538’s rating system is really dumb because it doesn’t account for methodological changes and it doesn’t seem to have any recency bias, which it really should. I guess that Marist was a really good poll for a long time, but if it’s been that bad lately it shouldn’t be given much weight. It’s equally absurd that Emerson is an A- even though all its good results came before it started using MTurk as a major part of its process. Also, it’s just absurd that Marist still doesn’t weight by education (I think?)
It really does feel arbitrary sometimes. Particularly not a fan of Silver's split-ratings ("A/B").
We had a former user here that would start every one of his polling posts with "From 538 A-plus, gold standard pollster..." as though that actually carried a ton of weight.
