Quote:
|
Originally Posted by theview
Yeah, and I think based on some old rough tests that people ran in prior version, this has been what people have roughly done.
The problem is that it doesn't work backwards. So, while a 80 in range for a CFer may equal a +15 UZR on average, we can't tell what actually happened in any particular season (and, so, 80 in range would be considered the same as 89, but would be different in kind than 79).
But, we don't know if the fielder was actually a +15 that season or if their ratings just showed them to be likely to be a +15 that season. (Or, put in other terms, the player was a True Talent +15, while the combination of luck and park factors [and variance] caused a +7 or + 21.) Because defensive UZR numbers are more variable than, for example, OPS on a season-to-season basis, any specific season may be a good deal away from what their rating projected.
(I write all the above really for others; I assume you knew all that.0
|
Well, thanks for the benefit of the doubt, I guess...
My thinking was that if you can't track balls precisely, and the best you can do is some kind of estimate in the vein of Fielding Runs Above Average or Win Shares, you're better served working backward from perfect information about true talent than introducing another layer of error with all the assumptions you have to make to estimate ball distribution. That +15 in any given year may be a bit sketchy, but your confidence grows with the sample, and you never end up whiffing on Chipper Jones to the tune of 15 or 20 runs every year for a decade because the Braves have an odd distribution of ground balls.