He had his best years in the mid-to-late part of his career, and I expect I’m going to do the same thing.

At baseball-reference.com, Young’s numbers at his age compare with Barry Larkin, Craig Biggio and Julio Franco

Another example of the weakness of the comps system used there. Better to use PI, one of the best things at B-Ref. In terms of PA and OPS+ through age 22, Young has lots and lots of close comps, the closest ones being Don Money, Lonny Frey, Doug Rader, Edgardo Alfonzo, Denis Menke and Tony Cuccinello (all right around Young's current ~6000 PA and 105 OPS+, and all infielders). Really, that seems much closer to Young's league than Larkin, Biggio, or even Franco, who is a tier below them. Young's comps were solid starters and deserving All-Stars in their few best years, but nowhere near any kind of Hall.

The Hall of Famers that Young resembles are guys like Freddie Lindstrom and Travis Jackson, IOW the guys who probably shouldn't be in themselves. George Kell, no inner-circle guy himself, had a 112 OPS+ in 6700 PAs at Young's age; he was a distinctly better hitter than Young, though the years that Young spent as a bad shortstop might give him some positional edge over Kell as a good third baseman.

“One of my favorite players of all time is Paul Molitor,” Young said, “just because he was a great player, a great competitor, and he was respected for the way he played the game.

How much did Molitor kick and scream about his various mid-career position changes?

Michael Young is a big reason I am wary of fielding metrics. He was the Brad Hawpe/Adam Dunn of 2005, a very good offensive player whose horrific defense negated a huge portion of their bat's value, and who very plainly needed to move down the defensive spectrum (to DH in Dunn/Hawpe's case, to anywhere besides C in Young's case). But I was looking over his WAR page over at Rally's site, and here are his TZ listings for SS (2004-2008):

04: -22
05: -29
06: 5
07: -6
08: 9

That's a swing of ~3.5 WINS(!) just on defense between 05 and 06. That seems unfathomable to me. His UZR numbers don't shift quite so drastically (but still markedly: >20 run difference between 05 and 06), but even assuming that these numbers are accurate representations of the quality of his play during these years, they make me highly reticent to ascribe much year to year predictive value to them.

they make me highly reticent to ascribe much year to year predictive value to them.

I tend to agree with you but let me put on my MGL hat.

How do you feel about Young's BA? In 2001-2 he hit 249 and 262; in 2003 he hit 306 (and hasn't looked back and peaked at 331). For 2007-9, his BA has fluctuated from 315 to 284 to 322.

Or OPS+ which was around 80 in 2001-2 and jumped to 97 then 109 then 131. From 2007-9, his OPS+ has bounced from 106 to 95 to 128. In RC/27 terms, he went from 3.8 to 5.5 in 2002-3 and from 4.8 to 7.0 in 2008-9. If you equalized playing time, that 2008-9 difference would be about 3.5 to 4 wins.

MGL has said that the year-to-year correlation of UZR is in the same range as the year-to-year correlation of BA. That's not awesome (around .5 if I remember right) but it's useful. We've just gotten used to big bounces on offense and they don't cause us to question entire systems of measurement and projection. The bounces in UZR aren't (apparently) any larger than the bounces of many offensive stats.

Now I do think there are reasons to think that "true defensive talent" shouldn't bounce around as much from season-to-season (or day-to-day) as "true offensive talent" does and therefore I think that measurement issues likely play a substantial role in the high variance. But that's just opinion and I won't argue that something which correlates y-2-y as well as BA is a major problem ... I'll just keep rolling my eyes and shaking my head. :-)

Now I do think there are reasons to think that "true defensive talent" shouldn't bounce around as much from season-to-season (or day-to-day) as "true offensive talent" does and therefore I think that measurement issues likely play a substantial role in the high variance.

This. Said more succinctly than I ever could, and it encapsulates all my ambivalence about defensive metrics and, by extension, catchall stats like WAR and WARP.