Wednesday, October 22, 2008

Call Off The World Series

You think Boston-Colorado was a mismatch last year? Check out what Diamond Mind is saying about Tampa Bay-Philly over at ESPN.com:

[O]ur simulations project the Tampa Bay Rays to continue their "Cinderella" run and defeat the Phillies. In fact, Tampa Bay won over 71 percent of our 2,000 series simulations, the largest winning margin of any postseason projection we've done for ESPN.com. When we projected the Red Sox as heavy favorites last year, we gave them a decent chance of a sweep (Boston swept 105 of the 1,000 series simulations). In contrast, Tampa Bay swept 299 of the 2,000 simulations we ran for this year's Series, or 15 percent, and the Rays won the Series in five games or less 626 times (31.3 percent). In other words, the simulations do not bode well for the Phillies, and the odds are against them even making it to a seventh game.
Put differently, Boston swept Colorado in 10.5% of the simulations last year, while Tampa Bay is sweeping in 15% now. That's pretty dramatic if you ask me. Other fun nuggets:

As good as Philadelphia's pitching has been in its run to the World Series, our simulations suggest that the Rays' young guns will easily outduel the Phils' pitching, with the exception of Hamels, who projects as the only Philadelphia starter to perform well . . .

. . . The bullpens both project to do well, though the Rays' bullpen did a bit better in the simulations. So it appears that it will come down to the starting pitching; unless the Phillies' starters can step up and outshine the Rays' rotation and perform like they did against the Dodgers, Tampa Bay will be claiming its first World Series trophy.

. . . A short series often will produce standout performances, sometimes from unlikely sources . . . [t]he simulations show that Tampa Bay's defensive star, shortstop Jason Bartlett, just might be that guy -- he came up with a big hit in a surprisingly high number of simulations. Looking for a stand-out game? Garza threw a no-hitter in simulation No. 1,493, while Shields had three one-hitters in the 2,000 simulations.
The two things that strike me the most from the simulations are how (a) the Rays' bullpen stacks up better than the Phillies, which is the opposite of what we've been hearing from many pundits; and (b) the degree to which the Rays' rotation outshines that of the Phillies. As for that point, I think everyone agrees that Tampa Bay has an advantage, but I didn't expect the difference to be as stark as the simulations would have it. According to Diamond Mind, the Rays' worst starter over the 2,000 sims is Andy Sonnanstine, who posted a 4.26 ERA. That's better than any other Phillies starter except for Cole Hamels.

I realize that many of you are going to dismiss this as nothing more than the outspit of a silly computer game. And that's fine, because, hey, the games are played on a field. But don't dismiss it too casually. Diamond Mind is a pretty spiffy program, and the guys behind it know their stuff.

If that doesn't convince you, know this: I was on a Diamond Mind sim league this year in which I drafted a strikingly mediocre team, known as the Matewan Massacre. That team finished almost exactly .500.

I know: Heavy.

(thanks to Keith Law, ever the company man, for shooting me the link)

3 comments:

Anonymous said...

In how many simulations does David Price pitch a gem, only to have an article written about his similarities with Barack Obama?

Daniel said...

It actually doesn't surprise me that the Rays bullpen does better than the Phillies, given that the Rays can throw out great platoon matchups all over the place with their lefties and the Phillies have several guys with pretty large platoon splits (notably Ryan Howard).

If Howard or Utley hits against a right handed pitcher from the 7th inning on in this series, Joe Maddon needs to wake up.

Mark S said...

Curious what would happen if we had 100 monkeys pick the winner or had 10 monkeys pick 10 times? I suspect their success rate would be equal to or better than any program could predict. Not trying to dismiss the science/math used by all the stats folks as I am an academic and use many myself but there are just way to many human factors that cannot be accounted properly in defining your variables used in your modeling. Granted you can get close but I suspect the monkey could produce similar results as study have shown in picking stocks.