Sunday, 12 July 2009

Hilaripus projection from 2002

Looking at Baseball Prospectus from 2002, one finds this projection (a 'Wilton' in the parlance of the day) for Jose Ortiz: .330/.385/.534.

Ortiz was traded from Oakland to Colorado in 2001, after putting up a horrendous .167/.217/.167 in 46 plate appearances.

In Sacramento, before going to the Rockies, he put up .273/.345/.449

Once in Colorado, his line read .255/.314/.495 in 224 plate appearances.

So where did the .330/.385/.534 come from? Well, for that you have to go back to 2000, in Scramento .351/.408/.575. At age 23. In over 500 PAs.

The thing is, this looked way out of line with his production in Vancouver in 1999: .284/.346/.483

Eyeballing the lines, you'd have asked him to do that 2000 season over again before you gave him full credit. Yet the fact is, if you just weight the lines appropriately, and throw in a boost from Coors, you probably would get a .919 OPS.

I wouldn't like to say we've advanced so far in projection that such an error couldn't occur again. I suspect it has something to do with just automating a lot of calculations, and not checking what might appear anomalous. I don't doubt, regardless of whether the formulae are better, that a lot of these outcomes just get spat out of the computer still. But maybe they check them more.

Oh, Ortiz' real 2002 line? .250/.315/.313. In Coors. In 215 PA.

No comments: