The tough thing about writing cohesively about election polling in this cycle is that any piece needs to push back simultaneously against incorrect assumptions in a bunch of different directions.

Have to emphasize at the same time:
A) Polls aren't irreparably broken
B) Polls aren't precision instruments

A) It's not probable Trump will win
B) It's not unthinkable Trump will win

A) Race has remained very polarized/stable
B) Small movements could have big consequence

A) Non-response and other sampling issues are a real concern for pollsters
B) "Shy Trump voters" probably aren't

A) Methodology matters for polls' credibility
B) There's not one silver-bullet gold standard method that's guaranteed to get it right, or even reliably outperform

A) Horserace numbers aren't the only political polling that matters
B) Horserace numbers aren't the only political polling that matters

(...wait)