Outlier seasons v. real improvement

[penbeast0]

I am curious whether anyone has ever done a study of whether players who have a new peak statistical season after they have been in the league 3 years (or after their 22nd birthday) are more likely to slide back down or are more likely to maintain the new level of production.

[tsherkin]

I am curious whether anyone has ever done a study of whether players who have a new peak statistical season after they have been in the league 3 years (or after their 22nd birthday) are more likely to slide back down or are more likely to maintain the new level of production.

I imagine it depends on the how. And are we limiting ourselves to 4th seasons, or just after 3 years in general?

Also, team context matters, but I guess that gets back to "how."

[penbeast0]

I'm looking for a statistical model for normalcy. The experience/age limit is because most players improve statistically in years 2 and 3, I'm looking for later on in the career when someone has a big year . . . is it more likely they regress to the mean and if so by how much so we can use it as a predictor when someone has that one big year after he's established his level in the league. Will it usually be a new level or an outlier where the player returns to what we expect.

As a Wiz example, Don McLean came here as a Kuzma type with less playmaking and had one year where he was one of the more efficient scorers in the league. Then went back to being pretty much useless for the rest of his career.

In terms of volume, Michael Adams made a massive jump in scoring/minute in his 6th year in the league going from a guy who would get you 12 to 18 pp36 to over 26 pp36, then regressed to 18 then down to 12-15 pp36 for the rest of his career.

Is that what we should expect?

[tsherkin]

I'm looking for a statistical model for normalcy. The experience/age limit is because most players improve statistically in years 2 and 3, I'm looking for later on in the career when someone has a big year . . . is it more likely they regress to the mean and if so by how much so we can use it as a predictor when someone has that one big year after he's established his level in the league. Will it usually be a new level or an outlier where the player returns to what we expect.

As a Wiz example, Don McLean came here as a Kuzma type with less playmaking and had one year where he was one of the more efficient scorers in the league. Then went back to being pretty much useless for the rest of his career.

In terms of volume, Michael Adams made a massive jump in scoring/minute in his 6th year in the league going from a guy who would get you 12 to 18 pp36 to over 26 pp36, then regressed to 18 then down to 12-15 pp36 for the rest of his career.

Is that what we should expect?

Lots and lots of variables, though.

Draft slot / expectations? Are they getting lots of run? Is the team bad? Are they good? Is there an extant start the offense is running through? What position/role do they fill? Perimeter vs. big? What style of ball do they play? Age? Prior experience? AAU crap-ball or pro league? Athleticism- or skill-based, or both? Shooter or no?

What specific adjustments does the team make to get the guy rolling, if any?

Does the performance come from unexpected and likely unsustainable shooting from a specific zone?

Like, you can draw out a graph with enough data, but there are still specifics which go into it that will ultimately matter.

[One_and_Done]

AD hitting 3s in 2020 never stuck. Whatever year Steve Francis supposedly broke out would also be a strong candidate.

[penbeast0]

Not sure what you are looking at with Stevie Franchise. I went back to check the stats and they look like I remember them. His rookie year he was at 17.9 pp36 and 4 of the next 5 he was between that mark and 20.1 pp36. Efficiency he started at .543 in his rookie year and ended his 6 year run in Houston at .533 ts%. Not sure you understand what I'm looking for.

To go the other way, Steve Nash after 4 years in the league was averaging about 11pp36, then pumped it to between 15 and 20 for the next 11 years. After 3 years in the league his efficiency was only averaging about .520 ts% then he pumped it to around .600 for the next 14 years. That's what I think of as a sustained bump.

Havlicek was pretty consistent. His second year to his 12th he kept his pp36 between 20 and 24ppg, but in 1970 (the year Russell retired), he kicked his efficiency from an anemic .460 ts% or so to over .500 ts% from age 29 on. Even his last two years in the league he was still at .495ts%. That's a sustained improvement rather than an outlier season.

We can look at reasons, and they are there . . . for example, Havlicek started playing more as an on ball player rather than an off ball one (plus in my opinion, expansion, increased drug use, and NBA/ABA league jumping led to more focus on scoring, less on defense, less team unity, and weaker average team strength). That's why I was asking if any of our many great statheads has looked at or seen a study that looks at this question over a broad spectrum of players.

[LukaTheGOAT]

I haven't personally done any studies, but somethings you may want to check out:

Here is one study:
https://digitalcommons.bryant.edu/cgi/viewcontent.cgi?article=1223&context=eeb#:~:text=Player%20performance%20statistics%20data%20was,after%20the%20age%20of%2032.

Spoiler:

This research paper examines the peak performance age of NBA players from the perspective of General Managers. It explores the age range where players are most effective and analyzes the relationship between age and performance using ESPN player statistics. The study concludes that NBA players typically perform their best between ages 27 and 31, with performance slightly declining after 32.


Another study based on PER and MP:
https://www.linkedin.com/pulse/analysing-predicting-peak-age-nba-players-data-science-marcus-chua/

The full dataset can be accessed as a CSV file on Github here: https://github.com/mchj2468/nbapeakageproject

There is a blurb here about NBA guys peaking at 27-28:

https://sites.dartmouth.edu/sportsanalytics/2021/11/10/peak-age-in-sports/


Another says they improve until 25-27:

https://harvardsportsanalysis.org/2015/05/player-progression-in-the-nba/

Forum post RAPM Aging Curve done by Jeremias Englemann from back in 2013:

https://apbr.org/metrics/viewtopic.php?f=2&t=8308

Englemann is pretty active on twitter, and he might actually have some helpful info to answer your question.

This one has more of a focus on NBA stars and therefore is perhaps not representative of larger league trends. Nonetheless, it suggests 27 is a good age, which is has been a consistent them in everything I've linked:

https://hoopshype.com/2018/12/31/nba-aging-curve-father-time-prime-lebron-james-decline/


Another study using box-score numbers:

https://courses.cs.washington.edu/courses/cse163/20su/files/project/archive/nba.pdf

Per this study:
https://www.mdpi.com/2078-2489/15/4/242

"Figure A3 shows that players’ salaries generally peak in their late twenties to early thirties (29–30 years old), while PER tends to peak slightly earlier (27–29). Salaries and performance metrics, such as PER, VORP, net rating, etc. vary with age, highlighting the relationship between a player’s career stage, their role, and their value to the team."


This person on Reddit from 11 years ago did a study:
https://www.reddit.com/r/nba/comments/2aa0r2/study_at_what_age_an_nba_player_reaches_his_peak/?rdt=54128

SPOILER:
The average peak age is 27 years and 2 months

In general, it would seem players reach new heights at 25 or after, and while perhaps these articles do not explicitly say that they are less likely to drop in immediately after, my hunch is that so many guys peaks after this age suggests they likely hold onto this performance for a bit of time.

[penbeast0]

Thank you Luka. I appreciate the help.

[tsherkin]

Thank you Luka. I appreciate the help.

Second this, good reading!

[1993Playoffs]

It gets tricky because outlier can also mean “asked to do more in this particular season “

[tsherkin]

It gets tricky because outlier can also mean “asked to do more in this particular season “

Indeed. Another thought which just struck me is that a lot of that stuff is a decade old, and the research which went into it includes many, many seasons where players didn't even begin their careers until 22.

I wonder what it looks like for guys who started 19 and younger?

EDIT: As in, does the age matter more, the role the team is asking of them, or the season of their career/total experience?