The problem with the oddsmakers or any predictive model is that they are dependant on accurate data. I believe the polling data is more baised and contradictary this election than any other. As they say, "garbage in garbage out." For example the national polls have Romney and Obama basically tied while the state polls show an Obama advantage. Also stories about increased Republican enthusiasm and Romney's advantage with independents contradict all the polls and suggest a Romney victory.
Several recent polls in supposedly solid Obama states also seem to seem to buck prevailing wisdom. A poll in Maine gives Obama 49% of the vote; polls in Oregon, Pennsylvania and Minnesota each give Obama 47% while a recent Michigan poll gives the President a mere 46%. Although Obama will likely win most of these states, this string of sub-50% polls definitely suggests luke-warm support. Even though Obama strategists scoff at Romney's attempt to expand the electorial map to MN and PA, Bill Clinton's presence at two Minnesota rallies, Biden's presence in PA along with increased ad buys in both states suggest that Romney/Ryan has team Obama rattled.
The (admittedly conservative) National Review article (see below) makes an interesting case for a potential Romney victory in PA. It basically argues that PA outside of Philly has been trending Republican in recent years and that Democrats must run up increasingly large totals in Philadelphia to counter Republican advantages in the rest of the state. The problem is the rest of the state is growing in population while Philadelphia is flat. To retain PA Obama must come close to matching his net gain of 480,000 Philly votes in 2008. The author further argues that the Obama campaign took PA for granted early and was caught flat-footed after the debates. They spent too much time and money in North Carolina when they should have focused on shoring up PA.
http://weeklystandard.com/blogs/mitt-s- ... tml?page=1