NHL players' peaks

[Bear of Bad News]

http://www.cbc.ca/news/when-nhl-players-peak-hockey-metrics-1.2646054

Short summary: National Hockey League forwards reach their peak scoring performance at age 28 and defencemen peak at age 29, while goaltenders show little change in performance based on age.

I disagree with the conclusion on goaltenders; based on what I've done so far, they have a similar peak period.

Click to read more...

 

[soireeculturelle]

strange findings, considering Eric T. talks about both forwards and defenseman peaking (points, ES time on ice and corsi) in their mid-20s. he also found that goalies drop off rather sharply, in general, past age 30.

"Brander also notes that "forwards who do not develop into consistent scorers by age 23 or 24 in most cases never will." - this seems consistent with past research, though.

As for the playoffs, I looked into it last week and it seems that 14 of the last 30 Conn Smythe winners were aged 25 or younger when they won it. In that sense young legs may have an edge deep into the playoffs.

 

[rosemount289]

To me......................???

To me it's 10 years in the NHL..........then the player starts to decline.........all the great players do not have better years after 10 NHL years......you can see that in Rick Nash...........who's played almost 11 years and is starting to decline.

 

[Bear of Bad News]

Most studies I've seen that attempt to estimate a player's age curve have the problem of censored data. Suppose that we develop a "perfect" measure of a player's overall value in a season (easier said than done, but bear with me).

At some point in a typical player's career, they're going to cross below the threshold where they have no more NHL value, and they'll be in the minor leagues.

The result is that when people attempt to measure the curve, they do one of two things - the first is to just take the average "value" of players at age N. That distorts results, since it implicitly assumes that the typical 40 year is the typical Selanne/Jagr, and we know that's not the case, but if you're not an exceptional player, you don't play to the age of 40 in the NHL.

The second is to measure on a player-level basis the drop between year N and year N+1. This is a better way to approach the problem, but you still run into the problem of truncated endpoints (for players who drop below the threshold between year N and year N+1) and censored data (anyone good enough to play in the NHL can probably play *somewhere* for as long as they'd like to, but we don't know what Andrew Peters would do between age 35 and age 36 because he's not given the opportunity).

 

[soireeculturelle]

Most studies I've seen that attempt to estimate a player's age curve have the problem of censored data. Suppose that we develop a "perfect" measure of a player's overall value in a season (easier said than done, but bear with me).

At some point in a typical player's career, they're going to cross below the threshold where they have no more NHL value, and they'll be in the minor leagues.

The result is that when people attempt to measure the curve, they do one of two things - the first is to just take the average "value" of players at age N. That distorts results, since it implicitly assumes that the typical 40 year is the typical Selanne/Jagr, and we know that's not the case, but if you're not an exceptional player, you don't play to the age of 40 in the NHL.

The second is to measure on a player-level basis the drop between year N and year N+1. This is a better way to approach the problem, but you still run into the problem of truncated endpoints (for players who drop below the threshold between year N and year N+1) and censored data (anyone good enough to play in the NHL can probably play *somewhere* for as long as they'd like to, but we don't know what Andrew Peters would do between age 35 and age 36 because he's not given the opportunity).

seems possible to adjust for survivor ship bias (the phenomenon you're describing)
http://nhlnumbers.com/2012/12/6/goalie-aging-survivorship-bias

perhaps this is where the analysis in the original article falls short, regarding goalies.

 

[Bear of Bad News]

Interesting article - thanks for linking it!

I'll have to dig into it in depth.

 

[soireeculturelle]

Interesting article - thanks for linking it!

I'll have to dig into it in depth.

if you dont take survivorship bias into consideration, you'll most likely end up with the conclusions presented in the OP article.

marginal skaters tend to wash out of the NHL after age 24-27, and the effects on goalies are obvious as well. both of those things will push the actual peak age of players down toward the early-mid 20s as opposed to late 20s and 30s.

as an aside, a researcher from simon fraser university is doing research on professional video game players (starcraft 2 is a big deal in some part of the world - in korean if you're an eGamer, you get as much attention as NHLers here). the peak age in reaction time is around 24. the type of complex decision making with time constraints needed to do well in those types of games mirrors the thinking needed in hockey.

 

[LadyStanley]

Study: Do NHL players peak before age 28 or 29?

http://www.washingtonpost.com/news/...-peak-before-age-28-or-29/?wprss=rss_capitals

http://www.degruyter.com/view/j/jqas-ahead-of-print/jqas-2013-0085/jqas-2013-0085.xml

Focused on points/game (rather than points/minute).

New study released on March.

