I think one of the most noteworthy things about this particular election was whole debate over polling and polling numbers. Never, it seemed to me, had the number and predictions, leading into the Day Of, been so far apart.
Part of that was, I believe, because of the evolution of intelligent polling aggregation, of the kind Nate Silver does (and which we'll see a lot more ofin the future), meaning that the natural variances between polls and the tendency to cherry-pick the ones you want couldn't hold up.
The other part of this was what I think was an explicit, intentional effort to simply present the numbers that the GOP and GOP faithful wanted to see. Whether it was wishful thinking ("All the polls are skewed to the Left, so we'll just shift everything 5% to the Right") or an intentionally deceptive act to encourage the troops, I have no idea. Perhaps some of both, depending on who was doing it.
Polls and surveys aren't perfect, of course, both because of methodological issues and because people don't make their decision, really, until they make their decision. "Dewey Beats Truman!" and all that. But they are of value, when used correctly (just as individual experimental results can, with care, be aggregated into larger studies), and I think we just got a good demonstration of that.
(h/t +Shannon Turlington; via http://xkcd.com/1131/)
Google+: View post on Google+
Certainly part of it is keeping up the enthusiasm that your candidate could win, but there needs to be some connection with reality.
I found one commentator today still saying you should believe your gut instinct. I'll post it later.