The basic reason the polls failed was upstream from averages generated by 538 et al. Those polls tragically expected a repeatable electorate ("everyone who vot

The basic reason the polls failed was upstream from averages generated by 538 et al. Those polls tragically expected a repeatable electorate ("everyone who voted in 2012, less the dead and plus some fraction of new voters, will turn out"). Arguably, the shy Tory effect was at play as well. We shouldn't celebrate poll averaging as some data science genius; it is only as good as the polls being averaged.