Lots of hot air has been expended on the Iowa and New Hampshire results. I don't really know much about American elections (who caucuses? just vote!), but at least I am cognizant of that fact. Our mainstream news outlets are apparently clueless about their own ignorance ("We were wrong? Must have been those New Hampshire racists!"). After Obama won Iowa, he was about to be crowned king and Hillary was finished - a few days later, and it's neck-and-neck. McCain was supposed to be finished this past summer, now he's the new front-runner.
Intrade.com has a fascinating graphic showing McCain's comeback. The graph shows how much Intrade.com traders valued contracts on McCain winning the GOP presidential nomination. The payoff of the contract, if McCain wins the nomination, is 100; otherwise it is zero. So in November and December of 2006, traders valued McCain's stock at 55ish. By July 2007 it fell to around 5, and after New Hampshire it has climbed back to around 35 - the highest of the Republican contenders.
National polls largely mirror this trend. I assume Intrade.com traders were trading on the information of national polls, Iowa, New Hampshire and other primary polls, the financial situation each candidate is in, etc. The problem is that people look at all this data and think it means more than it does, because it's all we have - and any pundit that says "I don't know" won't be invited back on TV.
Soft support for candidates means that poll numbers are flexible. High voter turnout also upsets polling data, which is based on sampling from different demographics based on who has historically turned out to vote in the past. I'm sure lots of other things confound poll numbers too. Thus the polling numbers that our chattering classes obsess over in truth don't mean very much. And so to take it one step further, the output of the chattering classes doesn't matter very much either (except to the degree that other people take it seriously).
My point is not that McCain was always going to make a comeback and that pundits should have known - there was nothing fated about his remarkable turnaround. My point is that the value-added of these pundits is minimal and that their assertions are almost always over-confident.