ButterflyEffect

I've been scratching my head on Virginia for a while. I think it's because very little vote from Richmond, Norfolk, and Virginia Beach have been counted so far. But I wouldn't call it yet like NYT has.

For some reason CNN hasn't called Florida for Trump yet. Those votes in Miami-Dade aren't coming, dudes.

I guess if Biden camp is disappointed about florida, they should be excited about ohio. Still early, but early returns look decent.

It's too early to tell. He's doing very well and could win Florida. That's about as much as you can take from it so far.

I mean, isn't that the case in most states? Many more rural, conservative counties than urban?

Jeeze, Florida results are coming in FAST

My final projection. I think it'll end up being tighter than polls predict, by a few percentage points. Florida could go either way in my opinion.

Of course no Miami-Dade data is skewing numbers somewhat, but spots like Pinellas are going to tell the tale of Florida today.

How the NPA's split is going to be huge.

Initial numbers out of Florida are very favourable for Republicans. Hold on to your butts everyone.

Glad to hear that you've both voted and that it was painless. Today's going to be a loooong day. I've only been in the office for an hour, but it feels like 12 hours already.

VA has been mostly ignored by good polling agencies. I guess, much like the candidates, they don't expect it to be at play. Best poll is the Washington Post +11, but it's 2 weeks old. VA flipping would indeed signify a catastrophic polling failure, or interference in the election itself.

Final words from my actual favourite aggregate & forecaster, Philippe Fournier of 338Canada. This is his first US election, so time will tell how close he will be. His models have been excellent in Canada, though Canadian elections are a lot easier to predict than the USA, even with less polling. https://www.macleans.ca/politics/washington/a-final-338canada-u-s-election-projection-where-the-race-will-be-won-or-lost/ In particular, this is a pretty good self-aware excerpt: Before concluding: Regular readers of this column know that I usually stick to the numbers and focus my analysis on historical precedents and what the data indicates. Nevertheless, I am inclined to add that some of these states projections do not reflect my personal gut feelings. These are difficult to handle. Should the data be off by a few points in the direction of what my instinct had in mind, I would naturally go “Oh, I knew it.” But the reality is: I didn’t know it. I was just expressing my biases—which can sometimes turn out to be correct. During my university studies, I worked as a croupier in a casino to pay rent and tuition. Night in and night out, I saw people who just knew it right after it was too late. I witnessed the extremely irrational and very human behaviour of trying to see clear patterns out of pure randomness. Yes, hitting on 16 when the dealer shows a five in Blackjack is playing against the odds, and players who did just that would regularly infuriate people around the table. Most times, they busted and lost their money. But I clearly remember times when a player would then get a five to hit 21—and win a hand he would have lost otherwise. What would the player say then? “I just knew it!”, but he didn’t. These numbers above are riddled with uncertainty. Polls have uncertainty, and projections don’t always hit the mark. Trying to forecast an election without polls and going with your gut feeling may get you some praises when you are right, but, as with blackjack, you are most likely to lose in the long run. Studying scientific polls is like counting cards at blackjack: you will lose some hands, but you will win more in the long run. This is why the 338Canada model has correctly called the winner in 90 per cent of electoral districts over six Canadian elections. Obviously, these projections are based on voting intentions. Should actual votes not be counted or should unexpected complications occur at the voting booths on Tuesday, we could be in for a few surprises. Unfortunately, these factors cannot be easily quantified—at least not before voting day. Let us hope for the sake of all Western democracies that the will of American voters be accurately translated into actual, counted votes.

Margins in PA are thinner than the other rust belt states, so its completely possible. 538 has PA as Biden +4.8, RCP has it as Biden +2.9. If the state level polling there were off as much tomorrow as they were in 2016 that becomes an insanely close race. FL, GA, and NC being won by Trump are completely reasonable predictions too. You keep saying 1,000,000 views like that's supposed to be impressive. I'm not impressed by 0.3% of the US population having watched it. Partisans will flock to partisan media outlets, and while I'm not sure exactly whatever "Right Side" is supposed to be, I can tell you that it's biased. All media has some form of lean bias, but what you've linked doesn't even try and hide it. Only an idiot would think there aren't 1,000,000 conservatives in the USA. So what exactly does this prove exactly? Sure, Trump could absolutely win tomorrow. But I don't think 1M views on a partisan video or a large group of people at a partisan rally are proof that it's coming. Also I'm not censoring you, you can watch whatever shit you want. That doesn't change how I feel about it.

