The Peak and Decline of Elite NHL Forwards

[Appleyard]

I know some work on this has been done before, and some posters here have done some work in this regard as well that I really enjoyed.

However, when I was looking through the studies that had been done, I noticed that:

A. There had been relatively few that focus only on forwards.

B. There had been few of these that only focus on very high tier forwards.

C. These studies had generally looked at Raw points instead of PPG.

D. None had seemingly taken era into account, so for example Jagr's 2005-06 season ranks similarly to 2000-01.

E. There had not been much in terms of dividing goals, assists and points.

So I figured since I had a few days free time after finishing my Masters I would look into it, first chance I have had to do any real in depth analysis of anything NHL related since I did one on draft pick probability a few years ago. Below are the results in article form.

http://alongtheboards.com/2015/09/peak-decline-elite-nhl-forwards/

Hope you enjoy reading it as much as I enjoyed researching it and writing it. If you have any questions feel free to ask!

Click to read more...

 

[Ohashi_Jouzu*]

It seems like you come to the same conclusion everyone else does that such players peak out between 26-28.

 

[Appleyard]

It seems like you come to the same conclusion everyone else does that such players peak out between 26-28.

Yeh, I would say from these figures that:

27 is most likely when a top forward hits his absolute peak.

Age 26-29 is likely when they will be right around peak level.

Age 24-26 and 30-31 they should be not too far off their peak... but with a noticeable gap.

But that a player who is more of a goalscorer likely peaks age 26-27 while a playmaker age 28-29.

I do think by disregarding injuries and looking only at top forward though it seems the absolute peak is a bit later than many have previously seen. (though only ~1-2 years. 25-26 seems most prevalent in a lot of studies looking at raw points or a wider range of players.)

What I personally found most interesting was:

These players barring injury are still generally 1st line (though middling 1st liners, especially after age 34) guys well into their 30's.

Assists don't change much per game from age 24-32. (yet goals have a more obvious bell curve.) And age 33-34 assists are at a similar level as at age 22-23. (the same cannot be said for goals.)

 

[lomiller1]

To what do you attribute the difference from other studies? Is it primarily due to the focus on top end players?

 

[Appleyard]

To what do you attribute the difference from other studies? Is it primarily due to the focus on top end players?

I think the difference of ~1.5 years in terms of peak compared to previous studies is likely a combination of a few things.

Looking at only top end players is likely one of them. These top guys play until age ~37 on average... while middle sixers are generally done a few years earlier, it makes sense that the extra ability they have acts as more of a 'buffer' between them and retirement/serious decline, as they have more leeway to decline but still be a valuable player. If a middle sixer lost ~30% of their production as they age they would be a 35 point player who is in their mid 30's... if that is the case and they do not offer exceptional D they are played down the line-up or out of the league, while if a once elite player loses ~30% of production they are still a valuable top 6 asset. The extra year or two in terms of decline is likely also accounted for by this.

But looking at PPG (well, effectively PPG) instead of raw points in my mind is certainly a contributory factor... players do miss more time as they age, it does not make them a 'lessor' player but it does mean that over a season they contribute less raw points.

The lack of adjustment for era in previous work could also have an influence, 80s and early 90s to late 90s and early 00's, and then early 00's to post lockout are such massively different scoring climates. If narrow samples in terms of age were taken and looked at in terms of peak it could skew a mean peak slightly. (for example guys born in the late 60's generally entered the NHL in a very high scoring period, but aged ~26-35 contended with a period of very low scoring, while a player born in the early 80's would have generally entered the NHL in a very low scoring period, before a brief spike in scoring climate aged ~24-26 (where virtually all those guys had in terms of 'raw, unadjusted' numbers their best year) and then one of the lowest scoring eras since the war.

I think when I have some time I will try and use the same methodology and look at maybe middle sixers or second liners... the definition of such a player will likely be harder to come up with and therefore the sample harder to determine, but the results in comparison will likely be elucidating.

 

[LeSandvich]

I wonder what the team-impact peak of a player is. Around 27-28 aswell? Or could an older player, say 32 years old, have the same, if not more impact, if he gives up points for defensive duties?

 

[charlie1]

I know some work on this has been done before, and some posters here have done some work in this regard as well that I really enjoyed.

