Wednesday, June 24, 2009

Exceeding expectations

The math says we should be 37-34 (.518 win pct.), but we are cruising along at 39-32 (.549), same record as the Damn Yankees. They find themselves--like us--in 2nd place, behind the best team in the league. The aforementioned math is from MLB and is their version* of the so-called "Pythagorean" or "Expected Wins" formula. The whole thing rests on only two numbers: runs scored (RS) and runs allowed (RA). That's pretty simple, it doesn't account for things like timing and luck, or veteran savvy clutchness, or the preponderance of gnarly mo-fos on your squad. Including tonight, we've scored 281 runs and allowed 270. The Damn Yankees have scored 386 runs and allowed 350, and the math says they should be 39-32. The Giants have allowed the fewest runs in the majors, quite an impressive feat. We got another impressive start, this time by The Great Old One. I continue to be impressed by this team--amazed, in fact. Improbabilities and intagibles are still improbable and intangible, but we seem to have an abundance of both. Then again, maybe we really are this good. After all, we are only "exceeding expectations" by two wins--that's not so surprising considering the vagaries of the game and the many ways one team can prevail over another. Our formula is "don't give up runs." Lately, we are including the other part, "score some runs," and by-golly, it works. I especially liked the two-run HR by Olympic Nate!



Bob said...

We hoped Lincecum would not be a flash in the pan. We hoped RJ would not not be just an embarrassing shell of his former self, going through the motions of getting a few more wins to embellish his career stats. We hoped that Cain's success would some day match his ability. We hoped Zito would not suck every time. We hoped Sanchez would be good.
Four out of five is not bad.

Zo said...

This is probably an example of a stat that either needs a margin of error like they give to polling results or else is better expressed as a bell curve. The high point on the curve, or the most probable result given only runs scored and runs scored against, is that we could be 11 games over .500. That is the single most probable result of all possible results, but is not likely (but amazingly, hey! we're not far off). We could also be much better, having lost by blowouts (didn't we get outscored in 1999?) but won more games by one run. Or we could have won by blowouts about once a week and lost more. Maybe someone with programming skills could run this on a computer and generate the curve. I'd like to see it, and I would bet that within a 95% confidence level, we could go from a losing record to pretty close to where we are. Since I never see this kind of thing when people talk about baseball stats, I assume that they attribute more relevance to the stat than it deserves. How 'bout that Unit?

M.C. O'Connor said...

Actually, if you followed the link, they do some regression work with various exponents to "improve" the formula. Believe me, if you want in-depth stat stuff, you can find it. This formula does not deal well with outliers like winning all your games by one run (and they admit that). They just tried to find a "best fit" empirically.