In 2016, the overall polling aggregates were pretty good on the national side.
There was shitty state-level polling if I recall.
In 2020, the error was worse.
One of the major things has been getting the weighting right. In 2016, they didn't weight by education much, for instance, but...