Julien's Faster than Light Blog
Jump to top
Tuesday, July 29, 2003
my thinking on this has become so uptight!
we are settling into a tighter statline:
hitters and pitchers: ab wal con pow pra
at-bats you know. the rest i made up.
wal, con, and pow measure the three hitting skills: drawing walks, making contact, and hitting for power.
wal = (bb + hbp) / (ab + bb + hbp)
con = (ab - k) / ab
pow = .273 + .285 * ((tb - h) / (ab - k))
ed: pow has changed slightly.
they can be used to predict traditional metrics, like avg, obp, and slg:
pav = con * pow
pob = wal + (1-wal) * con * pow
psl = con * pow + con * (pow - .273) / .285
i'm working on a formula to account for speed, which is an important factor for certain extreme players, such as ichiro suzuki.
pra is predicted run average, ie how many runs a team of entirely that player would score in a game. it's based on bill james' runs created per 27 outs. ready?
pra = .9 * pob * psl * ab * 27 / ((1 - pav) * ab)
the pob * psl * ab part approximates runs created. the 27 / ((1 - pav) * ab) part is the "per 27 outs". the .9 is a scalar that accounts for other types of outs people get, like caught stealing and double plays. it'll make the hitters look worse, but it's closer to reality. a more precise formula, which will be used when the data allow, is:
pra = (bb + hbp + pav * ab - cs - gdp) / (ab + bb + hbp) * (psl * ab + .24 * (bb + hbp) + .62 * sb) * 27 / ((1 - pav) * ab + cs + gdp)
had enough? i have. for more information, and the mission statement, go here.
Comments: Post a Comment