The problem is, the general election is high turnout.
here is an âacademicâ example, i.e. that does not look like anything like the presidential election, and is not in terms of âmodel vs realityâ but just 2 different probability experiments.
Imagine you do one of the two following things :
-
you have a dice with two sides labelled 1 and 3. you roll it 100 times and your result is the sum.
-
you do the same thing, but with a 8-sided dice, with labels (0,4, and 2 on the other 6 faces).
In Both cases, the average of one dice is 2, so the law of large numbers tells you that the result should not be too far from 200.
Also in both cases the result of one dice has the same standard deviation (=1), and same goes for the sums (=10).
So the central limit theorem tells you that in both cases, outcomes a few standard deviations from the mean will behave the same, and will be close to Gaussian probabilities.
For instance, the probability that the result is between 190 and 210 should be, in both cases, about 68%.
However, the further you get away from the mean, the less true this becomes. (but of course, probabilities are small in both cases).
For instance : in the first case, the probability of being >300 is 0 (obviously), while in the second case, it is very small but positive.
(Note : I assumed that 100=10^2 is close enough to infinity for CLT to be applicable, thereâs a chance itâs not true, but for sure if you take 10 000 rolls (and replace numbers accordingly) the above should be correct. Also I donât have time to do it but itâs probably easy to run a code that does the above, plot histograms and compare the two distributions.)
1 Like
Seems like weâve had several election where an unelectable guy simply tried to but an election with massive funding and failed. There seems to be a limit to how many votes you can win outright just with ad spending.
He was a state sen, not an outsider with big money, vs a guy who spent nothing. Definitely interpret it as an engagement issue. If engagement is significantly asymmetric in Nov. then Trump will lose. Canât trust registered voter polls, need likely or certain voters.
In a reversal of one of the most familiar patterns in American politics, it appears that Donald J. Trump, not President Biden, would stand to gain if everyone in the country turned out and voted.
In New York Times/Siena College polls over the last year, Mr. Biden holds a wide lead over Mr. Trump among regular primary and midterm voters, yet he trails among the rest of the electorate, giving Mr. Trump a lead among registered voters overall.
this should privacy go in the Poor Media thread tbh
1 Like
A truism of American politics is that older voters prefer the Republican and younger voters prefer the Democrat. But the old rules donât seem to apply in this yearâs presidential election, where both candidates are old men and their fitness to serve is a top issue. Older voters are gravitating to Biden, and younger voters are taking a look at Trump.
Despite the shift of older voters in the direction of Biden nearly three-quarters of registered voters ages 65 and older said Biden was too old to be effective as president compared with less than half who said the same about Trump. Those figures tracked with the population at large.
âJoe isnât one of the most effective presidents of our lives in spite of his age, but because of it,â first lady Jill Biden said at a campaign stop in Wisconsin this week,
It really says something about American politics that these are our two choices. In a way, I donât see how Trump doesnât win with 1/3 of the electorate having a cultish adoration for him and the other 2/3 not liking either oneh
Stand selling Biden buttons and magnets outside of Starbucks. I think that will really help to Jordan Peeleâs Get Out the vote. Sorry for bad pic.
If you were building models to evaluate the chance of a candidate getting 475 EVs, youâd build different models. Youâre missing a key point if you think that a modelâs point estimates for every single outcome should be accorded the same weight. Of course they would go broke if they had to use a single model to predict 50 outcomes and take unlimited action in either direction on every single outcome. Itâs a virtually impossible task.
Very likely these âsimulationsâ arenât just Monte Carlo rollouts, they are likely tuning some model parameters across a range of values. If you took the best estimate for each parameter, you probably get Trump with a ~0 percent chance of winning NY. In some sense, thatâs a better model than the one maximally distorted model that spits out a 450+ EV landslide. But at the same time, any one setting for the model is as arbitrary as any other. The point of the exercise is to see how the prediction varies across reasonable ranges of the modeling parameters.
For greater clarity, imagine a black box that has 10 knobs on the back, each labeled âHIGH,â âMEDIUM,â and âLOW.â The black box is a perfect oracle, and when you set a combination of knobs, it tells you the exact results for all contests of interest in the universe where the respective properties of the universe match the selected values. If you tune all the knobs to LOW, Trump wins 491 EVs, taking NY by 0.5 points. On other settings, Trump loses NY.
What should you do with this box? Unclear, but one thing you certainly shouldnât do is assume that it tells you that Trump has a 1 in 1000 chance of winning NY. If you wanted a box to estimate that outcome, the current box is a piece of shit! At a minimum, youâd want your knobs to have gradations 1-10 so that you can see exactly how close to LOW you have to get before you see the outcome of interest. Quite possibly you want different knobs entirely. But the current box doesnât really help you. That said, itâs still quite informative to sweep through all the possible knob combinations and see what happens, and for outcomes where you see both results frequently, you can start to draw some quantitative conclusions about how likely those results are.
https://x.com/ChrisCillizza/status/1802729147506032672
Very surprised that Trump is agreeing to show up, feels like a real gamble.
He 100% thinks that he will mop the floor with Biden in any debate. No matter what happens, he will a) believe he won and b) claim it was rigged.
Plus it will allow him to get out of jail for the day.
Trump himself has likely not seen these rules (yet) and thereâs still plenty of time for him to back out.
1 Like
When will Trump back out of the June 27 debate?
- More than one week before
- 5-7 days before
- 3-4 days before
- 1-2 days before
- Day of the debate
- ITâS HAPPENING
0
voters
He probably should back out simply because heâs in the lead and every day where nothing changes is another win for him, but I doubt he will because of his ego and he thinks Biden has full blown dementia. That said I donât think a debate in June really matters for anything barring a cataclysmic meltdown by one of them and it will be long forgotten by the time voting starts.
Jesus I did not realize the Atlanta debate is in nine days.
Iâm worried that Biden will be underprepared because he will have been no-lifing Shadow of the Erdtree during the week leading up to the debate
10 Likes
In nine days Sleepy Joe will still be trying to find Mohg, smh.
3 Likes