NHL Age Curves and xG/60 Impact

[LGR4GM]

xG/60 stands for expected goals per 60min of ice time. It is a metric that tries to take into account where goals are scored and how frequently to help filter out some of the noise with the randomness of hockey (if you shoot higher than average you will outproduce your xGF like Thompson has and if you shoot below average you see the inverse... Cozens). Anyways what does that have to do with this? Well here is the average age curve for NHL production. This is important when thinking about timelines of players and where they are in their careers. NOTE! you can absolutely outperform the curve but it shows us an average and most players will fall within 1 standard deviation of that average. Stop assuming every player is a special snowflake, Tage Thompson is an exception, Mitts is probably the rule. Anyways, here is the average for offense and defense. If you are really curious, there is a good discussion in this thread by Micah on this. 

 

Edited November 28, 2022 by LGR4GM

[tom webster]

I would like to see these lines overlayed with actual goals scored. Is that available?

[LGR4GM]

This point is very important. It isn't that good players defy the curve per say but that they are coming from such a higher peak they still positively contribute.

 

12 minutes ago, tom webster said:

I would like to see these lines overlayed with actual goals scored. Is that available?

Idk if he does that but xGF models try to fit goal scoring curves so they should overlay closely. 

[MattPie]

I'm curious if that bump after 37 on offense is that all the average players retire leaving only the aging stars who were beating the curve all along.

[Marvin]

I wonder what the adjusted +/- would look like.  For instance, Alexander Mogilny had his best scoring season in Buffalo.  But in Vancouver and New Jersey, he became a reliable back-checker although he was scoring progressively less.

[bob_sauve28]

Granato said the other day the league is getting younger. Is that true or not? 

[Brawndo]

What's cool is Micah references a research paper written in part by Michael Lopez, who is considered one of the leading minds in Sports Analytics and is employed by the NFL as Their Analytics Guru. Two of His main proteges are Dennis Lock(employed by the Bills) and Sam Ventura. 

[Curt]

I don’t know.  I’ve seen charts like this before, but they just never seem to make sense.

Players at age 18 produce more than players at age 27?

Players at age 20 produce more than players at age 26?

Players at age 40 produce more than players at age 35?

 

[Weave]

3 minutes ago, Curt said:

I don’t know.  I’ve seen charts like this before, but they just never seem to make sense.

Players at age 18 produce more than players at age 27?

Players at age 20 produce more than players at age 26?

Players at age 40 produce more than players at age 35?

 

I guess the point about players at 18 producing more than players at age 26 is due to the very small number of exceptionally skilled players that play at 18 have better average stats than the much larger population of players at average much lower skill at age 26.

More of the same at 20, although the number of players at 20 is higher.

Reverse of the same at 40 over 35 as theorized in a previous post.  I mean, Jagr and who in recent memory.  Thornton?  Really small but highly skilled sample size at 40.

[Curt]

20 minutes ago, Weave said:

I guess the point about players at 18 producing more than players at age 26 is due to the very small number of exceptionally skilled players that play at 18 have better average stats than the much larger population of players at average much lower skill at age 26.

More of the same at 20, although the number of players at 20 is higher.

Reverse of the same at 40 over 35 as theorized in a previous post.  I mean, Jagr and who in recent memory.  Thornton?  Really small but highly skilled sample size at 40.

The thing is, the chart isn’t tracking raw totals.  It’s tracking amount of change in production.

[Weave]

18 minutes ago, Curt said:

The thing is, the chart isn’t tracking raw totals.  It’s tracking amount of change in production.

Thats what I get for not looking at the Y axis closely.

Interesting that at all age groups forwards never, on average, have a positive change.

[matter2003]

I suspect if you showed the G/60 scored it would peak maybe 2 years further to the right as players learn to "do more with less" as they gain experience from playing. Just my guess, would be interesting to see if I am right or not. 

Goals are what count, not xG. I don't care if a player has 10 xG and only scores 1 actual goal versus a player that has 7 xG but scores 3 actual goals. I'm taking the actual goal scorer 10 times out of 10, which is why without an overlay showing actual goals this is pretty meaningless.

Edited November 28, 2022 by matter2003

[Pimlach]

[matter2003]

I can't really buy into this... basically it says a player peaks at scoring at age 20, declines til age 35 but then somehow starts going back up at age 36 until age 40?

 

Edited November 28, 2022 by matter2003

[mjd1001]

A different (slightly) approach to the same idea.... https://hockey-graphs.com/2017/03/23/a-new-look-at-aging-curves-for-nhl-skaters-part-1/

Edited November 28, 2022 by mjd1001

[Curt]

22 minutes ago, Weave said:

Thats what I get for not looking at the Y axis closely.

Interesting that at all age groups forwards never, on average, have a positive change.

Honestly, after looking at it a 2nd time, I understand it less.

Regardless, I can see that it’s heavily skewed at each end by the early entry, and longevity of elite players.  To me, that makes it harder to say with any confidence that it’s correctly identifying the peak age range.

[matter2003]

10 minutes ago, matter2003 said:

I can't really buy into this... basically it says a player peaks at scoring at age 20, declines til age 35 but then somehow starts going back up at age 36 until age 40?

 

Just looking at the top 10 goalscorers from last year this seems flawed. 2 players had their best season at age 30 or 31, 2 at age 27, 3 or 4 at age 25 and the only one at age 20 or 21 was Ovechkin. 

