That's fundamental to this election mode. Most swing states were within the predicted range, they just happen to all be correlated (which is expected) and swung in the same direction having a huge effect on the electoral college.

I disagree. There's a big difference between saying "kamala will win, it's certain", and "we don't know".

That's true, lacking confidence is less of a failure than confidently getting it wrong. But they weren't actually saying "we don't know". They were predicting a split election. Do pollsters even have a way to report that they lack enough confidence to give a prediction? I rarely see CIs on reported poll results so presumably they'd have to just refuse to publish any prediction at all, which clearly, they weren't doing.

Nate Silver has recently written about the clear problems in polling, and in particular the herd-like way they were reporting implausible numbers:

https://www.natesilver.net/p/theres-more-herding-in-swing-st...

I dont belive that claim is actually true?

the most likely result predicted by 538 was 312 for trump [0]

the issue with the model was the 2nd most likely result was 319 for harris.

they thought the odds of a recount being decisive was around 10%.

That hardly seems evidence of "predicting a split election". which prediction are you thinking of?

[0] https://projects.fivethirtyeight.com/2024-election-forecast/

538 is not longer run by Nate Silver, he runs the silver bulletin on substack now.

Apologies if you thought I meant this. I was using them as a reference for what people modeling the election from polls were predicting!

I don't know what Nate Silver was predicting. Was he predicting a near-split election or the situation where "someone is decently likely to win decisively, but we don't know who"?

Fair enough. Nate Silver was predicting a toss up, but was upfront that the model uses many simulations. Since the GP mentioned Nate Silver, I mistakenly took your comment about 538 as disagreeing with that since Silver did used to run 538.

Silver's predictions were extremely close to 538's. He predicted a toss up (50-50% chance of Harris or Trump winning the presidency), but many of his simulations were not particularly close.

What sources are you thinking of? Everywhere I looked, I saw "the polls are very close, the result probably won't be so close but we don't know which way it will go". I don't recall seeing anyone outright predicting a very tight result (beyond "here's what happens if there's a tie" articles -- background info rather than prediction).

>They were predicting a split election.

Who was? A 50% chance to win does not imply that the vote count will be close.

Also: statistical uncertainty is a feature not a bug. A lot of the idea behind statistics is the ability to quantify the certainty of the point estimate. As another commenter put it: a statistically sound "idk" is a better result than a confidently incorrect estimate, from a statistical standpoint.

That's an amazing analysis on systematic bias!

The data he has to back up his "too close results to be true random polls" is fantastic.

First, the words predict and forecast are not interchangeable. Polls do not predict outcomes, they merely forecast them. Since predictions are purely subjective, saying they 'failed' is inappropriate. You can disagree with a subjective prediction, but you can't really say they failed. Forecasting relies on historic data to extrapolate. That same data basically said "its a 50/50 coin toss" so the polls did not, in fact, fail. You just thought they failed because the precise poll value was not 50/50, but rather 49.xx/51.xx which does not account for statistical variances.