Models also aren't measuring the randomness of the electorate. They measure the variance inherit to the polls.

All,

I have a new theory that our country really did not want Donald Trump, but we’re too embarrassed to tell their friends/colleagues that they were supporting Joe Biden.

Hence why Harris is beating Trump in the polls.

-RS

Maybe, but I know several people that dislike trump a lot, but didn't care for biden either. Sort of a hold your nose situation. My very small sample size is excited about kamala.

538 also has sports models which can't beat vegas - not that that's an easy task and they should be expected to do that, but more shows their arrogance to think they could do that and not only figure out how to unlock untold millions but also do it casually, then the model kept on losing slowly and instead of scrapping it or finding the leaks and adjusting, they just left it alone figuring it'd eventually start winning - as a result, 538 has gotten a big reputation for having garbage in garbage out methodology and just rolling with it

my theory was they believed themselves to be infallible largely in part to the astonishing amount of praise they got early on


the main issue with their polling stuff is they didn't know how to properly value polling data, they just took it raw and shoved it in there

this will work well enough to trick yourself that you have a good forecasting model at times, but it's ultimately just guessing

before they ultimately shut down 538, disney did make an attempt to fix all his flaws that he was too stubborn to change by removing him from the team

and of course one of the first things the new people do is revamp everything and even throw a little shade at nate disguised as a compliment

https://fivethirtyeight.com/features...ling-averages/




predicting hillary at 71% was just absurdly off


you can look at the polling data and see it was a neck and neck battle with about 10% of the voting population being unknown


it was very clearly going to be a variance fest just based on who was going to show up to vote and who wasn't - and importantly, how many of the declared 3rd party voters would get cold feet and decide not to "throw away a vote" and choose hilldawg or donnyboy at the last minute - then if you factor in that a lot of the people responding with "i'm undecided" actually voting for trump but too embarassed to say it out loud along with the virtue signaling ev of declaring you're on team hilldawg and then he's a clear favorite despite that the polls indicate he'll win

yes, hillary was likely 71% favorite if A) the polling data were infallible and B) an exact representative of respondents showed up but neither of those were true

clinton should have been the favorite without a doubt (the undecided people were secretly voting for trump is from hindsight and would be a dangerous assumption to make before the election) but it should have been a smaller one, like 58% of so, still a massive favorite but not the bulldozer that nate silver predicted

but seeing the polling data above vs the actual turnout, it's clear that they were pretty accurate on hillary with no real surprise that she got 48% of the vote but had all severely underestimated trump

and... ironically, his prediction likely led to more trump people voting and fewer hillary people thinking it was necessary

Trump's 2PM press conference today is likely to be the sh*t show we've come to expect from his recent desperate ramblings. You know what they say about finding yourself in a hole with a shovel... Nah, he'll keep digging.

I think the response to Kamala shows that a lot of people are still very worried about a Trump second term. They just needed a candidate who they could argue with a straight face was up to the job.

Kamala also has come off as energetic and engaged since the switch, which was much needed and a sharp contrast to Biden. I suspect that she was expecting more resistance among Democrats than she received and feels very buoyed by how quickly the party coalesced around her as the nominee. Without question, this has been the best three weeks of her political life.

I also think Walz stumbled onto something useful when he started calling Trump and JDV weird. I expect to hear that refrain all the way to November.

Those polling numbers aren't neck and neck, Hillary was absolutely curb stomping Trump in polls

Its like calling a 55/45 a coinflip. Maybe in a vaccum thats kinda close to 50/50, but over 100 million sample.....

Yes, lack of independence both in that some future information that makes a candidate tank in MI will very like make them tank in WI as well as if polls are off to begin with in PA that increases the chances of them being off the same direction in MI etc.

you're eating red herrings here

you forget those are national polls and given the surplus of votes in NY and California it's always that way

Only one republican has won the popular vote since 1992 and that was Bush getting his 2nd term with the 9/11 boost (it was early but he was unlikely to win re-election prior)

we are excluding time prior to that because Nixon and Reagan were from California and thus won those states and then Bush Sr got a bit of the Reagan/California boost and won that state too - he probably had a decent shot at it in 1992 but Perot got 20% of the california vote that year

1992 Clinton +5
1996 Clinton +8
2000 Bush -.5
2004 Bush +3
2008 Obama +9.5
2012 Obama +5
2016 Trump -3
2020 Biden +7

if it were a popular vote and not an electoral one then she should have been like 95% to win

If the contest was on a popular vote basis, policy platforms, political communication, and potentially even candidates would be different, so it's disingenuous to say "she would have won 95%+ if".

