https://www.270towin.com/2024-presidential-election-polls/
Those polls, several of them, do not appear to qualify for the RCP average. There are fairly strict requirements as to sampling with a consistent turnout model.
This morning’s CBS News poll, in apparent keeping with RCP requirements, explicitly noted in their text that they oversampled blacks in comparison to their previous sampling for their turnout model. This moved them from T+2 to H+1.
But if you do not maintain your turnout model constant, you have not really taken a measurement that can be compared to your previous measurements.
For the zillionth time, it does not matter what a poll says because they have turnout models that may or may not be correct. What DOES matter is a change in a poll from week or month to month, holding the model steady. If you that, you can measure change. If you don’t hold it steady, you are not measuring attitude change. You’re just showing what a turnout model change does.