John D. has a great post

The Corner on National Review Online: "A physics Ph.D. candidate at a very respectable university: 'Mr. Derbyshire---I felt a strong need to actually figure out John Kerry's odds of winning Ohio in a somewhat rigorous fashion. Not too promising for the Kerry campaign: I figure there are 175000 provisional ballots out there, as that seems to be the high end of official estimates. If we assume that each has a 50% chance of being counted (and that's way high, it's really probably more like 15-20%) then having more ballots than Bush's margin of victory is a 360-sigma event, which comes out to a probability of 2.8*10^-28145. [That's a number with 28,144 zeros to the right of the decimal point, then some nonzero digits: 2, 8,...---JD] But hey, it could happen. And if every one of those votes goes for John Kerry (probability of this is around 10^-850000 or so) then Kerry might just pull this off.' "