For republicans it pays off to enrage 3 million coastal residents to gain 300k voted in the Midwest basically, for example

yup, no disagreements there

we'd get remarkably different candidates as well as party platforms if it were just about the popular vote

Yeah, what’s funny is everyone gets mad when a 2:1 dog ends up winning, but nobody will ever get mad if a 2:1 favorite wins 4 times in a row. Media pundits aren’t really rational in evaluating these models. I think 538 was probably close to correct in 2016 given the stuff that lined up for Trump; direction of polling error is not generally predictable and happened to go his way, Midwest tipping point states happened to be slighter redder than the country as a whole after being bluer in the Obama years etc.

the electoral college is for example, the only reason why we remain staunchingly anti-Cuba and maintain our policy of embargo and isolation vs them

we don't care about the politics anymore but the cuban american vote is large enough to possibly swing florida and they are staunchly anti-Castro so nobody will dare dissolve that outdated policy out of fear of upsetting those voters and turning them against them

however, there's hope for reform as each year florida drifts further away from being a swing state and thus the unreasonable amount of influence that small voting bloc has on national policy will get weaker over time

Maybe, but I'm sure there are more guaranteed Trump voters hiding among those polling as undecided

what'd drumpf say?

Caved and will show up for ABC debate + 2 more not confirmed yet

The sensible decision, Harris would beat him silly over it if he punked out, and IMO she's not anyone I'd be particularly afraid of debating anyway. Not to say she can't have a good showing but in a close race where she may be edging ahead he can't skip it. I don't know about the other two, if he's still angling to get one on Fox or something

desperate times call for desperate measures

I am not sure Cuba is the right example.

The clear example for me looking at national polls is federally mandated maternity leave. It clearly polls at the very least in the high 50s, easily in the 60s depending on the exact question asked (and I saw it as high as mid 70s), and it's a blatantly "we don't do this because quirks in the Senate yet" policy.

Why? Do you believe there's something incorrect in his example?

Yes, Florida is fully republican today not a purple state anymore.

Cuba might have been a reasonable example when Florida was very tight rep v Dem, not today.

What I mean is democrats could reverse all cuban policies without any cost for them in the electoral college today, and they had a trifecta recently and decided not to do so

But aren't those wildly different, at least really close to the election? Unknowns in football are variance during the game, in polling it's lack of data and accuracy of models. If you could do perfect polling with perfect model for interpreting the results in the morning of the election, you'd be right extremely often. Can't do the same with football even with perfect information (unless you make a perfect call to the coach)

Man that press conference was some kinda something even by trump standards

I guess it comes down to the nature of what the % probability is measuring. In some cases, it's measuring absence of information which would lead to a perfect prediction, in other cases, it's measuring the fact that even with perfect information, there is an element of variance involved, and in other cases it's a combination.

For example, we know that the probability of the bottom card at the bottom of a freshly shuffled deck being a spade is 1/4. No amount of additional information (assuming we are not allowed to examine the composition of the deck in any way) is going to improve that estimate. It will always be 1/4.

On the other hand, we get to the river and villain jams. We suspect he is doing this 2/3 of the time for value and 1/3 of the time as a bluff. However, we notice that he took a bite of his oreo, so we improve our estimate to 3/4 value 1/4 bluff. There might be further information available to us that helps us improve our estimate even further.

I don't know if I've explained what I mean very well, but I guess you're saying that the football match situation is more like the first example and election polling is more like the second?

I thought it was relatively good. He looked good, sounded good. Just some minor unhinged rambling here and there. Seemed to have told a below average number of lies

Yep pretty much what I meant, if bad card is 25% to come and it hits not much there to do, if a model says candidate 75% for X to win and they lose, there might be bad modelling involved or missing information that could have been used or obtained. Wasn't really getting to any real message there, just that I don't think "it was 3 to 1 so the favorite will lose quite often" isn't imo that helpful mental model to rate the model accuracy in a vacuum as the odds themselves are partly a reflection of the lack of accuracy of the models and information and not some underlying variance or hidden undetectable information in the world (while obviously perfect modelling and information is impossible)

Yeah I was expecting more, proud of Donny there how while getting his ass kicked bigly he hasn't really gone off the rails as badly I thought he'd be. Yet.