It’s a deeply flawed poll that has been torn to shreds by critics.

From the poll’s methodology page:

Those in the know, supposedly? In other words, inside the beltway types. Yes they do seem to love that poll, don't they?

No, they don't. The vast majority of recent polls are 1,000 to 1,500 respondents. That range is quite common. Except for Morning Consult, none of the last dozen or so were over 1,500.

https://www.realclearpolling.com/polls/president/general/2024/trump-vs-biden

"we give more weight to respondents from demographic groups underrepresented "

All pollsters adjust for demographic over- or under-sampling.

That was one of my favorite lines from the movie “Network.”

It’s still early. The numbers are already moving in the right direction.

First the statement though many other challenges create additional sources of error is a mathematical truism not some sort of plot. The MOE only measures error that comes from random chance. It doesn't account for error that people lying may cause, the wordings of questions may cause, problems with the selection of the sample may cause, or any other of the myraid of things that might cause error in a poll. It only tells us the error that can arrise from randomness and even then it only does at 95% confidence meaning 1 time in 20, on average, the poll will be even more in error from just that source.

Second, the MOE. There is a standard formula for the MOE. It is (1.96) square root [(p) (1-p) / (n)] where p is the proportion for one candidate, 1-p would be proportion represented by everyone else, n would be sample size. A sample of 1500 would have 2.53 MOE. I get their number using their sample size (and 50% as the proportion which maximizes). In short, their math is correct.

That said, the weighting can be an issue and the cross tabs will have much higher MOE. It is for the cross tabs that polling companies do larger samples so they have enough of the smaller groups.

It's just now become too obvious to hide in biased polling questions. When actual voters have voiced their actual opinions in actual elections, the Democrats have been kicking major ass. The Arizona Supreme Court ruling that came down last week put Arizona in good position to go blue again for Biden, and the crazies in Alabama have struck a cord with women in nearby southern states.
The facts as I see them differ greatly with the poll numbers I've been seen the past few months since Biden started actively campaigning.

Maybe there methodology has changed. That was a headscratcher.