Hockey analytics: Does size really matter in the NHL?

[Curt]

42 minutes ago, PerreaultForever said:

Well, firstly, keep in mind I'm not talking about producing, I'm talking about winning, and a winning team needs to be a balanced team.

Second, "bigger" isn't really the thing, rather stronger. Already had to argue past the semantic points and want to stress it's never just about size or height. It's strength. 

Now as for producing, there are big strong producers. Ovechkin is a beast. Crosby isn't big, but pound for pound he is strong and quite powerful. Getzlaf in his prime was maybe the perfect combination of size, strength, grit and skill. Marchand is a little guy, but I'd bet pound for pound he's about the toughest in the league and he positions himself and leverages his body perfectly driving right past and through bigger defenders. There's no real point in listing more names but there are lots. 

As for the smaller guys being producers, yes, they are, in the fair weather, but they disappear in the playoffs if they aren't in the tough strong group. Mitch Marner for example. In the tough games, Draisaitl usually is bigger impact than McDavid . That is a size and strength issue. 

Why are smaller guys often more skillful? Probably because they have to be or they never make it to the NHL. 

I don’t really disagree with you.

However, my question is relating directly to the subject of this thread.  In the NHL, larger players do not produce, or create production for teammates, at a higher rate than smaller players.  Why is this?  What could it mean?

If thats not a conversation that you are interested in having, that’s ok.  It’s more a question for everyone, as opposed to you specifically, anyway.

[SwampD]

37 minutes ago, Curt said:

Can you back this up with any data?

Yes. but I don't care that much to post a bunch of links, but here's this.

https://www.eliteprospects.com/league/nhl/teams-physical-stats/2020-2021

Click on weight.

Edited November 20, 2021 by SwampD

[PerreaultForever]

36 minutes ago, Curt said:

I don’t really disagree with you.

However, my question is relating directly to the subject of this thread.  In the NHL, larger players do not produce, or create production for teammates, at a higher rate than smaller players.  Why is this?  What could it mean?

If thats not a conversation that you are interested in having, that’s ok.  It’s more a question for everyone, as opposed to you specifically, anyway.

I gotcha, and I was alluding to that in the last part. I would think you need a better breakdown of the stats by taking out the goons and big bodies brought in just for that reason. Little guys never make the NHL as tough guys, they are only fast and skilled but big guys can be there as tough guys as well as skill guys. So if you take a half dozen small guys and a half dozen big guys the stats will tell you the small guys put up more points on average even if a few of those big guys are bigger producers. I doubt it's anything more complicated than that.

As I was saying, small guys have to be fast and skilled or they never get in the NHL. Big guys excel if they are skilled but can also make the league as tough guys with less skill. 

Does that not explain the math? I think it does but am open to being proved wrong. 

[Taro T]

5 minutes ago, PerreaultForever said:

I gotcha, and I was alluding to that in the last part. I would think you need a better breakdown of the stats by taking out the goons and big bodies brought in just for that reason. Little guys never make the NHL as tough guys, they are only fast and skilled but big guys can be there as tough guys as well as skill guys. So if you take a half dozen small guys and a half dozen big guys the stats will tell you the small guys put up more points on average even if a few of those big guys are bigger producers. I doubt it's anything more complicated than that.

As I was saying, small guys have to be fast and skilled or they never get in the NHL. Big guys excel if they are skilled but can also make the league as tough guys with less skill. 

Does that not explain the math? I think it does but am open to being proved wrong. 

Mike Peca & Brad Marchand say 'hi.'

[PerreaultForever]

1 hour ago, Taro T said:

Mike Peca & Brad Marchand say 'hi.'

Oh come on, I mentioned Marchand earlier as a perfect example of a little guy who was "pound for pound as strong as anyone". Briere was really strong for his size too, but these guys are outliers and exceptions to the general rule. 

[Taro T]

7 minutes ago, PerreaultForever said:

Oh come on, I mentioned Marchand earlier as a perfect example of a little guy who was "pound for pound as strong as anyone". Briere was really strong for his size too, but these guys are outliers and exceptions to the general rule. 

