But I have never really seen any stats attempting to prove this, probably because without DA-style numbers its hard to get a handle on how valuable someone's defense is. Once the data became available, though, it became possible to check our hypotheses about defensive value.

So, given the basic validity of Sherri Nichols's DA method, we ought to try and translate those numbers into something directly comparable in the offensive realm. I chose to model my DR numbers after the two endproducts of David Tate's MLV analysis: MR/G (number of runs per game contributed offensively to an average team) and TMR (total number of runs contributed by a player to an average team, for a whole season). I find the idea of breaking down a player's contribution into runs an aesthetically pleasing way of describing his value. We all understand what runs are, and we know that ~10 runs adds a win in the standings, which ultimately is what we are after.

Okay, with all that in mind, here are some points that need to be made about DR:

- The DA numbers are not park adjusted, so neither are the DR
numbers. I am curious as to what effect this will have on both sets of
data. Mike Gimbel has done the only study on defensive park effects I know
about; David Grabiner translated his numbers and posted these articles to rec.sports.baseball.
- I assume that some players do not have a more difficult set of opps
than others. I don't know how to check this hypothesis - it's the same
sort of problem as measuring the effect of catchers "calling the game".
The set of pitchers a fielder plays behind is not random (although the
batters he faces probably is, to first order), so its hard to distill out
the effect of the pitcher. I was trained as a physicist, so when I
encounter a problem I don't know how to solve, I simply ignore it.
- Note that NorH and NorXB are calculated differently - NorH straight
from DA and NorXB from the rates of giving up those XBs. There are in
places some discrepancies between these two elements of the DA data, and
can give some weird results. For example, look at Ken Griffey's line in
the AL 1992 data.
- From DR, we can see some limits on how valuable defense really is.
The most valuable defensive player season I've seen is Devon White's 1989,
at -41.26 runs below an average AL CF. The worst is Kirby Puckett's 1993,
at +41.94 runs above average. A more typical spread each year is -30 to
+30 for most positions. In absolute value, these are about half of Barry
Bonds's offensive value (above average) in 1992.
- Yes, I am using league average rather than replacement level as a
baseline. For one thing, its easy to calculate and gives meaningful
answers (you can tell right away by the sign on DR/G whether the player is
a good fielder or a bad fielder). A player's DR/G or TDR is probably most
meaningful relative to other players DRs, anyway, and so that it doesn't
matter what baseline you use. Finally, replacement level fielding might be
very close to league average. Weak (replacement level) hitters can't also
be weak fielders, or they would never make in professional baseball. Plus,
it is much harder to evaluate defense than it is for offense, and for most
positions your bat is what gets you brought up to the majors, anyway.
Therefore, I claim it is reasonable to expect the typical replacement level
player to be an average fielder at his position.

MR/G DR/G "Total Value" Blauser: +.141 +.081 +.060 runs per game Belliard: -.302 -.058 -.244 runs per game(Total value = MR/G - DR/G, since negative DR/Gs mean good fielding).

Not even close - Blauser's bat beats Belliard's glove, hands down.