There's a reason polls are reported with a margin of error. Often in the press. a final margin error of anywhere between 3-5% is reported on the big question, and the margin of error is just a guess based on the end results compared to the polling data that was taken.
In the most recent poll, the final NBC/WSJ poll before the election, Obama leads McCain by eight points nationally among likely voters, 51 to 43 percent, which is down three points from his 53-to-42 lead in the poll nearly two weeks ago. It's being reported as having a margin of error of 3.1%, which could mean the numbers are actually 48-46, a statistical dead heat due to the electoral college, or they could mean as much as 55-39, a landslide.
The final NBC/WSJ survey before the 2004 presidential election had President Bush with a slim one-point lead over John Kerry, 48 to 47 percent, and was reported as a 2.9% argin of error. Bush went on to win that election, 51 to 48 percent, a 3 point spread, but a statistical error of 5.88%.
The final Reuters/C-SPAN/Zogby poll has things a little tighter, with McCain trailing by 6 points, 50 percent to 44 percent among likely voters in the three-day national tracking poll, up from a 5-point advantage on Saturday. That poll has a margin of error of 2.9 percentage points.
In both cases, if you examine the numbers carefully, you will end up with this, where it's McCain's turn to mark his next square today, Obama then marks his tomorrow: