You can read all the polls you want, you can crunch all the numbers, you can come up with all sorts of scenarios, and you can consult all the fortune tellers and Sacred Scrolls that predict that Barack Obama will win the election in a landslide, but it’s not November 5, 2008 and won’t be for little more than three weeks from now. He could still lose. Walter Shapiro in Salon.com points out how that could happen.

What polling mavens too often forget is that an election is not a computer simulation or a contest decided by the best use of regression analyses in analyzing published data. As a one-time event, all that is required is for a winning candidate to get lucky, very lucky, on Election Day. And a passionate embrace from Lady Luck is probably now the only way that John McCain will ever find himself behind the desk in the Oval Office.

I’ve been around long enough to know that very rarely do trends this far out from Election Day switch dramatically, but it’s happened; vide President Dewey in 1948 or Jimmy Carter’s second term in 1980.

I’m not being a total Eeyore; I think Barack Obama will win the election, but I don’t think it will be a landslide. I’m going to go out on a limb and predict that he won’t get over 50% of the popular vote, even with all the newly-registered Democrats and energized younger voters. If I’m wrong, I will gladly eat some crow. But I also know that up until the last twelve seconds of the game yesterday, the Miami Dolphins were beating the Houston Texans 28-23.