from Amazon
Wednesday, December 30, 2009
Fangraphs park factors for pitchers
My answer must be: Yes, you are totally right, in response to this reader mail:
I had a thought about the FIP-based pitcher ratings on Fangraphs, and was interested to see if you agreed:
Fangraphs bases its pitching win values around FIP, rather than around runs allowed. This is an entirely defensible decision in the effort to remove fielding effects from the value of the pitcher, even if it will miss any and all non-FIP value accrued by the pitcher.
Here’s the problem:
http://www.fangraphs.com/blogs/index.php/pitcher-win-values-explained-part-six/
If I’m reading that correctly, the park adjustments they use are based on actual runs allowed. But part of what makes a park factor is how the park affects BABIP. You can’t give every pitcher the same theoretical BABIP and then use a park adjustment that will vary based on how the park influences balls in play, or you’ll be adjusting for the same bias twice.
Example: Rockies home/road BABIP splits over the last four years.
Hitters: .323/.286, .316/.300, .333/.311, .331/.289. Average inflation of 29 points.
Pitchers: .311/.297, .314/.312, .300/.287, .311/.304. Average inflation of 9 points.(I used both hitters and pitchers to avoid being biased in either direction by home-field advantage.)
So Coors seems to inflate BABIP by roughly 19 points, which is presumably a large factor in why it’s a strong hitter’s park. When evaluating the Rockies pitching staff, Fangraphs uses a system that ignores hits on balls in play entirely, and then gives the pitchers a bonus for toiling in what is a strong hitter’s park largely because it inflates hits on balls in play.
Two questions here, I suppose.
1. Am I right about this?
2. How big is the effect, in terms of, say, a 200-inning starter?
Basically, the FIP adjustment, that “3.2” should be calibrated not by league, but by park, if you are going to then use standard run-based park factors. Otherwise, you need to use a FIP-based run adjustment.
Cool insight…