards.com

Using only roster size and league scoring average you end up with this:

NHL & WHA Single Season Leaders and Records for Adjusted Points | Hockey-Reference.com

1. Howie Morenz (190), 2. Wayne Gretzky (170), 3. Cooney Weiland (168), 4. Wayne Gretzky (166), 5. Mario Lemieux (165), 6. Wayne Gretzky (163), 7. Frank Boucher (160), 8. Wayne Gretzky (159), 9. Connor McDavid (158), 10. Wayne Gretzky (156)

www.hockey-reference.com

Giant proportion of the best season being from 26 to 30 (35% of the Top 20), 17 of the top 37 from 4 seasons, trying to judge what first pp unit players scoring look like instead of league wide is probably worth it.

. very close to perfect.

Do you really think that ?

Rank	Player	PTS/A	Season
1.	Howie Morenz*	190	1927-28
2.	Wayne Gretzky*	170	1985-86
3.	Cooney Weiland*	168	1929-30
4.	Wayne Gretzky*	166	1984-85
5.	Mario Lemieux*	165	1988-89

Is very close to perfect ?

 

[WarriorofTime]

Do you really think that ?

Rank	Player	PTS/A	Season
1.	Howie Morenz*	190	1927-28
2.	Wayne Gretzky*	170	1985-86
3.	Cooney Weiland*	168	1929-30
4.	Wayne Gretzky*	166	1984-85
5.	Mario Lemieux*	165	1988-89

Is very close to perfect ?

Yes, absolutely, once you factor out that trying to gauge anything Pre World War II is basically an impossible direct comparison. Seriously, just start at 1945 and tell me what doesn't pass the "smell test" from your perspective? It's like Gretzky Gretzky Gretzky Gretzky Gretzky Gretzky with a couple peak Lemieux and a couple of peak McDavid in the midst there also, perhaps there could have been a peak Crosby if he managed a full season there. Again filter out the pre-100 year ago players, keep going and find someone who looks "weird"... Jagr, Esposito, Howe, Orr... you gotta go quite a ways...

 

[MadLuke]

Yes, absolutely

You really think

This:
Points

1.	Howie Morenz* • MTL	51
2.	Aurele Joliat* • MTL	39
3.	Frank Boucher* • NYR	35
	George Hay* • DTC	35
5.	Nels Stewart* • MTM	34

Should adjust by a large amount higher than this:
Points

1.	Wayne Gretzky* • EDM	205
2.	Paul Coffey* • EDM	126
3.	Michel Goulet* • QUE	122
4.	Peter Šťastný* • QUE	119
5.	Mike Bossy* • NYI	118

I will go ahead and say that I do not believe you, you either did not really check before answering or getting in I am debating on the Internet mode.

This has been quite talked about in this history section of the website the shortcomings of the hockey reference technic that make a lot of sense in baseball but not at all in hockey.

Seriously, just start at 1945 and tell me what doesn't pass the "smell test" from your perspective? I

The 06 is clearly hurt (having less player on the roster which often was playing with 4D instead of 6 should not hurt how much first liner forward star that much), 92-93 and 95-96 overrated like most very high PPO seasons. Before 1945 just make the exact same problem between any season of that technic more obvious, but they are always there in the very same ways, just of a lesser magnitude.

 

[jigglysquishy]

Yes, absolutely, once you factor out that trying to gauge anything Pre World War II is basically an impossible direct comparison. Seriously, just start at 1945 and tell me what doesn't pass the "smell test" from your perspective? It's like Gretzky Gretzky Gretzky Gretzky Gretzky Gretzky with a couple peak Lemieux and a couple of peak McDavid in the midst there also, perhaps there could have been a peak Crosby if he managed a full season there. Again filter out the pre-100 year ago players, keep going and find someone who looks "weird"... Jagr, Esposito, Howe, Orr... you gotta go quite a ways...

The entirety of 1942-1967 doesn't pass the smell test.

Howe, Hull, Beliveau, all look bizarre using the Hockey Reference method.

 

[WarriorofTime]

You really think

This:
Points

1.	Howie Morenz* • MTL	51
2.	Aurele Joliat* • MTL	39
3.	Frank Boucher* • NYR	35
	George Hay* • DTC	35
5.	Nels Stewart* • MTM	34

Should adjust by a large amount higher than this:
Points

1.	Wayne Gretzky* • EDM	205
2.	Paul Coffey* • EDM	126
3.	Michel Goulet* • QUE	122
4.	Peter Šťastný* • QUE	119
5.	Mike Bossy* • NYI	118

I will go ahead and say that I do not believe you, you either did not really check before answering or getting in I am debating on the Internet mode.

