1) The last batch of swing state polls that came from FOXNews/Rasmussen show the closest numbers in a while. This includes McCain leading in FL (+1) and NC (+1). They show a tie in MO and OH. And they show McCain within striking distance in VA (-4) and CO (-4). Data here.
2) The major battleground states narrowed. Look at the RCP graphs to see the movement: FL, NC, VA, CO, MO, OH. All the graphs show the same spike for McCain and dip for Obama. If that movement is real, we could be seeing the break of undecideds toward McCain.
3) The national polls may be oversampling Democrats. In 2004, the party ID breakdown was 37-37 (D+0). In 2006, it was D 38-36 (D+2) while Independent voters broke heavily to the Ds creating the landslide. Almost everyone expects the D advantage to be greater than 2 this year, but how much greater? Looking at the last national polls released (who give their party ID breakdown), we see the following assumptions: FoxNews(+6), CBS/WSJ (+9), Rasmussen (+6.5), Gallup (+10), and Hotline (+5).
If reality is more like D+3 or D+4, that would mean all of the polls are oversampling Democrats. Combined with undecideds breaking toward McCain, we could see 3-5 point D margins in the polls turn into McCain wins.
4) The gap in favorability ratings has narrowed notably over the last few days. Specifically, Obama’s unfavorable numbers have spiked to their highest level ever, breaking the 40% barrier in 3 polls for the first time since April. And McCain’s favs have spiked up while his unfavs have spiked down in the past few days. Both men have favorability ratings well over 50%, so this election will be decided by people who like them both (unlike in 2004, when it was decided by people who swung between two people they disliked).
That’s the overview. If you’re watching for an optimistic outcome, you want to see PA and VA too close to call when polls close. And you want to see NC or FL look good, early.