NHL Points-per-game peak age estimation

[danishh]

i'm sure some most of you have already seen this 2010 post by behindthenet (now moved to arctic ice hockey)

http://www.arcticicehockey.com/2010/1/21/1261318/nhl-points-per-game-peak-age

I looked at all NHL players born 1962-79 who played exclusively in the NHL, AHL or IHL from Age 21 to Age 29. I assumed that AHL and IHL points were worth 45% of an NHL point. The average points-per-game at each age came out as follows:

it's not that deep but an interesting reference, and perhaps something that someone here would like to expand upon/update in the future.


jesse spector took a much more limited look at a few star players this year and determined that post 26, things start going downhill.
http://aol.sportingnews.com/nhl/sto...rio-lemieux-pittsburgh-penguins-evgeni-malkin

 

[metalfoot]

As a general trend, yes, but not as a predictor; classic counter-example would be Daniel Alfredsson, whose peak in his career, points-wise, was from ages 30-35.

 

[Iain Fyffe]

Yes, this is an important fact to bear in mind, and makes you realize that by the time players are UFAs, their production is already declining. Yet that's when they are paid the most. The salary:value function in the NHL is really funky because of this.

 

[Ozamataz Buckshank]

Where does the 45% conversion factor come from? He doesn't explain that all, unless it just makes his data set look like a smooth trend

 

[Iain Fyffe]

Where does the 45% conversion factor come from? He doesn't explain that all, unless it just makes his data set look like a smooth trend

That's based on historic data IIRC, based on players who play in the IHL one year and the NHL the next or vice-versa.

 

[GKJ]

This is one of the fiercest arguments, because based on the study, no one ever gets a UFA contract within their prime, and people think that players are still in their prime when they turn 30. Or assume that someone being traded to a better team with supposed better teammates will cause their point totals to immediately sky rocket (Nash).

 

[Guest]

I think I'd like to see it comparing NHL to NHL, IHL to IHL and AHL to AHL better than this hybrid. I think it may skew the results by trying to squeeze in IHL & AHL stats.

It also is interesting to me because I have often found that players (historically) would break out about the time they are 25 years old. I remember getting list of players each year based on their age and trying to pick out which players were going to break out that year. I know players can break out at any age, sometimes younger and sometimes older but I found that 25 years old was the bell curve of the majority of players that did break out. Since then I have often judged a player's value or worth at their performance at age 25 whether being a bust or someone with potential.

 

[CanadianHockey]

I know I keep saying it whenever this pops up on the Sens board, but I'd like to see this study repeated with controls for different scoring rates in different eras, as well as creating creating separate data sets for each position.

 

[GordieHoweHatTrick]

Further evidence as to why it's important to build from within in a cap world.

 

[mindmasher]

Where does the 45% conversion factor come from? He doesn't explain that all, unless it just makes his data set look like a smooth trend

http://www.behindthenet.ca/projecting_to_nhl.php

 

[Czech Your Math]

I'm looking for any other studies of value or total number of players by age.

The more comprehensive (by position, various metrics, etc.) and more historical, the better.

 

[danishh]

here's another one from dan tolensky (hockeybuzz). Looked at the goal-scoring trends by age for the top 100 goal-scorers in history (at that time, the study is from 2007).

http://www.hockeybuzz.com/blog.php?post_id=8663

The above graph is more than a little messy to say the least. On the X-axis you have players ages ranging from 18 to 42 and on the Y-axis are goals scored. Each line on the graph represents one of the top 100 goal scorers. I was going to remove these lines and just show the averages but I thought it was interesting to see how certain players stand out above the rest. If you're familiar with the stats then right away you'll be able to spot Wayne, Mario, Espo, Gordie, Golden Brett, Teemu and other

and just the black line for clarity:

 

[danishh]

Harri Mikael of Smoke Degrees (A washington capitals blog)

http://smokedegrees.blogspot.ca/2011/05/study-when-do-great-goal-scorers-peak.html

takes a look at the nhl's 500-goal scorers and the distribution of their top-3 individual goal-scoring years by age.