You cut off my entire post about starting post World War II? Entirely different leagues.

 

[MadLuke]

You cut off my entire post about starting post World War II? Entirely different leagues.

I think we both agree that you did not find that

Rank	Player	PTS/A	Season
1.	Howie Morenz*	190	1927-28
2.	Wayne Gretzky*	170	1985-86
3.	Cooney Weiland*	168	1929-30
4.	Wayne Gretzky*	166	1984-85
5.	Mario Lemieux*	165	1988-89

was very close to perfect then, and yet answered Yes absolutely close to perfect to the question. (either not checking before answering or because you are in I am debating on the Internet mode)

 

[WarriorofTime]

The entirety of 1942-1967 doesn't pass the smell test.

Howe, Hull, Beliveau, all look bizarre using the Hockey Reference method.

How so? Howe's 4 straight Art Rosses average out to 102 points per an 82 game season. Adjusting helps him immensely.

I think we both agree that you did not find that

Rank	Player	PTS/A	Season
1.	Howie Morenz*	190	1927-28
2.	Wayne Gretzky*	170	1985-86
3.	Cooney Weiland*	168	1929-30
4.	Wayne Gretzky*	166	1984-85
5.	Mario Lemieux*	165	1988-89

was very close to perfect then, and yet answered Yes absolutely close to perfect to the question. (either not checking before answering or because you are in I am debating on the Internet mode)

OK? Close to perfect starting with Post World War II, not great for an entirely different and not very similar Great Depression era league. Let's not start getting mega pedantic for the sake of an internet argument. There's very little issue with league adjusted scoring for any of the players actually being discussed - i.e., Gretzky, Lemieux, Jagr, Crosby, McDavid... regardless of whether it's not particularly effective when discussing Cooney Weiland.

It's not like "well if it doesn't work for a Howie Morenz and Wayne Gretzky discussion, it doesn't work for a Crosby and McDavid discussion, end of story"

My only issue is that 2021-22 was entirely divisional so you really had 4 different leagues running simultaneously while the player pool was allocated across all four of those... so it's admittedly funky.

 

[MadLuke]

How so? Howe's 4 straight Art Rosses average out to 102 points per an 82 game season. Adjusting helps him immensely.

102 per full 82 games is lower than Glen Murray in 2002-2003 or Kovalev in 2001.

The way Hockey reference use roster size seems in a very obvious way hurt Howe not help him.

Why do you say that it helps him, how ? he goes from one of the most dominant Art Ross winners to people missing the top 5 in season top talent were quite injured level.

 

[WarriorofTime]

102 per full 82 games is lower than Glen Murray in 2002-2003 or Kovalev in 2001.

The way Hockey reference use roster size seems in a very obvious way hurt Howe not help him.

Why do you say that it helps him, how ? he goes from one of the most dominant Art Ross winners to people missing the top 5 in season top talent were quite injured level.

Because without era adjusting, Howe is a one time 100 point player that peaked at age 40 and your grandpa is wrong for thinking there was anything special about him, which is obviously laughably false.

How does Howe's 1952-53 Art Ross become less dominant? It's one of the best era-adjusted seasons ever (131 on HockeyReference) and nobody else that season is anywhere in the vicinity.

 

[jigglysquishy]

How so? Howe's 4 straight Art Rosses average out to 102 points per an 82 game season. Adjusting helps him immensely.


OK? Close to perfect starting with Post World War II, not great for an entirely different and not very similar Great Depression era league.

Gordie Howe's peak is recognized as the highest offensive peak outside the rest of the Big Four. But if you take his four years and stick it into the 2022-23 season along with the 2nd and 3rd scorer those years you get

Rank	Player	Points
1	McDavid	146
	Howe 52-53	131
2	Draisaitl	122
	Howe 50-51	118
	Howe 53-54	114
3	Pastrnak	108
	Kucherov	108
5	MacKinnon	105
	Howe 51-52	105
6	Robertson	104
7	Tkachuk	103
8	Rantanen	100
9	Nugent-Hopkins	98
	Lindsay 52-53	98
10	Pettersson	97
11	Karlsson	96
12	Hughes	94
13	Marner	94
	Richard 53-54	93
14	Point	91
15	Crosby	89
	Thompson	89
17	Panarin	88
	Richard 50-51	88
	Linsday 53-54	87
18	Zibanejad	86
	Bentley 50-51	86
19	Stuzle	85
	Lindsay 51-52	85
	Abel 50-51	84
	Schmidt 50-51	84
	Delvecchio 52-53	83
	Richard 52-53	83
20	Nylander	82
	Lach 51-52	82

20 players from 2022-23 scored 82 adjusted points or more. Only 15 players scored 82 adjusted points or more across 1950-51 through 1953-54. This doesn't account for league size, but shows the massive undervaluing of 1950s hockey.

