Lebowitz Schmebowitz

I'm sorry, but considering the mountain of evidence available, there's just no way I'm letting anyone get away with this:

-----------

I honestly do not know how to calculate: A) PECOTA; B) The Pygmalion Win Theorem; or C) Win Shares. Nor do I care. How many numbers do we have to sift through to realize that these facts and figures (some of which had not only the Marlins going 68-94 this year and the Padres winning the pennant last season) aren't any more accurate than the judgment of those who take statistics and other factors into account.

Emphasis his. When he says "other factors", I'm pretty sure he's not referring to things like this.

The most confounding part to me is that he seems to actually think that his predictions (2008, 2009) "take statistics...into account". There is, I'm pretty sure, one statistic that drives almost all of them: games won the previous year. Beyond that, it's all about creating storylines (how cool would it be if the Marlins won the division with their tiny payroll!?), considering what additions each team has made ('08 Giants + Renteria + Johnson = 89 wins), incorporating whatever other biases you have, and essentailly randomly combining those for a win projection.

How hard is it to have a formula, plug said formula into a computer and come out with the same predictions as everyone else who's using the same formula is coming out with?

Being different certainly is important. For instance:

Average projected record, 2008
Every single statistical projection system in the universe: 81.0-81.0
Paul Lebowitz: 82.2-79.8

Hooray for diversity!  At least he has a good excuse:

I didn't have time (or the motivation) to get those numbers right given the time constraints with getting the book done as quickly as possible.

So, he knew they were wrong, he just didn't care. That's certainly an encouraging sign for the accuracy of his predictions. Because Lord knows it would've taken hours to get them to .500.

There's no analysis; no knowledge from actually knowing anything about the game or about people; it's numbers crunching and it's boring.

No analysis; of course. There's no research put in, no history considered, nobody who knows anything about baseball actually created these systems. They're just a bunch of random numbers in a spreadsheet, really.

What's worse, it's no more accurate than listening to someone who has an idea of what they're talking about and isn't just blowing smoke out of their asses like Joe Morgan; or babbling endlessly about the veracity of numbers as the hard core stat-geeks do.

Now this -- this -- is brilliant.

I really wish Joe Morgan put together win projections, but unfortunately that's not the case. So I guess we'll have to make do with these:

Average RMSE, 2008 predictions:
PECOTA: 9.69
CHONE: 9.76
HBT: 10.24
CAIRO: 10.25
DMB: 10.47
ZIPS: 11.70
Lebowitz: 12.31
Steve Phillips: 12.47
Buster Olney: 13.25

To be fair, when you take both statistics and other factors into account, maybe Lebowitz's predictions come out on top. I just wouldn't understand.

H/T: BTF.