OK.  Missed the sarcasm of your earlier post. Sorry.  It was a long week and our oldest gave us a good scare today.  Beer is good.

[Curt]

6 hours ago, PerreaultForever said:

I gotcha, and I was alluding to that in the last part. I would think you need a better breakdown of the stats by taking out the goons and big bodies brought in just for that reason. Little guys never make the NHL as tough guys, they are only fast and skilled but big guys can be there as tough guys as well as skill guys. So if you take a half dozen small guys and a half dozen big guys the stats will tell you the small guys put up more points on average even if a few of those big guys are bigger producers. I doubt it's anything more complicated than that.

As I was saying, small guys have to be fast and skilled or they never get in the NHL. Big guys excel if they are skilled but can also make the league as tough guys with less skill. 

Does that not explain the math? I think it does but am open to being proved wrong. 

I actually had the same idea regarding 4th line/defensive forward type guys, who I think are generally a little larger, dragging down the numbers.

I think that probably does go a long way towards explaining it.

I suspect that if you looked only at forwards receiving top-6 ice time, larger players would out produce smaller ones by a little.

I wonder about the numerical distribution by size throughout the lineup.  What’s the average size of a top-6 F, or top-4 D, vs the NHL average?  I’d be interested in seeing that as well.

[North Buffalo]

18 hours ago, SwampD said:

I'll just throw this pack of firecrackers into the room.

All things being equal: hands, speed, vision, skating,... would Tage thompson be as affective if he was 6'?

 

For the record, I don't think size makes that much of a difference in the regular season, but Newton's laws of motion absolutely come into to play a lot more in the playoffs.

Fair enough except that imo it is easier to find guys smaller maybe a matter of degrees... I wonder if there is a cut off in difference both on offense and defense in that equation... is 10 lbs and 2 inches?   or 20 lbs and 4 inches and when does size have diminishing returns based on ice size and the ability of a smaller player to get under and outleverage a bigger guy all things being equal at a playoff level.

[triumph_communes]

Selection bias. Only players who are talented get to play in the NHL. Only smaller players whose talents exceeds  their body limitations play in the NHL. Similarly, larger players with less talents still make it if there’s any benefit to their size alone. Therefore the production is leveled because of selection bias.

 

What changed from the 60s to the 90s?  Drafting!  (Aka the introduction of the selection bias).

 

Stats 101 is above journalism and sabrespace posting I see. 

Edited November 20, 2021 by triumph_communes

[Curt]

18 minutes ago, triumph_communes said:

Selection bias. Only players who are talented get to play in the NHL. Only smaller players whose talents exceeds  their body limitations play in the NHL. Similarly, larger players with less talents still make it if there’s any benefit to their size alone. Therefore the production is leveled because of selection bias.

 

What changed from the 60s to the 90s?  Drafting!  (Aka the introduction of the selection bias).

 

Stats 101 is above journalism and sabrespace posting I see. 

Could you expand a bit on the bolded?  How did the implementation of the draft create a selection bias that did not exist previously?

[SwampD]

What does selection bias have to do with two guys in their fifth and seventh years fighting for a puck in the corner?

Seems weird.

[Marvin]

Thanks, everyone, for all of the attempts to explain the data.

I see a lot of truth in a lot of the explanations, such as @triumph_communes on selection bias, @SwampD on the possibility that smaller teams might skew towards the bottom of the standings, and @Curt for noting that the bottoms of rosters probably skew big.  I expect that the answers are as complex as we are observing in our discussions.

[LGR4GM]

6 hours ago, triumph_communes said:

Selection bias. Only players who are talented get to play in the NHL. Only smaller players whose talents exceeds  their body limitations play in the NHL. Similarly, larger players with less talents still make it if there’s any benefit to their size alone. Therefore the production is leveled because of selection bias.

 

What changed from the 60s to the 90s?  Drafting!  (Aka the introduction of the selection bias).

 

Stats 101 is above journalism and sabrespace posting I see. 

That's not selection bias. 

You can say the draft has selection bias but what you're describing is not selection bias. 

