It's better than that as these guys are more accurate than 538 and they are not a profit making group. Clinton Nov. win probability: random drift 94%, Bayesian 97% http://election.princeton.edu/