I'm guessing it would be better to look at this by age cohort. For example, it's true that there is a decline in the 31-35 contribution in the last few years ... but that predictably follows on a relative trough in the 26-30 a few years before and a substantial trough in 21-25 for nearly ten years. Similarly, there's a big peak in 21-25 in the early 1910s, followed by a peak in 26-30, followed by a peak in 31-35, followed by a peak in 36-40. That's a cohort of good players moving through.

There's an interesting 21-25 peak in the mid-60s that is not followed by a 26-30 or 31-35 peak. Originally I thought that might be all the young pitchers breaking down but this is just position players. So I'm guessing this is the speed (baserunning not greenies) factor.

There are so many other factors, though. One of them is salary. If pay is going up, guys retire later when they are still above replacement level (one would predict, at least), and when there's a plateau or downtick, guys retire earlier. I sense that there was such a plateau in the mid/late-2000s, and that would be worth considering.

True, you expect at least somewhat longer careers in the FA era. Players have more incentive to stick around and guys with guaranteed contracts will continue to play. Of course if they're kinda crap by that point then they're not adding to the WAR total.

As I hint, we probably expect more youth during speed eras and maybe more old guys during power eras. That does seem to be there for speed -- the youth spike through the 60s and 70s that I noted -- but not so much for the age-power. There are a couple of old peaks in the 20s/30s and the early 50s when it was a good power game. But the early 50s thing pretty clearly seems tied up with the aftermath of WW2 (and maybe Korea) and there's no old peak in the late 50s and early 60s.

The youth drought of the 90s is interesting and, if you want to tie to roids, may coincide with the intro of minor-league testing (I'm not sure what year that was). Possibly the prospects weren't using until they got to the majors so teams stuck with the older guys. Still, the 31-35 peak and the later 36-40 peak in the "late roids era" aren't out of line with peaks seen in other eras.

I've gotta say that graph is really hard to read without some 5-year tick marks.

Might there have been a devaluation of defense in the '90s and a revaluation in the 10s that would lead to more older players then and fewer now? I feel like there are a lot of players in their late thirties who can still hit reasonably well, but no one will give them jobs because they can't play defense (Abreu, Guerrero, Jermaine Dye a couple of years ago).

#4 asks a pretty good question, I've wondered that myself.

Might there have been a devaluation of defense in the '90s

Similar to my suggestion that speed = youth. But yeah, we've seen new defensive metrics come out and maybe that has had an impact. The shortening of benches has also made it harder to carry a bat-only old guy or an aging RHB as a platoon guy (benches were short in the 90s and early 00s too, I'm thinking longer-term). It would make sense that a pitcher added is an old hitter gone (obviously not a perfect relationship). I'd like to see the same chart for pitcher WAR ... with some magical adjustment for the shift in starter usage.

Still, the 31-35 peak and the later 36-40 peak in the "late roids era" aren't out of line with peaks seen in other eras.

But are those peaks what you would have expected from those (e.g. players aged 21-25 in 1990, or in 1995) cohorts given their priors, eg their age 21-25 and age 26-30 shares of annual position player WAR relative to that of other cohorts? Seems doubtful.