Is peak at age 29?

J.C. Bradbury:

Though old players may not be what they once were, the evidence indicates they can still be valuable. According to my estimates, a hitter who has a .900 OPS at his peak would be expected to post around an .850 OPS at 35; a pitcher with a peak 3.5 ERA is expected to post around a 3.75 ERA at 35. Yes, age saps athletic skill, but the stock of skill being diminished is also important.

Is this true?

Let's look at Bradbury's original study (unless you subscribe, it costs about $30 to download). He says:

The results indicate that both hitters and pitchers peak around age 29. This is older than some estimates of peak performance, but it is in-line with the general understanding of human physiological function. ... For [linear weights],which is not normalized for playing time, players peak at 29.41 years. This is similar to the peak-age estimate of 29.13 for OPS, which is normalized for playing time.

How can we test this theory easily? Well, if peak age is 29, we should see players at age 29 have a higher OPS than they did at an earlier age, right? Some may (rightly) note there is a selection affect in who gets to play in successive seasons, and due to regression to the mean we should expect to see players decline from one season to the next - players who underperform are less likely to play in a successive season than those who overperform.

So let's take those players who were 29 in 2008 and compare them to how they did in 2006 at age 27. Thirty-five position players were 29 in 2008 (based upon age on July of that season, the same age criteria used in Bradbury's study). Of them, 27 also played in 2006. What we see does not seem to support Bradbury's conclusion - on average, players lost roughly .014 points of OPS from 2006 to 2008. That's not what we should expect to see if the average peak age is in fact 29.

Something that should also be noted - 43 players were age 27 in 2006. So a little over half of those players stuck in the league to age 29. (Of course, some of those players could have missed a season due to causes like injury or a demotion and could have returned in this past season - I don't have 2009 stats in my database yet.) We're looking at the survivors.

But that's just a one-year sample, right? What if we looked at all age 29 seasons from 1997 to 2008? (As for why 1997, I shall explain shortly.) We see a similar (but less pronounced) dropoff, of about .006 points of OPS.

And of course baseball isn't all hitting. The Baseball Databank has outfield positions broken down from 1995 on (and 1995 plus two years is 1997 - see how that works?)

What we see as that as players grow older, they move off the more important defensive positions (catcher, shortstop, center field) and move to the corners. Thinking of peak only in terms of hitting is not instructive if we want to figure a player's free agent value - defensive value needs to be considered as well.

Bradbury's model may do a good job of explaining in retrospect where peak age is for players who make it to age 35. But it does not seem to have a lot of predictive value for what will happen to a 27-year-old player going forward. Anyone looking to predict future performance should look to some other method of estimating the effects of age.

UPDATE: Here's another way of looking at the issue. Bradbury's study included all batters who debuted past 1920 and played 10 seasons between age 24 and 35. Now, looking at all players who debuted past 1920, 48% were out of the league before the age of 29, Bradbury's purported peak. Does that make sense to anyone? Or flip it around - the most common debut age for a baseball player is 23, and the average player who debuts at that age will play six years, or until age 29. In point of fact, if we look at the most common debut ages:

So, for player that debut from age 21 to 26 (that's 75% of players), the average career length takes them through their age 29 season. Either MLB teams are all highly irrational, discarding players the second they've reached their peak (when given the aging curve Bradbury proposes, a typical player should still have several productive seasons post-peak) or Bradbury's study is simply wrong.