APBRmetrics

Revisited the data at http://stats-for-the-nba.appspot.com/. From 2001 to 2011, there are 27 players (out of 36 total) that ever appeared in a yearly top ten for which one can also identify the career max. value of RAPM. The average age at which this was achieved was 27.

This is up to two years later than ASPM estimates that I recall. Whether or not this result obtains to the general NBA population or whether or not it would be substantively pushed back were aging curves added in are interesting questions.

From what I've seen, according to box score stats player peak at ages 27 or 28. Some types of players peak earlier, like fast point guards or undersized power forwards, while others peak later, like big centers who normally peak just a little later or skill-dominant players like Nash.

The reason why players even peak goes back to how a curve is even constructed: for a nice old upside-down parabola (so there's a peak) you need two competing variables. As athleticism wanes with age, skill (and experience) increase. If your value as a player comes from your skill, like a shooter/passer, then that factor will counteract the failing athleticism and you'll peak later. Certain stats also peak earlier with players, like offensive rebounding, while others (free-throw percentage?) peak later (or does it always increase?)

What I mean by all that is, players don't peak at the same time, so forcing them all to the average peak, like 27 or 28, is not an ideal solution. I think that's partly why aging curves weren't incorporated into models.

Unbeknownst to many, Tony Parker has a career history of large dropoffs in the playoffs.Unweighted by minutes, his annual WS/48 in playoffs have been about 62% of his RS. His eW/484 have been 83% of RS, on avg. Both are about twice the expected dropoff from RS to PO.

AcrossTheCourt wrote:From what I've seen, according to box score stats player peak at ages 27 or 28. Some types of players peak earlier, like fast point guards or undersized power forwards, while others peak later, like big centers who normally peak just a little later or skill-dominant players like Nash.

The reason why players even peak goes back to how a curve is even constructed: for a nice old upside-down parabola (so there's a peak) you need two competing variables. As athleticism wanes with age, skill (and experience) increase. If your value as a player comes from your skill, like a shooter/passer, then that factor will counteract the failing athleticism and you'll peak later. Certain stats also peak earlier with players, like offensive rebounding, while others (free-throw percentage?) peak later (or does it always increase?)

What I mean by all that is, players don't peak at the same time, so forcing them all to the average peak, like 27 or 28, is not an ideal solution. I think that's partly why aging curves weren't incorporated into models.

This is why my player future projections are directly tied to statistical skillsets. Some skillsets decline faster/earlier than others. Lower usage/good rebounding/lower defensive rate guys have a career curve like a tiny molehill - and often a later "peak" that doesn't peak much, with often a longer history of "hanging around".

The type of player the guy is will obvious influence his perceived "peak". Also - I think NBA mileage factors in too - two 28 year olds with very similar skills and production - normally the one whose rookie year was at age 23 and has a bunch less career NBA minutes than the other has a better chance of not being in decline - as opposed to the other whose rookie season was age 19.

So, yes, I completely understand the difficulty in getting aging curves incorporated into the models. Beating.