Sam's 0.98 and 0.91 probabilities of an Obama victory seem high to me. We have 8 days until the election. I count four eight-day episodes in the past five months in which Sam's median electoral-vote estimate has moved far enough in the right direction to give Romney a victory. That would seem to me to indicate that Obama's chances of winning are 1 - 4/(5 x 30 / 8) = 0.79 -- unless, of course, something has happened to make public opinion stickier (or less sticky!) than it has been since mid-May.
Seems to me reasonably likely that Sam Wang's 0.91 and 0.98 numbers lean rather too heavily on the assumption of a normal Brownian drift (the 0.91) and on the assumption of a normal Brownian drift and that he has specified fundamentals properly in his Bayesian prior (the 0.98).