 

[NativeHockey77*]

Players usually have their best years from 22-26 from what I thought. I remember reading a study on career years being in that time frame usually (not all cases obviously) and all the former hockey players ive talked to said they felt their best 21-23 ish range

 

[TheStroker]

Players usually have their best years from 22-26 from what I thought. I remember reading a study on career years being in that time frame usually (not all cases obviously)

You could make a case that forwards have their best years then. Defensemen and goalies seem to peak later. I would argue that those 2 groups seem to peak in their mid/late 20s to their early 30s.

 

[Fred Taylor]

Most superstar players are usually at their best statistically between the ages of 21-26.

 

[Sonofamitch]

I'd say a player is at his best at 26-27 before he hits UFA and then it's all downhill from there.

 

[Mischa]

my thought was always 25-31 for forward and maybe 26-33 for defensemen

 

[Snotbubbles]

It depends on the type of scorer. Goal scorers peak at a younger age and have a shorter peak than a passer.

 

[Bear of Bad News]

It depends on the type of scorer. Goal scorers peak at a younger age and have a shorter peak than a passer.

I think that this is probably true (although the ultimate point of analyses are to replace impressions with data).

Ideally (and overall), it would be worth doing this for players of different "types".

 

[Vesa Awesaka]

Not sure if they peak but generally you know what type of player someone is going to be by 25-26. If someone isnt a scorer at any level by that time they probably arent going to break out.

 

[JimmyChan9191]

It's weird because a lot of defensemen peak in their late 20's/early 30's but for scoring forwards, it's generally early to mid twenties.

 

[PositiveCashFlow]

there's some goalies who are decent well into their 30's

 

[thom]

Age 23-29 are the peak yrs.

 

[Plural]

I'll read the paper as soon as I got time. But I grinned at the numbers. I have a very strong suspicion that the most elite offensive superstars peak earlier than 28.

 

[plusandminus]

This is quite easy to look up.
When I did it myself a couple of years ago, it showed that players peaked offensively peaked at age 26. I think the peak age for goal scoring was slightly lower than the peak age for assists.

This does not necessarily mean they are at their best near age 26. It may just mean that they lose some of the attributes (like explosiveness?) that make for good goal scoring, or that they simply go from playing a more "(goal) scoring oriented" game to playing more "overall responsible".

 

[Bear of Bad News]

This is quite easy to look up.

It's certainly not easy to look up, and I'd be curious why you think otherwise. What's your go-to source?

 

[mkwong268]

This is the histogram I get when looking for player's highest PPG season where they played at least 20 games. Of course this is pretty shallow analysis and it's not taking many things into consideration but I thought it might add something to the discussion.

EDIT: This is for forwards only.

 

[plusandminus]

It's certainly not easy to look up, and I'd be curious why you think otherwise. What's your go-to source?

Well, hockeyreference.com is one of the places. There you can see how much every player has scored each season, and it also lists age.
(Personally I had all the data - and much more - stored in a SQL Server database, and could do lots of different kinds of studies on it.)

I actually posted about this a couple of years ago (it's probably in the history section as this section didn't exist back then). If you're the one I think (different name now?), I think you read it and even commented upon it. The graph posted just prior to this post of mine shows a result similar to what was then posted.
So... Even though "only" 26 years old, Crosby might not peak further scoring wise. Same with Ovechkin (28 years old).

(While I'm writing, I still wish that statistical analysis could focus more upon getting knowledge about things that are easy to measure (like in this topic), and use that knowledge to build further knowledge on. Instead, much focus is being spent on trying to quantify quite difficult things that depends on many factors working together. For example, +/- is affected by lots of things, like zone starts, strength of teammates, strength of opponents, strength of own goaltending, etc. For +/-, a good approach would be to frist try to determine how much goaltending affects it. By doing that, one should for example look at save percentage and try to figure out how much that is affected by the skaters' defensive skill, etc. Even that is difficult. I think the key is to try to gain knowledge about bit by bit, and then use that knowledge to learn about other bits, and so on... For example, how much is a offensive zone start on average "worth"?
Sorry for getting off-topic.)

 

[Bear of Bad News]

I actually posted about this a couple of years ago (it's probably in the history section as this section didn't exist back then). If you're the one I think (different name now?), I think you read it and even commented upon it. The graph posted just prior to this post of mine shows a results similar to what was then posted.

I change the name from time to time (have to stay one step ahead of my creditors ), so it was probably me - I'll have to go back and look at it.

I agree - the raw data is there and pretty easy to get to (the goalies' information is on my site). I think that having the analysis "easy to get" is where we might disagree; it's very easy to do this stuff wrong. To do it right - dealing with truncated data, for instance - is not easy to get to.