Each week I've been taking polling data from Votemaster Andrew Tanenbaum's site www.electoral-vote.com and running simulations of the general election. I do two versions, one using the 4% margin of error common to state polls, and another using a margin of error derived from a regression of 2004 polling data against the actual results. The former is a snapshot of where we are now, and the latter allows for more movement in states before the election.

This week's numbers are even better for Obama than last week's.

4% Margin of Error:
Obama wins 100.0%, averages 360.9 EV (low 307, median 358, high 387)
McCain wins 0.0%, averages 177.1 EV (low 151, median 180, high 231)
No electoral ties.

10.3% Margin of Error:
Obama wins 99.99%, averages 346.6 EV (low 268, median 348, high 419)
McCain wins 0.01%, averages EV (low 119, median 190, high 270)
No electoral ties.

Yes, out of 20,000 trials with two different margins of error, McCain won exactly once. Obviously the one trial that resulted in a McCain victory has no statistical significance, and if I reran the simulation, I'd likely get different results - perhaps a few more McCain wins, or none. But it's safe to say that based on current polling, McCain effectively has no chance to win.

Just for fun, I looked at the one case McCain won in more detail. It was mostly what we'd expect. He takes must-win states Florida, Ohio, Virginia, North Carolina, Missouri, and Indiana. He also wins Colorado, New Mexico, and Maine(!), while Obama wins the other Kerry states plus Iowa, Nevada, West Virginia(!), and North Dakota(!!). This highlights that my simulation assumes each state is a completely independent trial, as it's almost inconceivable that Obama might lose Maine but win North Dakota and West Virginia. I should also note that I'm not attempting to allocate Maine's or Nebraska's congressional districts separately: I'm assuming these states are winner-take-all like the rest.

The new polls are again almost all favoring Obama. He's shown consistent leads in Colorado (4 new polls, ranging from +4 to +9) and Virginia (3 polls, +3, +6, and +10), and  Pennsylvania (4 polls, +8 to +15). Florida had 6 new polls, one showing McCain up 2, but the other 5 all gave Obama a lead from between 2 and 5 points. Similarly Ohio had 5 new polls, McCain leading by 2 in one, one tied, and 3 with Obama leads of 2, 5, and 5. North Carolina is still a toss-up, with 3 new polls evenly split between a tie and 2 point leads each way. Missouri may be breaking towards Obama (4 polls, one with McCain up 1, the others with Obama leads of 3, 6, and 8).

Obama is right to be worried about overconfidence, because this polling data sure looks good.