As I have said before, a bookies' numbers represent the perceived reality only, and as time goes on, it also represents the cumulative effect of betting that has been taking place all season, which will naturally skew any result. Even then, the odds are such that if we ran the election ten times, the bookies think Romney would win three times. The problem is that we don't run the election ten times, only once.

Nate Silver has written about this extensively on his blog. When he says, for example, that Obama has a 62% chance of carrying the state of Iowa, that means that Romney still has a 38% chance. So if you ran the Iowa election ten times, Romney would win about four of them.

Your assumptions also ignore the fact that sometimes bookies take hits on bad calls. Just like the miraculous sports game that blows through the point spread can sometimes leave the bookie exposed (though theoretically they're hedging their loss well), the same can happen in an election.

You're essentially doing nothing more than stating the "more likely than not" outcome from the bookies' perspective.