I remember hearing Dick Morris after the 2000 debacle. (I hope I remember these numbers correctly.) He said that the electoral college can be won by one man, and the popular vote another, but it was extremely unlikely that it could happen with more than a 1 1/2% difference.
In other words, if Bush goes over 2%, it would be highly unlikely to lose the electoral college. This was a statistical analysis. Of course, nothing about this election is usual or ordinary. But if Bush wins by a number near 8%, I don't think there is a snowballs chance in hell Kerry can put the states together for an EV win.
You would be absolutely right if this were the entire electorate - but it's not. The sample of voters who have already voted is not necessarily a completely representative sample of those who will vote tomorrow. I hope it is! But it's not likely.
I would take an 8% margin in the entire electorate any day of the week and twice on Sunday.