... with undecideds, I think Nate Silver's explanation makes the most sense:
My other annoyance with the punditry is that they seem to weight all segments of the debate equally. There were eight segments in this debate: bailout, economy, spending, Iraq, Afghanistan, Iran, Russia, terrorism. The pundit consensus seems to be that Obama won the segments on the bailout, the economy, and Iraq, drew the segment on Afghanistan, and lost the other four. So, McCain wins 4-3, right? Except that, voters don't weight these issues anywhere near evenly. In Peter Hart's recent poll for NBC, 43 percent of voters listed the economy or the financial crisis as their top priority, 12 percent as Iraq, and 13 percent terrorism or other foreign policy issues. What happens if we give Obama two out of three economic voters (corresponding to the fact that he won two out of the three segments on the economy), and the Iraq voters, but give McCain all the "other foreign policy" voters? ... By this measure, Obama "won" by 14 points, which almost exactly his margin in the CNN poll.
If these numbers show up in wider polling, it seems like awfully bad news for McCain. If he can't get a bump from last night's showing, which struck me as a pretty strong, I think it's going to be awfully difficult for him to get a bump of any sort across the run of debates. It's hard to envision him turning in a vastly better performance where the focus is explicitly on domestic policy (if you thought we heard a little too much about earmarks last night, just you wait ...), and of course there's the looming Palin-Biden fiasco for his campaign to weather as well. If undecided voters didn't like what they saw from McCain last night, I don't know what, exactly, his campaign can do - given his range and limitations as a politician, and the obvious weakness of his running mate - to win them over going forward.