It has 3 of Howe's 4 years as "part of the pack" with off-peak years Pastrnak, 11-games missed MacKinnon, and off-peak Kucherov.

Does anyone really think that 3 players last year (McDavid, Draisaitl, MacKinnon) had a better years than all but one of Howe's best.

It's a systematic underrating of the era, because of flawed math.

The team adjustment size formula is fundamentally broken. It assumes when roster size increases Howe would see his minutes decrease by the same amount that Benny Woit would. It assumes that all the penalty killing time Howe had was an equal scoring opportunity as powerplay or even strength time.

 

[Michael Farkas]

Right, that's "our" point. Which is that H-R stinks at this.

EDIT: I was beaten by a much more informative response haha

 

[WarriorofTime]

The team adjustment size formula is fundamentally broken. It assumes when roster size increases Howe would see his minutes decrease by the same amount that Benny Woit would. It assumes that all the penalty killing time Howe had was an equal scoring opportunity as powerplay or even strength time.


Gordie Howe's peak is recognized as the highest offensive peak outside the rest of the Big Four.

Sorry, by who exactly? What are the people that claim that basing this off? What is their methodology?

I don't profess to have a strong opinion on the efficacy of the "roster size adjustment", why the creators of Hockey Reference chose it, what they used to base that on to draw inferences, etc. But I also am not going to blindly attack it, I'm sure their methodology like all can be made "closer to perfect", but it still works pretty darn well for the most part and I'm not going to throw the whole thing out over a statistical method not agreeing with some particular "general consensus"

Should add a pretty obvious point, even if you 'hate' it as applied to Gordie Howe, has very little (none actually) relevance to Crosby and McDavid who played in a fundamentally same NHL (roster size hasn't changed)

 

[jigglysquishy]

How does Howe's 1952-53 Art Ross become less dominant? It's one of the best era-adjusted seasons ever (131 on HockeyReference) and nobody else that season is anywhere in the vicinity.

You've just answered the question.

It's not that everyone else in 1952-53 sucked. It's that the math is fundamentally flawed and adjusts everyone lower than it should.

I don't profess to have a strong opinion on the efficacy of the "roster size adjustment", why the creators of Hockey Reference chose it, what they used to base that on to draw inferences, etc. But I also am not going to blindly attack it, I'm sure their methodology like all can be made "closer to perfect", but it still works pretty darn well for the most part and I'm not going to throw the whole thing out over a statistical method not agreeing with some particular "general consensus"

They picked it because they're baseball fans trying to compare Bonds and Ruth while chasing ad revenue.

It's no surprise that bad math derives from non fans rushing through on excel.

If you go by VsX, Vs5, VsNon Teammate, Vs average scoring of top 10. Howe's season stands above anyone besides 6 Gretzky years, 2 Lemieux years, maybe 1 Morenz.

If you go by Hockey Reference, it's 32nd all time. Or 18th if you're only looking at post war years.

They don't fix the formula because they don't care about it being right.

 

[WarriorofTime]

They picked it because they're baseball fans trying to compare Bonds and Ruth while chasing ad revenue.

It's no surprise that bad math derives from non fans rushing through on excel.

Sorry, this doesn't make sense? The issue you have is roster adjustment, Baseball has had a 9 player batting rotation since whenever stats started getting tracked so I don't see the 1 for 1 there.

To the extent you're able, will you please clarify what you mean by this more? I don't see what the gatekeeping "no true Scotsman..." fallacy has to do with statistical adjusting.

 

[jigglysquishy]

Sorry, this doesn't make sense? The issue you have is roster adjustment, Baseball has had a 9 player batting rotation since whenever stats started getting tracked so I don't see the 1 for 1 there.

To the extent you're able, will you please clarify what you mean by this more? I don't see what the gatekeeping "no true Scotsman..." fallacy has to do with statistical adjusting.

Exactly. If a baseball fan, with minimal familiarity with hockey, tries to adjust hockey stats there will be lots of things they mis-assume because they are thinking about baseball instead of hockey.

Baseball has always had 9 players. Hockey has tripled the skaters in 100 years. How would a non-fan adjust it? Simple, 1 for 1 ice-time adjustment.

Gordie Howe plays 25 minutes with 9 forwards. Well now that there's 12 forwards he plays (9/12) 18.75 minutes. And his point totals are therefore adjusted by 9/12.

