Yeah, agreed. They've definitely tuned their models well for day-of projections.
I saw some tweets that the data guys are basically saying that their models did really well in 2018 and in the GA special as well - essentially saying that there's a Trump X factor for 2016/2020 that they just can't account for.
One other thing that wasn't really noted much in November because there was so much going on was that Trump's "X factor" was quite different from 2016 to 2020. In 2016, just about every single state ended up redder than polling indicated. The pollsters post-mortems talked about how when calculating Trump's odds, they didn't have the states synchronized enough - IE Trump doing 2% better than expected in Ohio probably meant he was 2% better everywhere.
In 2016 it was a totally different story. Some states (OH, FL, TX) went _much_ redder than the polling showed, but outside of those few states, the polling was actually pretty correct. The RCP average was dead-on in the entire rust belt, Georgia, North Carolina, Arizona, and Nevada. So it was the exact opposite of 2016. TX/OH/FL seemed to decouple from the rest of the states and Trump drastically overperformed his polling there.
Will be some really interesting case studies and postmortems coming out over the next couple of years, once there aren't things like coups and treason keeping journalists occupied.