However, when I was looking through the studies that had been done, I noticed that:

A. There had been relatively few that focus only on forwards.

B. There had been few of these that only focus on very high tier forwards.

C. These studies had generally looked at Raw points instead of PPG.

D. None had seemingly taken era into account, so for example Jagr's 2005-06 season ranks similarly to 2000-01.

E. There had not been much in terms of dividing goals, assists and points.

So I figured since I had a few days free time after finishing my Masters I would look into it, first chance I have had to do any real in depth analysis of anything NHL related since I did one on draft pick probability a few years ago. Below are the results in article form.

http://alongtheboards.com/2015/09/peak-decline-elite-nhl-forwards/

Hope you enjoy reading it as much as I enjoyed researching it and writing it. If you have any questions feel free to ask!

Very nice work. I was dying for a chart though instead of those tables.

 

[Appleyard]

Very nice work. I was dying for a chart though instead of those tables.

Haha, sorry! If I have some time in the next day I'll get some charts from it into this thread.

 

[Luigi Lemieux]

Interesting analysis. Seems like on average there's a big drop from age 31 to age 32. Wonder why that happens at that age in particular and why it's such a steep decline.

 

[LeSandvich]

Interesting analysis. Seems like on average there's a big drop from age 31 to age 32. Wonder why that happens at that age in particular and why it's such a steep decline.

Pretty sure its just statistical noise

 

[Appleyard]

Interesting analysis. Seems like on average there's a big drop from age 31 to age 32. Wonder why that happens at that age in particular and why it's such a steep decline.

In terms of % from the previous year the drop in points is quite similar around that age:

29-30: 4.6% drop
30-31: 4.8% drop
31-32: 4.3% drop
32-33: 6.2% drop

And assists are pretty steady... so it is only goals that drop at that age in a more pronounced way, the mean and median both see a drop and then a stabilisation as well, and only 7 of the players had more goals aged 32 than 31, while 19 had more assists at 32 than 31. The size of the sample has not unduly decreased at that point either, both ages have over 40 players playing.

It could be a quirk of a sample of 'only' 50 (there are little quirks with such a 'comparatively' small sample... like goals rising slightly across mean and median at age 35) and the decline would a smoother curve in another sample. (I do think in the next few weeks I will look at a sample of other 1st liners who did not make the 'elite' cut or something along those lines to compare the results... and try and get a sample of ~100 or more.)

One of the only things I can really think of are that it is influenced by the average decline of fast-twitch muscle fibre... depending on the medical journals/professionals believed ~30 is the average age when the decline in muscle mass starts (after a peak in the late 20s)... and most of this decline is the loss of fast-twitch fibre, which would have far more influence on shooting ability than passing I suppose (and has been discussed here previously). 31-32 is not so far of ~30 to make it believable that this has an influence... though it is generally a pretty linear decline, so this throws a bit of a spanner into that one!

 

[Raccoon Jesus]

Love it Appleyard. We've been talking about this a lot on the Kings' forum with regards to Kopitar's next contract (realizing, of course, that he's not to be mistaken with 90-100+ point players). Will link this over there and you'll probably get a little bit more noise

 

[Appleyard]

Love it Appleyard. We've been talking about this a lot on the Kings' forum with regards to Kopitar's next contract (realizing, of course, that he's not to be mistaken with 90-100+ point players). Will link this over there and you'll probably get a little bit more noise

Tbf if we adjust Kopitars points for the era he plays in he is around ~85 points in his best 3 seasons... not too far off the 'average' player in this study offensively. (and he is far better defensively than the average guy included here.)

 

[Feed Me A Stray Cat]

Good work, however I think there might be a bit of selection bias in the sample you chose.

Simon Gagne and Miroslav Satan probably did not look very different than Jarome Iginla over the first seven years of their respective careers. Excluding them from the sample because of not enough upper echelon repetition, when regression due to age might have been a reason for that, seems questionable.

Meanwhile, someone like Daniel Briere is in the "elite" set. So is Todd Bertuzzi, Marc Savard and Jason Allison. Why are they included but someone like Olli Jokinen excluded?

 

[Appleyard]

Good work, however I think there might be a bit of selection bias in the sample you chose.

Simon Gagne and Miroslav Satan probably did not look very different than Jarome Iginla over the first seven years of their respective careers. Excluding them from the sample because of not enough upper echelon repetition, when regression due to age might have been a reason for that, seems questionable.

