Sam Wang's Princeton Electoral Consortium website http://election.princeton.edu is still hosed...
I was wrong: there is a Daily Wang today. They got something back up:
They give 19-1 odds against Romney if you are ignorant of background fundamentals, 99-1 odds against Romney if you make reasonable prior assumptions about background fundamentals and are Bayesian.
Me? I think Sam is doing something like using the Gaussian copula here--the odds are odds within the model, and the model is not the world. I'm with Nate, who still gives Romney a 1 in 5 chance.