Among the bullshit wrong predictions about who would win the Presidency or a local election that I thought were suspect:
* Some processor dude (Allan Lichtman) with a bunch of factors who was right in 9?/10 of the past elections. Yeah, given enough professors expressing opinions on the matter (a bunch) someone is bound to be right 9/10 of the time - it's simple probability - regardless of predictive algorithm - well, Lichtman's algorithm was reasonable, but I thought unreliable, because it was ad hoc and because the current election might have been so close-called (it ultimately wasn't) - and in spite of it not being close-called, Lichtman was dead wrong. He's now busy about explaining why he was wrong and presumably updating his ad hoc factors so that he's right next time. Yeah... whatever.
Give me math, statistics, probability and science - insofar as these can be made to apply - or perhaps even AI to study the problem and learn from past data to make predictions - that's what I'd be inclined to trust. Not some random dude who by chance was right in most previous elections.
* The "Gold Standard" Iowa J. Ann Selzer pollster who predicted that Iowa would go for Trump - again, given enough pollsters - there are a ton - someone, by random happenstance, will be accurate most of the time, regardless of how their predictive formula works and regardless of applicability to the current election. Selzer predicted this time that Iowa would go for Harris - way off. My take is that chance caught up to Selzer and that she isn't some guru - though I've not looked into this at all.
One that I thought might prove accurate and ultimately was:
* Bettors. Yeah, yeah, yeah - whatever you want to say against this point. There are two reasons that bettors are better predictors than polls and are probably the best indicator of outcome at least insofar as who will win the US Presidency is concerned or in other big, well-known betting markets: 1) It's not about who you want to win - it 's about who you think will win b/c you have money on the line - so you may vote A but bet on B; 2) betting is crowd-sourced and benefits from the surprising emergent phenomenon that crowds are somehow better predictors of future events than individuals, even than individuals with a track record of success.
Betting odds for regional USA political races - e.g., who will win a particular Senate or House seat - are off recently. Nonetheless, over the years, betting odds have been very good at predicting who will win the US Presidency. This may of course be due to random chance - but I think not. I think that "money on the line" coupled with "crowd sourced" for a huge betting market such as the US Presidential election - has built-in factors for robustness - some of the robustness wears off for regional markets where there's not enough knowledge for outsiders to place an informed bet - so these markets will drift towards "who do you hope will win."
This is, of course, not to say that bettors on the US Presidency will always get it right - I'm arguing instead, that such bettors are likely better predictors than any other currently known metric, at least at this time. And indeed - should a better predictive metric emerge - betting markets should be self-correcting - as the more reliable bettors place more/larger bets and move the market towards increased predictive accuracy.