First, yes I've changed the headline on this series of posts about Bill James' series of articles, for the simple reason that Bill's been writing about fielders and not pitchers, whilst "defense" might logically include pitching. I made a couple of bad decisions, but that doesn't mean I'm stuck with them forever.
Anyway, as mentioned in one of those earlier, poorly headlined posts, Bill is in the process of introducing, at his site, a redesigned version of his Win (and Loss!) shares, at least the fielding portions. Which is pretty exciting to people like me (and I hope you, since you're here). Here's a really interesting thing!
One way in which it is different is that I have abandoned the "causal link" between outs made and defensive responsibility. The number of innings that a TEAM must play in the field depends directly on how many outs they make at bat. For every three outs you make at bat or on the bases, you have to play one inning of defense.
For that reason, I had (in the past) a "link" in my defensive analysis which placed additional responsibility on a defensive player if he made more outs at bat. Gary Sheffield, for example, ranked better in my defensive analysis than in anyone else’s defensive analysis, because Sheffield, with his very high On Base Percentages, didn’t make very many outs, Since he didn’t make outs, he was not placing a defensive responsibility on the team. Conversely, Aurelio Rodriguez or Ozzie Guillen or Jerry Adair would look worse in my system than in others, because the number of outs that they made as a hitter creates additional responsibility to play defense.
It is an entirely logical link, and I would defend that link today as I would have two months ago. What is true of the team must be true of the individuals on the team. The failure to represent this in the statistical model is a failure of the statistical model.
Nonetheless, it is almost impossible to explain to a third party why Gary Sheffield should receive a higher defensive value than another player based on the fact that he has a good on base percentage as a hitter. That’s one serious problem, and another is that this "link"—like all links—makes the analysis more complicated, more time-consuming, more ... awkward. When you make a line of analysis more awkward, you limit what you can do with that line of analysis. I decided that, regardless of what I might think, I needed to disconnect that link in order to make the analysis work better.
Sometimes Bill writes (or says) something and I think, "Well, Bill's just wrong about that!" Which doesn't mean I know he's wrong; rather, I figure there's maybe a 90-percent chance that he's wrong (and I'm right, of course). But this isn't one of those times. This is one of those times when I figure Bill might be 100 PERCENT WRONG ... or he might be operating in a completely different sphere of thought, and has simply left me behind. Again.
The notion that Gary Sheffield's on-base percentage made him a more valuable fielder is ... well, obviously bizarre on its face. Which doesn't mean it's not true. But I think Bill's right about this: Making that bizarre argument loses in credibility whatever it might gain in logic or accuracy. Doubly or trebly so, I suspect.
So I'm glad to see this. Win Shares, for a few reasons, never has gained the currency I think it could and should have. Bill deserves some of the "blame" for this, but so do we. Maybe it's too late, but then again maybe it's not. And I would still rather see a 35 next to Mike Trout's value than an 8.3.
