I am a sturdy fan of Bayesian inference. This means we get to think like grown-ups, taking recent experience into account, rather than treating every event as if it were a tidy disjoint probability.
The same electorate and issues from 2016 are very much with us in 2018. That is our most recent data point. And since the politicians, pollsters, and pundits got it so wrong in 2016, their poor reliability is also the most recent data point.
So I feel good.