Now he believes that the aggregations saying it is toss up (or “close”, which is not actually quite the same thing) are pretty much right.
But previously he was saying they were successfully “gamed” in 2022 (weren’t, were scarily closer than usual) and not to trust them. That NC aggregations were successfully being gamed when the aggregate results were the same when only looking at the few top level pollsters.
Partisanship and optimism does not have to include delusionality.
ETA yes those are good reasons to exclude AtlasIntel, but they do not define them as Red Wave biased pollsters. They are also rated as minimally biased by 538 and have, unfortunately, been historically very accurate. I’m willing to personally believe they are wrong this time and their outsized impact on the results in the Blue Wall state aggregations IS problematic. Unless they’re right.