2022 MidTerms & State races- who ya got?

[Spartanforlife4]

Nate Silver with an interesting article on why the polls favoring the Dems may actually be closer to what happens, like in 18, than they were showing in 16 and 20.

https://fivethirtyeight.com/features...mocrats-again/

Of course, then he goes and follows it up with the most neutral, both sides, “idk, maybe they’ll be wrong again” tweet.

https://twitter.com/natesilver538/st...w4lckyDjw_DeVQ

[dxmnkd316]

Nate Silver with an interesting article on why the polls favoring the Dems may actually be closer to what happens, like in 18, than they were showing in 16 and 20.

https://fivethirtyeight.com/features...mocrats-again/

Of course, then he goes and follows it up with the most neutral, both sides, “idk, maybe they’ll be wrong again” tweet.

https://twitter.com/natesilver538/st...w4lckyDjw_DeVQ

The polls weren't wrong in 2016. The polls were very close. The problem was we have a winner-take-all system where a single vote can change an entire state's electoral votes. 40,000 votes swung the entire 2016 election. The pundits didn't account for that very well, which is why the deluxe and classic models were so far off but the polls (lite) was well within the margin for error.

[Kepler]

The polls weren't wrong in 2016. The polls were very close. The problem was we have a winner-take-all system where a single vote can change an entire state's electoral votes. 40,000 votes swung the entire 2016 election. The pundits didn't account for that very well, which is why the deluxe and classic models were so far off but the polls (lite) was well within the margin for error.

Thank you.

Silver: Trump has a 30% chance of winning.

< Trump wins >

Morons: LOL538!!!

Silver: God you people are stupid.

[dxmnkd316]

Thank you.

Silver: Trump has a 30% chance of winning.

< Trump wins >

Morons: LOL538!!!

Silver: God you people are stupid.

Silver's no saint in this. He was the least wrong of the pundits and his agglomeration of polling is good. He should have left it at that and stayed off Maher and the other talk shows. If he ditched the stupid deluxe and classic models it would be better. It's no different than massaging the data. Just let the GD data speak for itself.

Lite should be: Unadulterated Data
Classic should be renamed: Moderate ****ery - Pollster grades, size of polls, LV vs. RV, basic stuff.
Deluxe should be called: "Here's what I think the data should look like"

There was a great tweet i saw the other day:

whether you should trust the lite, classic, or deluxe 538 model:
if you believe that there are variables not present in other elections, lite;
if you believe that this is roughly the same election as 2014, classic;
if you believe that the punditry to date has been good, deluxe

It's perfect.

[Kepler]

CNN not entirely Nazi yet.

[ScoobyDoo]

The polls weren't wrong in 2016. The polls were very close. The problem was we have a winner-take-all system where a single vote can change an entire state's electoral votes. 40,000 votes swung the entire 2016 election. The pundits didn't account for that very well, which is why the deluxe and classic models were so far off but the polls (lite) was well within the margin for error.

The polls may have been very close but the entire media and pollsters reporting was Clinton is going to win. Only a few outlets and pundits tried to warn everyone that Trump was a good possibility.

[dxmnkd316]

The polls were close. The entire media and pollsters reporting was Clinton is going to win. Only a few outlets and pundits tried to warn everyone that Trump was a good possibility.

Fixed!

[ScoobyDoo]

fixed!

lol. You can keep trying to pretend it wasn't like I said it was but that was what was being reported and polled. I lived through it.

[Handyman]

Silver's no saint in this. He was the least wrong of the pundits and his agglomeration of polling is good. He should have left it at that and stayed off Maher and the other talk shows. If he ditched the stupid deluxe and classic models it would be better. It's no different than massaging the data. Just let the GD data speak for itself.

Lite should be: Unadulterated Data
Classic should be renamed: Moderate ****ery - Pollster grades, size of polls, LV vs. RV, basic stuff.
Deluxe should be called: "Here's what I think the data should look like"

There was a great tweet i saw the other day:


It's perfect.

