Some computer science and political science students at the University of Illinois have put together a mathematical model to predict the probability of each candidate receiving a majority of the Electoral Votes, and thus winning the presidency. As inputs they use well-known polls of voters in each state and then aggregate the results using a dynamic programming algorithm. You can even play around with the assumptions, depending on your view of how swing voters are likely to lean. The data don't cover the period between the time Palin's candicacy was announced and just before she gave her acceptance speech, but it is nevertheless clear that Sarah Palin was the biggest thing to happen to this election so far. Obama was almost a sure thing to win in late August, but now McCain has a modest lead. The betting on Intrade has generally reflected this same shift, although today McCain trails Obama by 4 points.
HT: Club for Growth