[LGR4GM]

What I'm saying is you can argue the league has selection bias at the draft but does size mean better production wouldn't be selection bias. Maybe I just don't understand the awkward wording around the "production is leveled"

[triumph_communes]

22 hours ago, LGR4GM said:

That's not selection bias. 

You can say the draft has selection bias but what you're describing is not selection bias. 

Yes it is.  You’re confusing the context. The dataset used is biased therefore the hypothesis was not tested properly.  The conclusion that size does not matter based on that dataset is improper. 
 

The proper assessment is even easier than that effort article partaked in. It would have been to just take the average size of NHL players, which are the selected set of the human population. I’m sure it’s higher than the regular population, therefore some size is beneficial.

 

Thing that’s hard for people conceptualize is percentages.  If the correlation only accounts for let’s say 15% of the variation, then yes size does matter, but it is just one factor of many.

 

I think if anything has been made more clear in this league is coaching is a higher factor than any group of players. Bottom feeder teams can change coaches and win the cup the very same year. It’s happened many times before.  This is why I have such problems with all the stats that get thrown out all the time, they are more strongly tied to the player usage and system than the player itself.  The stats that are all relative within a team is a step beyond that, but each time you do this the more data is needed to make it worthwhile. 

Edited November 21, 2021 by triumph_communes

[PerreaultForever]

idk exactly what's being implied by "selection bias" aside from the idea that if there are two similar guys with similar skills you might take the bigger one, but another factor teams consider that isn't being discussed really is durability.

  Teams often shy away from smaller guys figuring they won't hold up to the physical side of the game and for the most part it's not wrong. 

The game has of course changed and their are more places for small talented players, but there's never going to be a situation where you have two otherwise equal guys and teams take the little guy. 

[Curt]

On 11/20/2021 at 8:04 AM, triumph_communes said:

Selection bias. Only players who are talented get to play in the NHL. Only smaller players whose talents exceeds  their body limitations play in the NHL. Similarly, larger players with less talents still make it if there’s any benefit to their size alone. Therefore the production is leveled because of selection bias.

 

What changed from the 60s to the 90s?  Drafting!  (Aka the introduction of the selection bias).

 

Stats 101 is above journalism and sabrespace posting I see. 

 

On 11/20/2021 at 8:24 AM, Curt said:

Could you expand a bit on the bolded?  How did the implementation of the draft create a selection bias that did not exist previously?

Seriously, could you expand on this?  Some of us actually have not taken Stats 101.

Explain it it to us like we are 12.

[SwampD]

4 hours ago, triumph_communes said:

Yes it is.  You’re confusing the context. The dataset used is biased therefore the hypothesis was not tested properly.  The conclusion that size does not matter based on that dataset is improper. 
 

The proper assessment is even easier than that effort article partaked in. It would have been to just take the average size of NHL players, which are the selected set of the human population. I’m sure it’s higher than the regular population, therefore some size is beneficial.

 

Thing that’s hard for people conceptualize is percentages.  If the correlation only accounts for let’s say 15% of the variation, then yes size does matter, but it is just one factor of many.

 

I think if anything has been made more clear in this league is coaching is a higher factor than any group of players. Bottom feeder teams can change coaches and win the cup the very same year. It’s happened many times before.  This is why I have such problems with all the stats that get thrown out all the time, they are more strongly tied to the player usage and system than the player itself.  The stats that are all relative within a team is a step beyond that, but each time you do this the more data is needed to make it worthwhile. 

This post said nothing. Keep putting a bunch of words together that make it sound like you know stuff, but until you actually make a point, I choose to believe that you are a bot.

Selection bias has nothing to do with size and how it relates to play in the NHL.

[Thorny]

1 hour ago, SwampD said:

This post said nothing. Keep putting a bunch of words together that make it sound like you know stuff, but until you actually make a point, I choose to believe that you are a bot.

Selection bias has nothing to do with size and how it relates to play in the NHL.