But wait a minute. That doesn't make sense. Why would a coach, by going from 3 line hockey to 4 line hockey, take the 6.25 minutes away from Howe, proportionally the same as a third liner. Wouldn't you still try to play Howe more? 4th liners are only getting 8-10 minutes a game. Why are 6 of those minutes coming from your stars, and only 2-4 of those minutes coming from 2nd and 3rd lines together?

It completely misunderstands how the transition from 3 line to 4 line hockey happened from an ice time perspective. It's why the numbers look good for post 1990 hockey, look okay for 70s and 80s, and look downright confusing for 20s-60s.

This is literally the Hockey Reference formula.

 

[WarriorofTime]

Exactly. If a baseball fan, with minimal familiarity with hockey, tries to adjust hockey stats there will be lots of things they mis-assume because they are thinking about baseball instead of hockey.

Baseball has always had 9 players. Hockey has tripled the skaters in 100 years. How would a non-fan adjust it? Simple, 1 for 1 ice-time adjustment.

Gordie Howe plays 25 minutes with 9 forwards. Well now that there's 12 forwards he plays (9/12) 18.75 minutes. And his point totals are therefore adjusted by 9/12.

But wait a minute. That doesn't make sense. Why would a coach, by going from 3 line hockey to 4 line hockey, take the 6.25 minutes away from Howe, proportionally the same as a third liner. Wouldn't you still try to play Howe more? 4th liners are only getting 8-10 minutes a game. Why are 6 of those minutes coming from your stars, and only 2-4 of those minutes coming from 2nd and 3rd lines together?

It completely misunderstands how the transition from 3 line to 4 line hockey happened from an ice time perspective. It's why the numbers look good for post 1990 hockey, look okay for 70s and 80s, and look downright confusing for 20s-60s.

This is literally the Hockey Reference formula.

OK I understand your critique with the roster size methodology. It's a rather imperfect way to adjust for the fact that we have no clue what Gordie Howe's actual ice time was in 1952-53 or to what extent we care about it. I would presume McDavid on 20 minutes would be much more effective on a per minute basis than Howe would be on a 25 minutes basis (all else being equal), whereas Babe Ruth is going to be equally effective on a per bat basis if he bats one out of every six times versus one out of every nine times.

(I still think the "No True Scotsman" isn't entirely fair here). Regardless, that doesn't have much of an effect on a Crosby-McDavid discussion which was the original focus. I think we all agree adjusted scoring is useless in a pre-war setting, maybe slightly less relevant in an "O6" setting.. but if it's still useful in a.. "modern era" setting (call it whatever, last 50, 40, 30 years), I don't see any issue with applying it in this particular context since the OP motivation is entirely Crosby-McDavid driven.

 

[Michael Farkas]

we have no clue what Gordie Howe's actual ice time was in 1952-53

We do.

 

[WarriorofTime]

We do.

NHL didn't start tracking time on ice until 1998-99 IIRC

 

[MadLuke]

Gordie Howe plays 25 minutes with 9 forwards. Well now that there's 12 forwards he plays (9/12) 18.75 minutes. And his point totals are therefore adjusted by 9/12.

And that the nicest way to put it, sometime team had 4/4.5D and not that dissimilar amount of forwards.

but if it's still useful in a.. "modern era" setting (call it whatever, last 50, 40, 30 years), I don't see any issue with applying it in this particular context since the OP motivation is entirely Crosby-McDavid driven.

The issue will be quite similar outside the roster size which something that disappear around the 80s.

Does Crosby 2006-2007 should be by a nice little amount above 2013-2014 Crosby or that more the reflection that 2006-2007 was the 7th season in league history with the most PPO and 2013-2014 was #50.

Should this:

Sidney Crosby	1.00​
Ryan Getzlaf	0.84​
Claude Giroux	0.83​
Tyler Seguin	0.81​
Corey Perry	0.79​
Taylor Hall	0.77​
Phil Kessel	0.77​
Nicklas Bäckström	0.76​
Jamie Benn	0.76​
Alex Ovechkin	0.76​

Adjust lower than this:

Sidney Crosby	1.00​
Joe Thornton	0.95​
Vincent Lecavalier	0.90​
Dany Heatley	0.88​
Martin St. Louis*	0.85​
Marián Hossa*	0.83​
Joe Sakic*	0.83​
Jaromír Jágr	0.80​
Marc Savard	0.80​
Daniel Brière	0.79​

When from #2 to 10 Crosby separation was higher ? If so why ? For what interest us, Hockey Reference not being interested in scoring distribution can be a significant flaw and that can change fast, 05-06 and 06-07 was a different league than the 2011 league which was different than the 2021 bubble one.