Meanwhile, someone like Daniel Briere is in the "elite" set. So is Todd Bertuzzi, Marc Savard and Jason Allison. Why are they included but someone like Olli Jokinen excluded?

It was the trickiest bit, the selection of players... the gap between the bottom end guys who got in and the top end guys who did not was quite small. In the end I really ceded to what their absolute peak was and their position amongst peers at that point offensively. I though long and hard over Gagne and Jokinen especially, I think those two, Satan and Marleau were ones I had in the sample at one point but excluded, mainly due to their highest peak not making them a top 10-15 scorer or pacing for it. Briere was also at one point not in the selection.

Satan peak: age 28, 87 adjusted points (25th in PPG)
Gagne peak: age 25, 89 adjusted points (23rd in PPG)
Marleau peak: age 30, 89 adjusted points (18th in PPG)
Jokinen peak: age 28, 93 adjusted points (17th in PPG)

Briere peak: age 28 & 29, 97 adjusted points (11th & 11th in PPG)
Allison peak: age 25, 102 adjusted points (9th in PPG)
Savard peak: age 26, 109 adjusted points (3rd in PPG)
Beruzzi peak: age 26, 110 adjusted points (1st in PPG)
Iginla peak: age 29, 112 adjusted points (3rd in PPG)

It was hard to find the balance and not that 'open the doors' to guys like Bill Guerin, Corey Stillman, Jozef Stumpel etc... though on reflection I think Jokinen is a guy who could have been included without necessarily being a guy who opens the doors to 'non-elite' offensive guys.

I hope that goes some way to explaining the reasoning behind it. At some point in the next months or so I also intend to do exactly the same study but looking at the 'leftover' guys and others who were for a time '1st liners'.

 

[AlienWorkShop]

Just a very random thought that I wasn't sure where to put anywhere else, so sorry this doesn't directly comment on your work... but came to mind amid recent discussion on NHL partying.

Was thinking after seeing that a couple former elite/very good players who've been rumoured/publically admitted to having drinking issues had just signed in Europe at relatively young ages, it makes me wonder how peaks differ between players that "treat their body like a temple" and those that perhaps indulge in partying. Not necessarily those who truly go off the rails, like Fleury, but are regular boozers who perhaps have one eye on the game and one eye on the carnal pleasures of being young and rich.

For example, I've heard Stamkos is a teetotaller (that may be entirely wrong, but just as a hypothetical example), so I wouldn't be surprised if his productive years last into his late 30s. You could probably tally numerous examples of players whose careers/general effectiveness dropped dramatically by 30.

Pretty much impossible to truly control for, but there must be a correlation between career longevity and... "partying". We've seen plenty of hard partyers win Cups, but I'm just thinking from a long-term, career perspective.

I don't really cast any moral judgment on players who like to go out for some beers or perhaps even more illicit substances, but from a fan's perspective, it can be an indicator of how truly 'dedicated' a player is and whether hockey really is his end all, be all. Something I'm sure GMs and coaches take note of...

 

[LeSandvich]

Just a very random thought that I wasn't sure where to put anywhere else, so sorry this doesn't directly comment on your work... but came to mind amid recent discussion on NHL partying.

http://www.hockeyvips.com/2015/08/german_titov_on_jaromir_jagr/

“Am I surprised Jagr is still playing? No,†Titov said. “He doesn’t drink. The man loves the party scene but he never breaks his regimen. It’s not for him. He is very serious about his training.

 

[Appleyard]

Did a bit of a follow up to this article, as the topic fascinates me so much.

Available here:

http://alongtheboards.com/2015/11/the-peak-and-decline-of-1st-line-nhl-forwards/

This one looks at players who were producing at the level of a 1st line forward for some stage of their career instead of forwards who were 'elite' at some stage in their career.

It also compares the paths the elite forwards took compared to the 1st line forwards.

It has graphs in this time!

Any questions feel free to ask...

I think the last thing I will do in the next few weeks is split this '1st line' sample into a few smaller samples of players with similar career paths. (ie one sample of guys who came in, shone and tailed off before their early 30's, one sample of guys who did not start putting up good numbers until late 20's/early 30's, and one of guys who were consistently at a '1st line' level from early 20's to end of career.)

