The polls in South Carolina and Florida have been the statistical equivalent of a zero-gravity flight: Sharp climbs followed by steep dives.
On Monday January 16 - five days before the South Carolina Primary, Mitt Romney was leading Newt Gingrich by an average of 7.7 percentage points in the polls released since Friday.
By Tuesday, that lead had increased to an average of 13.3 percentage points.
Then came the Tuesday night debate in Myrtle Beach which was Newt Gingrich's best debate performance of the year.
By Thursday, Gingrich was leading in three polls by an average of nearly nine percentage points, while two others had Romney still leading by just over 8 points.
By the way all these numbers are courtesy of RealClearPolitics.com.
If you look at polls, you should not just look at the "head-to-head" numbers; but you should also look at when the polls were "in the field," that is, over what period of time was the polling firm actually making the phone calls.
Polls are not predictive. They demonstrate the state of the race during the time the poll was being taken. The closer to the election, the closer the polling results should be to matching the actual results.
Polls have been moving so quickly, that some polling firms have gone to releasing not just the total number, but revealing what the nightly results were to demonstrate whether a trend was developing over the two or three nights of polling.
For instance, in South Carolina Romney might have been leading in the first night's polling, Gingrich and Romney might have essentially tied in the second night, but Gingrich might have been very strong in the third night.
The total results are the same, but a trend like that would be very interesting to know.
By the time the Thursday night debate had been in Charleston, South Carolina - that was the Marianne Gingrich Debate - Gingrich had solidified his lead, and all three polls released Friday and Saturday had him with a seven point lead.
That represented a 20 point swing toward Gingrich in five days.
The swing grew to nearly 26 percentage points as Gingrich crushed Romney 40.4% to 27.8% when the votes were counted Saturday night.
The early polling in Florida was the exact mirror. Gingrich, in the afterglow of his South Carolina victory, came out of the gate with an average lead of 7.25 percentage points over Romney in polls released Monday and Tuesday.
After a weak debate performance by Gingrich on Tuesday night, polls released on Thursday had flipped to Romney who was leading by eight percentage points in two of them, and seven in a third.
Romney's best debate performance of the season on Thursday night resulted to his leading the three polls released yesterday (Sunday) by an average of 14 percentage points.
A swing of - are you ready - 21 points from Monday to Sunday toward Romney.
National polling has followed the in-state polling over the past 10 days. The Gallup organization has been conducting a five-day tracking poll since early December.
The way a tracking poll works is a polling firm will make some number of calls - say 100 - per night over three, four or, in Gallup's case, five nights. After the total is reached the results are released (either to the campaign if it is a private poll, or to the media if it's public).
Gallup released its numbers after five nights. On the sixth night, they recalculated the results leaving the first night off. On the seventh night they left off the second night's results. What this does is give you a very good sense of the arc of a campaign.
Going into South Carolina Romney had a 37-14 lead in Gallup's national five-day tracking poll. By Thursday the national numbers had change to a Gingrich lead over Romney of 32-24.
By Saturday the Gallup five day tracking poll had closed to a 28-26 lead for Gingrich meaning the later days in the track were strongly for Romney.
The Florida primary will be held tomorrow night and we'll get to compare the polling with the voting.