But I still DO NOT LIKE seeing her numbers sliding down on 538.

but you're right, I should just ignore it all...but I'd like to know what the "real" polls are looking like and better understand why Nat's 538 in general seem so much lower than any other site.

Because his predictions in the primary went so spectacularly wrong. He had trump <10% likelihod to win the nomination.

So he changed his model to react to every poll that comes in. Others don't so much.

... flipped from light blue to light pink is Arizona. Would it account for the drop in % likelihood to win?

All his states have correlations built into them. So if some states are strongly correlated to movements in MO, then their margins will slightly decrease too. This has a significant effect across many states.

Any thoughts on how you think the 538 numbers will move (a little more toward Trump? a little more for Hillary?) based on state polling numbers?

His model is too variable this year.

so if/when it happens I won't be in total freak out mode! What about other models (i.e., Sam Wang)? Thoughts on where you think they might land?

Also -- any ideas on how early voting may or may not impact the models? When state polls are gathered, at this point, are they still going for "who are you likely to vote for?" versus "Have you voted and if so, who have you voted for?"

I think they are finally revising their biased methodology so they fit closer to the real numbers.

Assessing polls by their final numbers only is BS.

If your polls were off for the entire election and the at the last second they make a "market adjustment" so they can be credited with scoring the race correctly well... that shouldn't count.

I've noticed over the last 2 Presidential cycles, conservative polls do this frequently.

There is no way this poll was a correct analysis of the race, and now at the end, they are lining up to get another good rating.

Those ratings should somehow include polling performance during campaign.

usually this works by polling companies getting skittish about releasing numbers they think are outliers. With tracking polls, that's difficult because they have to release one every day.

To cover the Republican bias of that poll.

This reversal is huge and only been a few days