I think what he's saying generally makes sense

[Thorny]

If size is being over emphasized more than other traits, fewer "spots" are available than there should be for players who maximize those other traits. Stiffer competition for those spots results in mostly the cream of the crop rising, therefore the statistical output of that group being higher than the average of the over-emphasized group makes sense, as that group will be watered down. 

The pursuit of a specific trait will water down it's whole pool 

This is a certainty considering the "low" average result of drafting/developing to begin with

And I don't believe emphasis/bias on a particular trait would "die" after the draft - I'd imagine that continues into roster building. 

Edited November 22, 2021 by Thorny

[SwampD]

9 minutes ago, Thorny said:

I think what he's saying generally makes sense

I had to reread it to get it,… kinda.

Selection bias still has nothing to do with it.

[Marvin]

I would like to see height, weight, position, ice time, production, and other stats run with a nice multivariate statistical analyses.  That would be the grist for a MS thesis, I should think.

[triumph_communes]

10 hours ago, SwampD said:

This post said nothing. Keep putting a bunch of words together that make it sound like you know stuff, but until you actually make a point, I choose to believe that you are a bot.

Selection bias has nothing to do with size and how it relates to play in the NHL.

All you’ve proven is this is over your head. I can keep rewording it until you understand or you can take offense and I can’t fix that.  Size does matter or NHL players wouldn’t be larger on average than most humans.  Take jockeys for example - much smaller on average then most humans because smaller size matters there.  But the meaning of “matters” is just one factor of many and can easily be overcome by skill, smarts, leadership, etc. Selection bias has to do with how a particular dataset is being interpreted and the OPs article succumbed to it. That has nothing to do with the concept of size, it’s just how that particular dataset came to be. 

Edited November 22, 2021 by triumph_communes

[triumph_communes]

14 hours ago, Curt said:

 

Seriously, could you expand on this?  Some of us actually have not taken Stats 101.

Explain it it to us like we are 12.

The fallacy the article has is taking all current NHL players and plotting their size versus points as the basis for assessing whether size matters in the NHL.
 

The fact is the selection whether size matters happened before the dataset of ‘NHL Players’ came to be.  This is the bias.  It happened in deciding who is good enough to be in the NHL. So to address the question you have to assess whether NHL players are different in size than the general population (more specifically the populations that hockey players come from) and see if there is a difference there and apply statistical tools. If there is a difference (I’m lazy and assuming yes hockey players are larger humans on average than the typical 5’9”), then you can conclude it is an important factor.
 

The ‘tell’ is how many NHL players are a certain height relative to how many people are a similar height.  If hockey players average 6’1” but people are 5’9” then yes, size “matters” plain and simple.  How much does it matter?  That’s where we’d have to get much more complicated 

let’s be honest in the 60s the pool of hockey players was smaller and more representative of a general population.  It was more about who had the money to afford equipment, spend their lifetime in a hockey rink being trained, etc.  and at the end of the day they plotted size vs points and larger players were better.  As the years went on and the privileges were torn away as teams were in pursuit of the best players, this difference also went away.  And that’s what the article saw by separating those time periods.  After teams did everything they could to select the best hockey players, size among those selected to play hockey no longer mattered anymore. And that’s not because size doesn’t matter, it’s because the players who couldn’t overcome size disadvantage didn’t make it to the NHL anymore. What could be interesting here is seeing how the size of players changed as a whole changed over those periods as well to prove what I just alluded to. 

[SwampD]

5 hours ago, triumph_communes said:

All you’ve proven is this is over your head. I can keep rewording it until you understand or you can take offense and I can’t fix that.  Size does matter or NHL players wouldn’t be larger on average than most humans.  Take jockeys for example - much smaller on average then most humans because smaller size matters there.  But the meaning of “matters” is just one factor of many and can easily be overcome by skill, smarts, leadership, etc. Selection bias has to do with how a particular dataset is being interpreted and the OPs article succumbed to it. That has nothing to do with the concept of size, it’s just how that particular dataset came to be. 

Thanks Mr Wizard. I think the disconnect is that I’ve only ever heard the term selection bias, as it relates to the NHL, being that teams are more likely to draft older players.