Slog - Comments on Witness the Magic of Regression Analysis...

Comment by Mr. Poe

Mr. Poe — Wed, 05 Nov 2008 16:14:52 -0800

Baseball is stupid.

Comment by emma's bee

emma's bee — Wed, 05 Nov 2008 16:17:55 -0800

@1: that may be, but, fortunately, statistics is smart.

Comment by A Non Imus

A Non Imus — Wed, 05 Nov 2008 16:22:58 -0800

That wasn't the regression analysis...that was the Monte Carlo simulations.

Comment by flamingbanjo

flamingbanjo — Wed, 05 Nov 2008 16:28:43 -0800

How accurate were the exit polls? Just curious.

Comment by boyd main

boyd main — Wed, 05 Nov 2008 16:32:26 -0800

538.com was my rudder in the storm of the last few months. I think my faith in their analysis is the reason that I never really doubted Obama would win (at least since the debates) and why I wasn't surprised and all that elated last night. The reality had sunk in for me slowly, not all at once.

Comment by anonymous

anonymous — Wed, 05 Nov 2008 16:45:10 -0800

I have a horrifyingly strong crush on Nate Silver thanks to fivethirtyeight.com

Comment by Fnarf

Fnarf — Wed, 05 Nov 2008 16:51:49 -0800

Baseball statistics is the highest form of human expression. Nate's great, but he's not even at the top of the pile; have a look at what Tango Tiger and Fangraphs get up to. Now, think of teams that use this kind of analysis as Obama, and teams like the Mariners that want to play Jose Vidro at first base as McCain, and look what you get.

Comment by Eric Arrr

Eric Arrr — Wed, 05 Nov 2008 16:52:21 -0800

For me, a big part of the election was not just whether Obama would win but also whether 538's prediction system would prove out.

The fact that 538 correctly called every state in the union (except Indiana) and nailed the popular vote within a tenth of a percent is the proof I've been waiting for that the game isn't rigged after all, that maybe the system can be relied upon to work as advertised, and that when we lose, we won't be left to wonder whether we've been the victim of some conspiracy that is beyond sight and recourse.

Comment by jrrrl

jrrrl — Wed, 05 Nov 2008 16:53:24 -0800

I've read his stuff at baseballprospectus.com for years, and won $thousands in fantasy baseball because of it. :) Their stuff (not just Nate) has been scary accurate.

Comment by Simac

Simac — Wed, 05 Nov 2008 17:03:09 -0800

Plus, Nate is geek-cute.

Comment by leek

leek — Wed, 05 Nov 2008 17:21:28 -0800

I feel vindicated after my comments yesterday were spat upon!

Comment by Will in Seattle

Will in Seattle — Wed, 05 Nov 2008 17:28:07 -0800

Darn, and I was hoping for 380 to 382.

At least I got my Blue Tidal Wave.

The reason baseball statistics are useful is so many people have been betting on them for so long and we have developed good statistical methodology to allow for many things.

Comment by onion

onion — Wed, 05 Nov 2008 17:31:15 -0800

I know a bunch of science/math geeks who are going to cum in their pants when they see this.

Comment by umvue

umvue — Wed, 05 Nov 2008 18:17:19 -0800

Regression Analysis (and Statistics in general) is magic but...

... mathematical statistical acrobatics are nothing without good data. Credit belongs to pollsters and the designers of sampler schemes etc.

Comment by dw

dw — Wed, 05 Nov 2008 21:56:57 -0800

Nate's great, but he's not even at the top of the pile; have a look at what Tango Tiger and Fangraphs get up to.

Tango Tiger doing election stats? That would be scary. He'd have calculated exactly how much Palin hurt the GOP within 10 votes.

I'm just amazed 538 was that dead-on accurate. PECOTA is usually a conservative projection.

Comment by nicejewishgirl

nicejewishgirl — Thu, 06 Nov 2008 00:23:53 -0800

Does anyone know if he's single? 'Cuz I'm totally asking him out.

Comment by Fnarf

Fnarf — Thu, 06 Nov 2008 10:31:04 -0800

Will:

The reason baseball statistics are useful is so many people have been betting on them for so long and we have developed good statistical methodology to allow for many things.

Once again you are wrong, Will. For starters, "we" haven't done anything; I was there, watching, and even participating a little tiny bit (PNOPS was pretty bad, but briefly served a purpose), when the modern wave of stats was just getting off the ground, and I didn't see you.

And betting has nothing to do with it. Bettors also played no role in modern stat development. Once again, you have absolutely no idea what you're talking about.