|
I don't have much to say today, so I'm going to talk about a weird little out-of-context statement by a genius.
So PeteAbe had a nice little talk with the Red Sox' second most famous executive yesterday.

James had some interesting things to say about defensive metrics:
"The only difference between our ability to evaluate defense and offense at this point is confidence. Defense, at this point, can be evaluated with the same degree of precision and the same degree of agreement among different methods as offense," he said.According to James, we trust the offensive metrics more only because we are more familiar with them. Defensive stats still seem a little mysterious. "We haven't been doing it all our lives," he said. "But we've had pretty good methods now for five or six years. I've been doing the (offensive) stuff all my life. I know what's a normal gap between two seasons (offensively) and what isn't. I don't know the same (defensively)."
Really? I mean, I'm about as big a fan of UZR and plus/minus* as anyone, but are we that sure that's something we can say? Seems kind of crazy to me. Then again, he's probably got access to better stuff than what you get on FanGraphs or his own site.
* So yes, thanks to the positive feedback the The Metrics System "series" is definitely going to become an actual series, and eventually I'm going to do one on UZR and probably include plus/minus too. But in the meantime, primers on UZR seem to be the thing to do in the blogging world right now, so you really don't need me. Check out this one and this one.
What I do know is, as great as the stuff we have is, there's just no way it's as reliable as the various ways we have of measuring offensive stats. I think the reason the difference is "confidence" is that it's a little crazy to be as confident about "pretty good methods" that we've had "for five or six years" as we are about offensive metrics that have been in the making for a hundred plus. Even wOBA, new as it is, essentially takes all the stats we already had and gives them a good tweaking; UZR is, in a lot of ways, a brand new idea (or one based on older, very obscure ideas). The advanced defensive metrics are great tools, way better than anything we had before they came along, and it's great to use them. But if we're less "confident" in them, it's for a pretty good reason. If Bill James says he's more confident in them, it's probably because he and the Red Sox have better numbers to play with. Or maybe he's just talking (he does that sometimes).
|
Re: “Defense, at this point, can be evaluated with the same degree of precision and the same degree of agreement among different methods as offense” But precision and agreement among different methods are necessary but not sufficient conditions. Put it this way: we can probably be very “confident” that whatever we are measuring when we use advanced defensive metrics is "real," and reasonably confident that what we are measuring is “defensive skill.”
But where we can’t be quite so confident, I think, is in relating that to the bottom line – i.e., number of runs saved (and thus indirectly wins). With offense we can do that pretty easily (as anyone who reads this knows, by using team runs scored as a “check” on formulas for evaluating individual players). Defense is different, for reasons James himself identified when he first developed the win share system. There are ingenious ways to get around that problem, but there’s an additional level of abstraction that rightly gives us somewhat less confidence in the results.
And that doesn’t even take into account the issue of correctly apportioning “credit” (or blame) between pitchers and fielders. Yes, there are ways to estimate that, but there is no equivalent issue for hitting.
Bottom line: there are simply more steps in the process for possible error to creep in, as compared to hitting stats.