 

[danishh]

Budd Bailey of the buffalo news (formerly worked for the Buffalo Sabres)

http://buddbailey.blogspot.ca/2012/07/peak-performance.html

(july 2012)

took a bunch of active players who had played for at least six years in the NHL, and in some cases many more, and wrote down their points year by year. A player's best season got a score of 100, and his others were a percentage of that figure. So, if Dave Andreychuk scored 100 points at age 28 for a career high, age 28 received a score of 100. If he scored 50 points at age 22, that received a 50. Do that for about 80 players, average the numbers, and the classic bell curve emerges.

the data (i might upload a graph of it myself later, but bud didnt provide one)

Age 18 - 53.9 Age 29 - 81.2
Age 19 - 70.1 Age 30 - 78.6
Age 20 - 78.1 Age 31 - 75.7
Age 21 - 92.8 Age 32 - 69.8
Age 22 - 93.6 Age 33 - 72.9
Age 23 - 89.5 Age 34 - 62.9
Age 24 - 96.5 Age 35 - 64.0
Age 25 - 100.0 Age 36 - 59.2
Age 26 - 98.3 Age 37 - 66.9
Age 27 - 87.2 Age 38 - 89.1
Age 28 - 85.8 Age 39 - 43.9


so 21-26 is the peak zone. He doesnt really give that much detail about his selection process, so i have no clue what types of bias may be involved, and does mention that ages 18, 38, and 39 are skewed by a small sample size.

 

[barneyg]

so 21-26 is the peak zone. He doesnt really give that much detail about his selection process, so i have no clue what types of bias may be involved, and does mention that ages 18, 38, and 39 are skewed by a small sample size.

Thanks for sharing, this is interesting stuff. One tidbit from your last post:

took a bunch of active players who had played for at least six years in the NHL, and in some cases many more, and wrote down their points year by year. A player's best season got a score of 100, and his others were a percentage of that figure. So, if Dave Andreychuk scored 100 points at age 28 for a career high, age 28 received a score of 100. If he scored 50 points at age 22, that received a 50. Do that for about 80 players, average the numbers, and the classic bell curve emerges.

Given the results, there's obviously a 2nd standardization to 100 with the average numbers, as there's no way _every_ player's best season was at 25.

I think I'd like to see it comparing NHL to NHL, IHL to IHL and AHL to AHL better than this hybrid. I think it may skew the results by trying to squeeze in IHL & AHL stats.

IMO, using a good minor-league adjustment factor and including those years is better than ignoring them. If you only use NHL numbers, you get selection bias because your curve will only include players who got in the league early, i.e. the best players. That's why the graphs for the top 100 goal scorers and the 500-goal scorers suggest that players reach their peaks at 23, while the BTN graph (in the OP) has a peak at 25. I think stars peak earlier than 2nd and 3rd line players, and incidentally that's the type of graph I'd like to see.

 

[danishh]

Thanks for sharing, this is interesting stuff. One tidbit from your last post:



Given the results, there's obviously a 2nd standardization to 100 with the average numbers, as there's no way _every_ player's best season was at 25.



IMO, using a good minor-league adjustment factor and including those years is better than ignoring them. If you only use NHL numbers, you get selection bias because your curve will only include players who got in the league early, i.e. the best players. That's why the graphs for the top 100 goal scorers and the 500-goal scorers suggest that players reach their peaks at 23, while the BTN graph (in the OP) has a peak at 25. I think stars peak earlier than 2nd and 3rd line players, and incidentally that's the type of graph I'd like to see.

i saw a graph somewhere that put league years vs point performance. Maybe that's a factor that actually has more relevance than age. I'll try to dig it up at some point and post it here.

 

[DL44]

I did a similar analysis after the 2010-2011 season for the Post-lockout era.

The Chart may not accurately represent what a player's peak performance is post-lockout (since many haven't played out their curves... but it does show at what ages so far have been the most productive..

But you see what we've seen with the previous other charts in this thread...

Here's some post-lockout numbers...
-The newest era shows a little differently than past eras under similar analysis.





In list form:

19 36.0
20 40.1
21 46.3
22 54.9
23 58.1
24 58.1
25 63.3
26 62.3
27 62.4
28 60.6
29 58.7
30 57.7
31 57.6
32 56.8
33 55.4
34 51.6
35 47.5
36 42.2


- Performance starts a slight decline after the age of 27, and then takes a good downturn after the age of 32.
- Player's Prime yrs appears to be 24.5-28. (NHLPA is aware of this an was the prime reason for pushing the FA age down to 27.. ability to cash in on their best work instead of at 32 when a decline starts setting into the stats.)


Took numbers from only post lockout seasons of the following 79 players and categorized according to their age around the start of the campaign...