 

[charlie1]

Nice work (and thanks for the graphs!). You've shown pretty convincingly that an average 1st line forward will decline less rapidly than an elite forward.

I didn't see you speculate much as to the reason for that. Any thoughts on underlying mechanisms?

 

[Appleyard]

Nice work (and thanks for the graphs!). You've shown pretty convincingly that an average 1st line forward will decline less rapidly than an elite forward.

I didn't see you speculate much as to the reason for that. Any thoughts on underlying mechanisms?

Why? I imagine it could be a few things, but here would be my main suggestion regarding that:

The elite players have further to fall...elite forwards are still very useful players (top 6ers almost exclusively) as they decline... even if they do so by say ~50% of production, if a 1st liner declines by 50% they are generally not in the league any-more, so do not figure into the data-set any-more.

When the individual players are looked at, generally once the 1st liners suffer a big down year they are out of the league the year after, unless they are very good defensively... while the elite forwards can have a similar down year ratio wise and still keep playing, and impacting the averages negatively.

A lot of the 1st liners never had the 'chance' to seriously decline over a few years!

For example, percentage of each data-set who had played their last 30+ game season at certain ages:

Age 31: Elite (2%:1/50) 1st line (14%:12/86)
Age 32: Elite (4%%:2/50) 1st line (17.45%:15/86)
Age 33: Elite (6%:3/50) 1st line (29.1%:25/86)
Age 34: Elite (18%:9/50) 1st line (39.5%:34/86)
Age 35: Elite (32%:16/50) 1st line (48.8%:42/86)
Age 36: Elite (44%:22/50) 1st line (64%:55/86)

In effect the 1st liners decline is being statistically masked a bit in comparison to the elite players as those 1st liners who started to obviously decline production wise were not afforded more years in the league to decline even more.

So if we lay out a scenario:

Your team has a good 1st liner in his twenties. Your team also has an elite player the same age.

If they are both still in the league at ~33-35 the 1st liner will be playing at a good level, probably better relatively than the ex-elite guy.

But the odds are far worse that the 1st liner is still in the NHL at that age, while the elite guy is very likely still playing. I have not calculated it off these data sets, but I looked at elite players ages of retirement last year for my own interest... and if I remember correctly elite guys average retirement age is ~37, while it seems like 1st liners go on average at ~35.

 

[Appleyard]

Though also of interest is the fact that the difference is virtually all down to the drop in goalscoring, as the rate they compile assists at as they age is pretty similar ratio wise to their peaks... the elite forwards goals per game has a greater decline up until the players age 32 season, but not too markedly... but then the elite players suffer a far bigger relative drop at ages 32, 33 and 34...

So really the lesser decline is due to the 1st liners maintaining their goalscoring better than the elite forwards.

That is not accounted for by the greater attrition rate in 1st liners as well... so that in itself is interesting.

Maybe as a result of the elite guys still generally being 'the' man as the age, while the 1st liners are generally the 2nd or 3rd most important guy on their line... so the elite guys are declining, and still the driving force on the ice, so we see that decline in goals due to ageing more markedly, while the 1st liners are generally more complimentary players and don't drive the offense as much, so possibly can 'leach' more off others as they age to keep totals higher, while elite guys don't have that luxury!

 

[PSGJ]

Great work. I do wonder though when the peak in overall performance is. Because I got to figure that a guy in his early 30's is a whole lot more useful in the defense than a guy than a guy who scores just as much and is in his early 20's.

 

[The Shadow]

It seems like both Crosby and Ovechkin hit their scoring peak at age 24 or so.

McDavid is closing in on 24. Can he hit another tier and hit a higher point total with the cast he is playing with ?

 

[authentic]

Love it Appleyard. We've been talking about this a lot on the Kings' forum with regards to Kopitar's next contract (realizing, of course, that he's not to be mistaken with 90-100+ point players). Will link this over there and you'll probably get a little bit more noise

That ended up working out okay for you guys last season.

 

[authentic]

Great work. I do wonder though when the peak in overall performance is. Because I got to figure that a guy in his early 30's is a whole lot more useful in the defense than a guy than a guy who scores just as much and is in his early 20's.

Probably still 25-30, but overall early 30s over early 20s. Just compile two teams of players those ages, early 30s would win easily IMO.