The whole process of polling is a pile of shit.

I analyzed some polls for recent elections, and they were off by an average of 16 percentage points. They only write that “margin of error 3%” thing because if they were honest about it, no one would listen to their poll, and they’d just use the ones from all the other people who still were lying about their MoE.

The “changed their algorithms” thing is just adding random fudge factors to the dogshit methodology until the answers come out right, but it never makes them predictive at all of the next election, because the underlying methodology is still dogshit. Basically, they call people on the phone, 1-2% of them answer, and then they bark a bunch of questions and whoever stays on the line for the whole thing, they assume that’s a good sample for everyone else and won’t be correlated with any other factors. Which, when you start thinking about it, makes it make sense why it might be off by 16 points.

There’s actually an argument to be made that a change in the polling is pretty reflective of a change in sentiment, even if the overall accuracy might be off by 10 or 20 points, is part of why I posted this. But mostly, I posted it to throw into sharp relief the dishonesty of the overall media narrative “Biden is fucked in polling cause debate,” which is created through a mixture of asking cunningly crafted questions to make it not an apples to apples comparison, mindless journalistic groupthink, and lying.