[extropy-chat] Kerry to win, 2.5 to 1

[Eliezer Yudkowsky]

Robin Hanson wrote:
> Tradesports, IEM, Betfair give Kerry a 71 to 74% chance to win.

[Eliezer] Whoa!  Robin Hanson says that the markets are breaking 2:1 in 
favor of Kerry.  IEM at 67%, Tradesports at 63%.  Maybe they're not worth 
rotten petunias for predicting the outcome a month in advance, but they 
might be good indicators of who's winning on Election Day - we'll have to see.
  <MitchH> "Quick, dump your Bush shares!  The markets must know 
something!"  Bandwagon in the markets is probably one of the most 
pronounced of all, hehe.
  [Eliezer] True.
  <James> yeah, those markets are reactionary
  <James> like the early exit polling that showed Kerry leading
  <James> with totally non-representative internals
  <James> sent the markets plummeting
  <James> including NYSE/NASDAQ
  [Eliezer] hold on, non-representative intervals don't change the Bayesian 
interpretation of the data - if it swings wider than expected given the 
bias, that's evidence
  <James> *internals*
  [Eliezer] ok, what's non-representative internals mean?
  <James> meaning that the sample population was not representative
  <James> in this case 60/40 women/men
  <James> among other things
  <James> I've seen some analysis that normalized the internals on those 
exit polls
  <James> which gave Bush +1-3, depending on the method
  <James> Exit polls are pretty bad anyway
  <James> not a good track record in many cases
  <transBeing> (your favorite is Kerry?)
  <James> My favorite is Cthulhu
  <James> Or maybe Kodos
  [Eliezer] heh, I was just about to say Kodos

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence