Today, one of the premier sabermetric websites, Baseball Prospectus, ditched one of its signature stats, saying it wasn't doing the job. Luddites will, of course, crow about this, but they should be eating crow instead.
What BP is doing is exactly what the hidebound can't abide: moving forward to improve the state of the art. The BP crew spent a year-and-a-half tweaking SIERA, the ERA predicting tool, until they finally decided that it was not a significant improvement over other tools already in use.
SIERA attempted to measure how much of a pitcher's ERA was his pitching and how much were non- repeatable factors outside his control, strip out the noise and then project what the pitcher would do going forward without luck's finger on the scale. There are already simpler statistics that strip out fielding and fluctuations in BABIP from a pitcher's work. And there are already a plethora of sabermetric crystal balls that perform marginally better than an educated person's estimate. So BP is leaving in the starter and consigning SIERA to the eternal bullpen.
Learning is not linear. We gain knowledge in fits and starts, by trial and -- yes -- error. Another seamhead mixing up a new statistical brew will sprinkle SIERA concepts into his cauldron and we'll all benefit collectively.
b
No comments:
Post a Comment