[Thorner]

6 hours ago, LGR4GM said:

xG/60 stands for expected goals per 60min of ice time. It is a metric that tries to take into account where goals are scored and how frequently to help filter out some of the noise with the randomness of hockey (if you shoot higher than average you will outproduce your xGF like Thompson has and if you shoot below average you see the inverse... Cozens). Anyways what does that have to do with this? Well here is the average age curve for NHL production. This is important when thinking about timelines of players and where they are in their careers. NOTE! you can absolutely outperform the curve but it shows us an average and most players will fall within 1 standard deviation of that average. Stop assuming every player is a special snowflake, Tage Thompson is an exception, Mitts is probably the rule. Anyways, here is the average for offense and defense. If you are really curious, there is a good discussion in this thread by Micah on this. 

 

Very interesting and definitely lines up with the word nowadays that on average, offensive prime output comes at a much earlier age than often thought. 

Win when your guys are young 

[Curt]

4 minutes ago, Thorny said:

Very interesting and definitely lines up with the word nowadays that on average, offensive prime output comes at a much earlier age than often thought. 

Win when your guys are young 

It’s hard or impossible to quantify, but I’m also not positive that “prime offensive output” exactly equals “prime hockey player”.  There is also the defensive facet of the game.

[Thorner]

1 minute ago, Curt said:

It’s hard or impossible to quantify, but I’m also not positive that “prime offensive output” exactly equals “prime hockey player”.  There is also the defensive facet of the game.

That’s why I said prime offensive output lol

[Thorner]

The game is obviously leaning younger over time so it’s important to remember things like, just because Tage is putting up a 40 goal pace now at 25 doesn’t necessarily mean he’ll be putting it up in 3 years once Levi is finally ready. Also fits the idea it’s better to pay early rather than for past performance. 

Of course plenty round out their full games to a greater degree at a later age, hopefully those are the guys who could be available at a slight shave of cost do to declining raw production 

The very best of the best I think are in their primes a bit later. Crosby became a truly elite, top 10 worthy all time player later in his career when the production technically wasn’t as high. But your average guy peaks a bit earlier. 

[dudacek]

40 of the NHL's top 50 scorers right now were drafted before 2015.

Four of the top 50 are 22 or less.

Not sure those numbers are saying what people think they are.

Edited November 28, 2022 by dudacek

[Thorner]

4 minutes ago, dudacek said:

40 of the NHL's top 50 scorers right now were drafted before 2015.

Four of the top 50 are 22 or less.

Not sure those numbers are saying what people think they are.

The chart is saying that prime offensive output is 23

By expected goals for 

It’s not that confusing of a chart

[Thorner]

6 hours ago, LGR4GM said:

xG/60 stands for expected goals per 60min of ice time. It is a metric that tries to take into account where goals are scored and how frequently to help filter out some of the noise with the randomness of hockey (if you shoot higher than average you will outproduce your xGF like Thompson has and if you shoot below average you see the inverse... Cozens). Anyways what does that have to do with this? Well here is the average age curve for NHL production. This is important when thinking about timelines of players and where they are in their careers. NOTE! you can absolutely outperform the curve but it shows us an average and most players will fall within 1 standard deviation of that average. Stop assuming every player is a special snowflake, Tage Thompson is an exception, Mitts is probably the rule. Anyways, here is the average for offense and defense. If you are really curious, there is a good discussion in this thread by Micah on this. 

 

Any idea if the curves change when looking at expected goals FOR and not just expected goals? The current chart is using just individual expected, correct? 

And it also includes PP mins in the per 60, correct? 

Edit: actually, it says “impact on” so perhaps this is representative of expected goals for and not just individual expected goals. I feel like sometimes people are a bit loosey goosey with the terminology which leads to confusion 

Edited November 28, 2022 by Thorny

[mjd1001]

Ok, another take on this....

I looked at POINTS per game (not goals, not per 60, but points per game)

I looked at some of the top players over the past 10-20 years and wanted to see where they had their peak years in terms of where they ranked.  Ranking is important to me because some years have more scoring than others, so where a player ranks is more important than their raw numbers.

I picked some random players that were in the top 10 in scoring often and were over 30.

The result summary:  Peak offensive production does appear to be between the ages of 25-27.

The players I looked at:

Crosby:  His peak in terms of PPG ranking was between 19-27 years of age

Malkin:  Between 22-26

Ovi:  21-26

P. Kane  23-27

St. Louis:  23-29

Thornton:  23-27

Giroux: 22-24

Tavares:  22-23

Taylor Hall 20-21.

 

Some players had individual peak years later (Giroux at 28, St. Louis at 31, Kane at 29) but for the most part, the numbers above hold true for Star level players.  Year 26 seems to be the 'peak ranking' in terms of ppg for forwards, at least for the sample of the guys above I looked at.  For many of them, their 'peak' year in PPG also is close to the same years their team was at the peak (at least in terms of fighting for the top of the standings on a regular basis)

-Absolute peak year, averaging all their rankings out by age is 23 years of age for this group.

-Ages 22-26 are very close to peak year.

-Some Dropoff at age 27...staying level until age 29.

-Larger dropoff starting at age 30 and continuing each year to get worse.

-Production at age 20 is about equaly to age 29 or 30.

Edited November 28, 2022 by mjd1001