AMacdonaldAlfredArnottBRichardsBertuzziBoyesBriereBrownCarterCheechooCleary
CloweColeComrieDatsyukDoanDruryDSedinEStaalEliasErikssonFranzenGaborikGagne
GetzlafGiontaGomezGraboskiHavlatHeatleyHejdukHorcoffHossaHuseliusIginlaJThornton
JokinenKaryiaKeslerMkoivuSKoivuKopitarKovalchukKunitzLacavalierLaddLangkow
LegwandMalhotraMaloneMarleauMorrisonMorrowMrichardsMSLNaslundNylanderPalevski
PerryPlekanecPonikarovskyProspalRibeiroRolstonRyderSamsonovSamuelssonSavard
SedinSharpSmythSpezzaStastnyStillmanTanguayTruutuUmbergerVanekZetterberg




Some notes -
- a limited sample size relative to the total number of players - since i focused on more top 6 or former top type players.
- limited selection of seasons - 6
- favored selecting scoring forwards for the most part... (altho that may be seen as a pro for what i was trying trying to see here) why add players whose primary roles aren't scoring right.
_ Many of the players included are young and haven't hit their peaks, or have played out their curves - so it may have something to do with the earlier peak represented..

So the Chart may not accurately represent what a player's peak performance is post-lockout... but it does show at what ages so far have been the most productive - hopefully that makes sense.

 

[DL44]

I updated the above chart with the listed players 2011-2012 seasons ^^^^

http://i45.tinypic.com/2w6bh20.jpg

from lockout to lockout:

from lockout to the end of 2010-2011:

EDIT: Putting them on the same scales would of been nice... sorry about that...

 

[Ginu]

It would be interesting to use a rolling window approach in 3-5 year intervals to see how the age for top production declines over the years. I believe that players are reaching their peak earlier now than in the past. I'd like to see a graph of average age (of points leaders) by year to see if there's a downward trend.

 

[charliolemieux]

So this proves that by their late 20"s when players are considerend in their "prime" they have started to pay less focus on offense and started to play defense.

Is this not the argument epitiomed by Yzerman? Sacrificing pure offensive talent to play a more complete 2-way game? And therefore enter the Prime(most effective part) of their carreer.

 

[DL44]

So this proves that by their late 20"s when players are considerend in their "prime" they have started to pay less focus on offense and started to play defense.

Is this not the argument epitiomed by Yzerman? Sacrificing pure offensive talent to play a more complete 2-way game? And therefore enter the Prime(most effective part) of their carreer.

Exactly... prime scoring doesn't exactly represent the actually prime performance of a player.

The development of IQ, and the effect of experience continues and usually results in a superior player than the higher scoring earlier versions of themselves.

scoring prime of a player based on all the data: 24-29
prime player performance of forwards, my pure estimate: 26-32

Later for Dmen: 27-34/35
By about 28, in general what you see in a player is basically what you get in terms of production.. but development always continues till the body really begins to breakdown - which is unique for each player.

 

[tarheelhockey]

None of the studies above separate defensemen from forwards -- that seems like a pretty hefty flaw in the methodology, given the well known difference in their rates of development.

 

[MadLuke]

As a general trend, yes, but not as a predictor; classic counter-example would be Daniel Alfredsson, whose peak in his career, points-wise, was from ages 30-35.

Why a strong trend would not make a good predictor, do not get the counter example, this would be a good counter-example if someone have said all player peak between 21-26.......

 

[danishh]

So this proves that by their late 20"s when players are considerend in their "prime" they have started to pay less focus on offense and started to play defense.

Is this not the argument epitiomed by Yzerman? Sacrificing pure offensive talent to play a more complete 2-way game? And therefore enter the Prime(most effective part) of their carreer.

it proves that players tend to peak in offensive production at the ages shown. Anything else is not supported by any evidence (unless you'd like to offer some?)

None of the studies above separate defensemen from forwards -- that seems like a pretty hefty flaw in the methodology, given the well known difference in their rates of development.

i would agree with this. The main reason i posted this was to see if anyone had seen or would do themselves any further analysis on the topic. Positional analysis is certainly one of those areas that could benefit from further study.

 

[DL44]

None of the studies above separate defensemen from forwards -- that seems like a pretty hefty flaw in the methodology, given the well known difference in their rates of development.

I used only forwards in post #17.

I did another one of a smaller sample size using only players that had hit 70 pts in their careers.. i.e. high end players.
Maybe i'll dig that up if i can.

But the 79 players i did use are listed in the post above.


I did a dman graph as well... but my sample size was too small to come up with any sort of a decent graph... it looked like a roller coaster.