Prognosticators behave like gas prices with movement towards Trump and Harris

Ever notice that the decline in gas prices takes place over time, but then a single "jump" wipes out all of the decline, generally overnight?

So here is the odd thing. Yesterday there were several state polls released, including the very good Atlas polls that updated every battleground state. When RCP was done including all of the new polls, Trump's average in the battleground states actually increased by 0.1 to 1.0. It has seen gone back down to 0.9% as it was at the start of yesterday, largely because of some CNN polling in Wisconsin and Michigan that is out to lunch. If I look at Nate Silver's site there is little change in the state polling averages from where they were, and every state except Michigan was still showing trending red. Silver's last update yesterday was that the Atlas polling had pretty much wiped out any other polling gains that Harris might have had.

Now I have no clue where the economist came up with 65 new polls in one day, unless they are accepting polls run by users on X or Facebook. But nobody else saw that much polling no matter how hard they looked. But over the course of almost two weeks, Trump went from 50-50 to a 56-44 favorite on their site, only to see the entire lead wiped away with one day's worth of polling that others saw as a wash. You see something similar with 538 moving from 54% to 51% literally in one day (after taking weeks of good Trump polling to get to 54%). It seems like their gut instinct is to push hard for something to show that Harris is winning.

I look at a couple of states for an example:

• North Carolina: This is a net +4 state for the GOP that Trump has won twice and leads in every aggregate by over a point. The Hill shows Trump up 1.4% in their aggregate polling. 538 shows his chances of winning there as 59%.

• Michigan: This is a net zero state (38-38 D/R) and a state that Trump won once and lost once. Up until this morning he was ahead in Michigan. According to the Hill, he is now down by 0.4% for the first time in a while. 538 shows Harris with an identical 59% chance of winning?

So how does Trump have the same chance to win a +3 state he won twice and is ahead by over a point, versus Harris winning in a even partisan state that Trump split and where she barely holds a lead? How is this "statistically" possible without outside variables pushing the statistical needle in Harris's direction? I don't believe it is, and this is why I believe 538 has had a history of missing election results to the left. That bias is built into the model.