The general assumption has been that NBA players peak at about age 27, along the lines of the peak age Bill James found in baseball many years ago. While the Journal strikes an ominous tone about another James, LeBron (due to turn 25 next week) having already peaked, Berri points out on his blog that the matter is not quite so cut and dry.
the key issue is not the specific point in the player’s 20s where the peak occurs, but rather that performance after age 30 has a noticeable drop-off. In the player’s twenties the slope downward is quite gradual (and not something you would probably notice if you watched the player). In other words, LeBron will still be LeBron – barring injury – for a few more years.
Aging studies depend heavily upon the method used, which won’t be known until the follow-up to Wages of Wins is released next spring. These methods are a source of some debate in baseball circles; fewer attempts have been made in the NBA. I’ve never really done much of a comprehensive study, but my SCHOENE database allows a quick and dirty method similar to the one used by Mitchel Lichtman for the Hardball Times. It shows a peak age of 27.1, which is somewhat later than Berri found. [One technical note: SCHOENE uses three-year weighted averages adjusted for age rather than simply the previous season, so some of the noise has been removed.]
However, the bigger takeaway is a peak age range rather than a single number. Here are the percentage of players in my database (minimum 250 minutes played in consecutive seasons from 1979-80 through last year) who improved the following season, based on their age in the current year:
Despite fairly large samples, the percentages at age 25 and age 26 are not high enough above 50 percent to be statistically certain that players to tend to improve on average at those seasons. So peak age could be anywhere from 25 through 27 by this method.
Two other notes: First, SCHOENE is built on the assumption that there is no such thing as a single aging curve, but instead unique ones for different types of players. Second, we do know that various stats have different aging curves. For example, I discussed not long ago on Unfiltered the tendency for rebounding to peak early. Since Berri’s Win Score method values rebounding especially heavily, this could explain part of why he finds a peak age slightly earlier than the one I find.