Sunday, January 25, 2009

Reliability

I hate hate HATE articles like this one. The author--and readers--are searching for some magic point where a statistic (in this case, BABIP) suddenly gains credibility. One might as well ask how many dates you have to go on before your girlfriend spontaneously switches from 'gold-digging whore' to 'future Mrs. Sabermetrician'.

In a way, I feel the linked post is the best of its kind, because rather than pretending to be scientific, it admits (albeit in the last paragraph) that it has been a waste of the reader's time:

"This isn't a very functionally useful finding for evaluating players or predicting what they will do."

Translation: forget everything you just read, you'll never use it. Pizza Cutter should pitch a new book idea to Michael Lewis: "Chaosball: The Art of Not Predicting a Random Walk".

1 comment:

jwilli7122 said...

i don't know. i don't think the article is trying to find a point at which BABIP becomes useful. it's just trying to point out that it's not totally determined by chance. pitcher A in yr X probably has a true BABIP that is different from pitcher B's in yr X. we can't determine, for sure, whose is lower, but that doesn't mean that we can assume they are the same. (well, we can in a model, but not in real life.)