A few days ago, Freepers were rejoicing about the Gallup numbers which had Bush up by 5%, and crowing about how accurate Gallup is:

Take Gallup for instance. Their final poll was actually amzingly accurate measuring George Bush. They came within 1/10th of 1 percent. But Gallup figured Nader would garner about 4%. That’s the norm for all pollsters: Nader would get 4%. Some had 3% (NBC, ABC); others had as high as 5%. Nader wound up with only 1.4 percent, losing 2.6%, or 2/3rds of his total, to Gore.
…
So before anyone panics, keep this in mind: In 2000, Gallup correctly predicted that Gore and Nader would receive 50% of the vote to Bush’s 48%. In 2004, Gallup (so far) has Bush leading by 5 percent. That amount may shrink with the release of today’s poll, but I would very definitely be surprised if the lead disappeared.

Well, guess what. The lead disappeared, and Gallup’s final poll has Bush and Kerry tied at 49% each–a 2-point loss for Bush and a 3-point gain for Kerry. And guess what else: Freepers suddenly flip-flop and find that Gallup is notoriously inaccurate about these things:

Beware the Gallup Poll (The Big Trick They Played to Get That Sudden Tie):

That 49-49 result is due to Gallup’s decision to allocate the remaining undecided voters based upon a tried-and-true formula — 9 to 1 for the challenger.

Pew, on the other hand, decided to allocate the undecideds they found equally. This is what Pew had to say: “Pew’s final survey suggests that the remaining undecided vote may break only slightly in Kerry’s favor.”

So, it appears that Pew’s data told them something that the CW of undecided voters, i.e. their breaking toward the incumbent, is off. Gallup just did what my 9th grade geometry called a “plug and chug.” As Polipundit notes, The New York Times survey indicates the same result as Pew. They found that undecideds do not seem to be breaking for Kerry.

Talk about infidelity. The first Freeper was not even comparing similar polls–he took a final poll from 2000 and compared it to a daily tracking poll 4 days before the 2004 election. But since it put Bush out ahead by 5 points, he was in love. Now Gallup is predicting a tie–and they were predicting a 2-point win for Bush in 2000, not surprising in light of the fact that they usually over-weight on the Republican side. And so suddenly Gallup is playing “big tricks” and can’t be trusted.

How consistent.