I've been finding it vaguely odd to watch liberals grow agitated over mild tightening in the final days of polling. These are the same folks who've been loving Nate Silver's FiveThirtyEight.com, but one of Silver's innovations was to examine past elections and build his algorithm such that it predicts that the race will tighten in the days before the election. It tightened for Kerry, for Gore, and for Dole. It's a common feature of elections. That's part of why his model has a sounder claim to accuracy than competing methods of prediction. No one who's been reading his site should be surprised by this.