This is a pretty good breakdown of what the (non idiot) critics of Nate Silver say. Then he usually opens his mouth and the nimrods go hyperbolic about it and troll him which is fun to watch but not very helpful to the discussion. (then his defenders dial it up to 11 in defense and away we go!) He was always better when he just put out the data and left the ****ery out of it. The lite version of the data is the only one that should be used for real discussions. The rest should be for hypothetical style analysis.

There are other problems with certain aspects of his methodology (according to nerds who understand this more than I ever could) but there is no system that is ever going to be able to figure this out or predict with any real sense legitimacy because the baseline data is in and of itself flawed. Polls have bias, sample sizes are small and often random, opinions change on a dime for sometimes weird and unknown reasons, the polls are conducted stupidly or at the wrong times, people lie...etc. Short of someone literally polling everyone in the US of voting age multiple times over the course of multiple years asking dozens of questions there is no real way to fix that.

[dxmnkd316]

lol. You can keep trying to pretend it wasn't like I said it was but that was what was being reported and polled. I lived through it.

You're confusing polls with pundits. Pundits use polls for whatever they want to say and we willingly consume it. The polls are what they are. You can go back and look at them. I did a couple months ago and they were shockingly close. They missed that small fraction of people who were undecided and those people broke to Trump near the end. Not sure we know why they broke to trump so late. It's a complicated oddity, muddled, even yet.

[dxmnkd316]

This is a pretty good breakdown of what the (non idiot) critics of Nate Silver say. Then he usually opens his mouth and the nimrods go hyperbolic about it and troll him which is fun to watch but not very helpful to the discussion. (then his defenders dial it up to 11 in defense and away we go!) He was always better when he just put out the data and left the ****ery out of it. The lite version of the data is the only one that should be used for real discussions. The rest should be for hypothetical style analysis.

There are other problems with certain aspects of his methodology (according to nerds who understand this more than I ever could) but there is no system that is ever going to be able to figure this out or predict with any real sense legitimacy because the baseline data is in and of itself flawed. Polls have bias, sample sizes are small and often random, opinions change on a dime for sometimes weird and unknown reasons, the polls are conducted stupidly or at the wrong times, people lie...etc. Short of someone literally polling everyone in the US of voting age multiple times over the course of multiple years asking dozens of questions there is no real way to fix that.

Yeah, I think they operate on the same theory I operate on, "With enough information, nearly anything is predictable to a reasonably certain degree." THe disagreement comes, as you hinted at, what constitutes enough info?

I think they fall in love with their model and overestimate its infallibility. If you just say there's a 95% chance the range falls between x and y, that's all you need. You can even add the comment, if the 95% range is 45-51, that 51% is a big ****in deal if you're Florida or Pennsylvania.

I think if I were Nate, I'd show more of the data. Overlay the bell curves for the states (lite model!) and run the monte carlo showing several of the actual state results. Ok, so Clinton wins in this result because Florida looked like this <----H-->, PA looked like this <---T--->, etc. Show the national result but show the individual state bell curves and how the simulation's result led to the 330-208 result. Or don't. Just show each state's bell curve and don't make a national one.

I really liked the NYT, and (patting myself on the back) the live model I made during the 2020 election. I said, "Here's where we're at with known votes. Here's how much Trump would have to outperform the rest of the votes to win." That moving target, similar to the NYT gauges, gave you a time-based indicator of the momentum, the state of the vote, etc.

[Kepler]

Um.

[Deutsche Gopher Fan]

Dusty old bish can’t even rhyme

[dxmnkd316]

https://twitter.com/proptermalone/st...9hoMp6wWDlNXVg

trafalgar has some interesting artifacts in their data

[Kepler]

The only decent ads for Biden are STILL being done by Republicans.

I don't know if it's the bunch of 80-year olds who have run the DNC for 30 years, or if it's ANY bunch of 80-year olds, but good god the Dems should defenestrate every person who has worked their campaigns whose name is not Stacey Abrams.