Monday, November 17, 2008

The "Monkey Pollster" vs fivethirtyeight.com

Moved to:

http://tax-tricks.blogspot.com/2009/06/monkey-pollster-vs-fivethirtyeightcom.html

Comments:
Regarding "Marcel": he beats the pull it out of your butt prognosticators but usually lags a little bit behind the most sophisticated "systems" for projecting.

Regarding the "Monkey Pollster" (who in 2004 was masquerading as an astrophysicist), are you saying that you predicted each state winner correctly except Indiana, or did you miss some others as well? 538 only missed Indiana.
 
The only state where the simple projection was "worse" than 538 was North Carolina. The "Monkey Pollster" predicted NC at McCain 0.4%; 538 had it Obama 1.1%; and it went Obama 0.3%. This is not exactly a failing of the simple system.

538 also didn't call Missouri, so it missed two states.

I still think the most insightful concept that 538 introduced was a probabilistic distribution of electoral votes using Monte Carlo simulations. But its mean projections weren't necessarily better (in this election.)
 
Wrong. FiveThirtyEight got Missouri right. It missed only Indiana and one of the electoral districts in Nebraska.
 
A little late - I can't find the projections on 538 anymore. But if I recall correctly, on the final electoral map, Missouri was white, indicating that it was too close to call. Nate indicated that he thought that McCain would win it, but his model did not predict it with any level of confidence.
 
Post a Comment



Links to this post:

Create a Link



<< Home

This page is powered by Blogger. Isn't yours?

Subscribe to Posts [Atom]