But, as said upthread, they predicted downballot elections in 2016 pretty well too. So unless there's been an improvement in downballot predictions - not just good absolute performance - why should we expect any difference in 2020?