Post Election debrief by Atlas Intel

[qrdlelections]

https://www.youtube.com/watch?v=Xg5jqiQhfEs

Must see youtube.   My favorite bit was how they discussed the polling error caused by men versus women doing the interviewing. 

AI uses 'digital recruitment', ie: iphone app.

Whatever they are doing, it clearly is working.

[ProgressiveModerate]

Briefly skimmed through the podcast - have to admit they're less hackish than I thought they would be and made some decent points about different types of biases. One bias they talk about that hasn't really been discussed is who the interviewer is - for instance depending upon whether the interviewer is a man or woman some people may give different responses.

The only thing is they keep claiming they were the most accurate pollster in 2024 and 2022, but weren't they like really off in 2022 (mostly overestimating Rs)? They're 2024 polls were eerily accurate - curious to see how they hold up 2026.

[qrdlelections]

They are not hacks, they are totally for real.   They are likely using big data and other inference techniques collected on people with their digital recruitment which makes them far superior to a lot of old school pollsters.  

Think SurveyMonkey but serious AF scientists.

Silver and Cohn are basically having a meltdown as this is likely the end of the aggregators, imho.

https://x.com/NateSilver538/status/1858882128957550891

That said, RCP had yet another truly great year.   They've beat out 538/silver every year since 2016 that I've looked at with a dead simple model.  

I like RCP a lot because it doesn't overly rely on Atlas Intel and their opaque method for truth, even if it might be superior.

However, my guess is there will be a lot more AIs going forward as they all realize this is the way.


Here is a summary of their historical results - https://www.reddit.com/r/YAPms/comments/1gf4bub/atlasintel_polls_vs_final_results_2020_2022_and/

I think they will be even more accurate in 2026, but will have more competitors.

[Lacking All Conviction; Filled With Passionate Intensity]

They're going to have the reputation of being "R hacks" on this site for 2026 and 2028 no matter what, just like how RCP underestimated Trump in 2020 and still gets called a hack outlet. Results are forgotten; perception of bias is permanent.

[qrdlelections]

Oh yeah, there will always be people who are biased and ignorant for sure.  

Trick is to track down and only listen to the people who care about the facts and for whom the partisan crap is just noise.

Imho, I find the xenophobia around immigration to be silly, but I also live on a very white island.   So what do I know, right?  

If people are saying they care about immigration it's probably because they care about immigration.  Doesn't make sense to me, but last time I checked the US was still a democracy and I'm not a dictator.

[JohnCA246]

I was wrong about them mainly due to unfamiliarity. I conflated them with a bunch of other R pollsters that popped up.  I will totally take them seriously henceforth even if they are wrong next time because it just means they messed up and are not simply just saying their team will win. I will not take groups like Trafalgar seriously because they really seem to be homers saying their team will win every time. Atlas is very legit.