I don't mind if things are super close in a swing state or two where a recount could dictate the results of an election. I don't want to see shenanigans where the results are very clear and we see interference in attempt to prevent votes from being counted.

"Right side" so I assume they're an amateur, conservative-leaning outlet. So am I surprised that conservatives flock to it, watch it, and like it? Echo chambers are weird.

Polling for primaries has been known for quite some time to be really inaccurate. To the point where I'm not even sure what the point of primary polling is for. Polling gets worse the further down the ticket you go, with presidential polling be the most accurate, and a drop off below it. I haven't paid much attention to house or senate races at all for that very reason. Pollsters are predicting a Democrat senate victory, but I personally don't buy it. Great excerpt from my new bff Nate regarding primary polling:

While your examples are pretty hyperbolic I agree in a sense. Not a lot of people think of polling as a science, but it really is. The further off the beaten track you get towards the fringes of the left and right there is a strange distrust towards science.

5% is a huge MoE too, probably didn't sample nearly enough people. The poll isn't wrong, but you'd feel that had they found a way to sample even 250 more people that they could have likely had a better result (I have no idea the number of poll'd for that particular poll). Sample today was only 502 likely voters so I'm guessing MoE is similar. They did do pretty good in 2018 though. If all the polls had a 5% MoE then Biden would be in major trouble. Even seeing Biden +5 averaged in Pennsylvania has me a bit concerned regarding that race. I did read a good article this morning about how polling agencies changed their methodologies between 2016 and 2018-present. I sort of figured that they had to account for the errors in 2016 but I didn't know how, nor where exactly the errors arised from. Turns out it was because many of them didn't weight responses by education. Non-college educated whites were undersampled. Agencies are now weighting by education so this shouldn't happen again, unless of course something else in the weighting is out of whack now. https://www.cbc.ca/news/politics/grenier-uspolls-trust-1.5765695

I don't place any weight on rally size or mass gatherings size. It's the equivalent of looking at 20,000 people in an arena watching a concert and proclaiming that they've never seen the city as a whole so enthused about the band before. Those people want to be there. There are large number of Trump supporters and a large number of Biden supporters everywhere. Even a large crowd of say 25,000 people in Los Angeles only accounts for 0.6% of the cities population. --------- Going to be a busy day for polling data (I think). Too bad I'll be at work missing most of it. The only thing I've seen so far is a Biden +5 in Pennsylvania from Monmouth. Monmouth is pretty dang good, I'd like to see them release some more state-level stuff today.

I'm neutral on Nate, though I do appreciate his very in depth work to establish polling track records. I also appreciate that he isn't very sensationalist. Sure he's likely Liberal, but I don't think it impacts his work much. I've seen a lot more partisan aggregates, like those Lean Tossup clowns, absolutely rip him apart for including republican-biased polls in his model. I get it, polling agencies also have house effects, but if its minor you need to include it unless the polling methodologies are completely flawed. In 2016 when nearly everyone else was giving Trump 10% or less odds on winning, I appreciate the fivethirtyeight 29%. Nearly 30% is actually dang good odds. He has Trump at 11% chance of winning right now.....not as good as 2016, but also not inconceivably small that there's no path to victory.

So was your post with that Florida result supposed to be your gotchya moment? Where I concede that Trafalgar is actually good and that all other polling agencies are trash? People a lot more invested in US politics have painstakingly analyzed agencies track records, polling methods, and sample demographics to determine whether they are "high quality" or not. And that tweet is false. Nate Silver had the odds of a Clinton victory at 71% (Trump 29%). He was also the only person with the balls to warn people that a Trump victory wasn't super outlandish. All it took was a standard polling error.....and that's exactly what happened. https://projects.fivethirtyeight.com/2016-election-forecast/ And his final post about the election where he clearly states that there's a lot of uncertainty and that standard polling error could completely eliminate Clinton's lead: https://fivethirtyeight.com/features/final-election-update-theres-a-wide-range-of-outcomes-and-most-of-them-come-up-clinton/

I don't know, you tell me what you think of Trafalgar

New high quality state level polls from YouGov/CCES Florida: Biden +2 Georgia: Biden +1 North Carolina: Biden +4 Pennsylvania: Biden +8 Texas: Trump +2