from Amazon
Monday, September 29, 2008
End of year Sabermetric stats
Courtesy of Patriot.
I don’t really have much to add. Patriot noted that he uses a 73% offensive replacement level, likening it to a .350 OW%. Using PythagenPat and 4.5 RPG per team, I get .364. No biggie. Just wanted to point out that he should probably be saying .360 not .350. However, what if you look at it as one replacement guy with 8 average guys? In this case, this team will win .486 games, making our replacement level -.014 wins per game (or more accurately per one-ninth of a game slice). Adjusted to a per game basis, that’s -.014 times 9 equals -.126, or a .374 win%.
That is, rather than presuming 9 replacement-level hitters with a team of average defense, we presume 1 replacement-level hitter, 8 average hitters, and average defense. That gives you a .486 win%. The marginal impact is .014, which you “annualize” by multiplying by 9. Kinda like ERA for relievers. Anyway, to get it to my replacement level, I’d use 74% or 75%. We’re pretty much in agreement here.
With starters, if we repeat this process, but presume 5.4 IP per replacement start, and the bullpen gives him average support, then Patriot’s 125% gives you a starter win% of .390. To make it .380, you’d want 1.27 or 1.28. So, 125% is perfectly fine.
For relievers, it’s the same process as hitters, if you presume 1 IP per replacement relief. You’d want 106% or 107% of league average.
Anyway, basic core agreement, with just a smidge of disagreement on the peripherals.