I'm not necessarily sure that "polls" like that are so useful, as they are opt-in rather than sampled by the polling company as a (hopefully) representative sample.
As a case in point, here's a recent poll from the Telegraph on Brexit:
http://www.telegraph.co.uk/news/2016/09/20/nigel-farage-needs-to-stop-telling-me-why-i-voted-for-brexit/
As I type the figures have balanced out rather, but early on in the polling the option "I'd really rather we remained" was topping something like 70% or so of the vote. Do you believe this is a representative sample? Of course not, and nor should you. But some pro-EU groups saw the poll and took the opportunity to express their opinion and get friends and acquaintances to do so too -- weighting the pill heavily in favour of the participants.
Bad sampling gives bad results. I'm not saying Trump didn't win the debate, but online polls hosted by news site where you opt in are the most meaningless form of survey ever and deserve to be binned.