Forget spread in the polls and look at Obama's numbers. The RealClearPolitics average shows Obama and Romney in a tie, in other words, the spread between the two is negligible. So here are the poll numbers for surveys ending on October 20:
Obama 47, Romney 47
Obama 47, Romney 49
Obama 45, Romney 52
Obama 48, Romney 42
If you focused on the the spread between the two, you would see a dead heat. I don't believe that's the correct reading of these polls. This is a race with an incumbent, and regardless of the vast swing in the numbers for Romney, Obama's numbers are packed in a tight range, averaging 46.8%. (The RCP rolling average is 47.1%, not statistically different.) This is very bad news for the President, because he is the known quantity in the race and he can't break the 49.5% or so he will need to win the election. Further, these polls are lagging indicators, but the trend line towards Romney is significant.
Many might object that the election is not based on the popular vote, but the electoral vote. True enough, but they seldom diverge by much. In the electoral college the race seems to be coming down to Ohio, where Obama has maintained an advantage to date; with one recent poll calling it tied and a mainly Democrat polling firm showing Obama up by one. If Ohio follows a national trend, and the trend follows past experience of breaking to the challenger, I would predict a slim Romney victory. The conventional wisdom has not caught up with this and tomorrow's debate could of course change things. Intrade odds still sitting around 60% for Obama. However, the Intrade odds are heavily influenced by the available polling data, which might be lagging. In 2008, the betting site got 48 states right, but the two that bettors missed were Indiana and Missouri. In an election that might easily come down to one close state wrong, getting that one state wrong, that would also mean the site gets the election wrong. Right now, Intrade odds are at about 58% for Obama to win Ohio; so we see a strong correlation with the national odds. I am not sure why the CW hasn't caught up; but it doesn't matter yet. The race is tight and the debates are making a difference and there is one to go.
Nate Silver does something called a nowcast, where he has the popular vote at 49.9 to 48.9 favoring Obama. My nowcast, given past behavior breaking of voters breaking for the challenger, has Romney winning the popular vote 50.5 to 48.5 with one percent going to other candidates.