polling guru nate silver has pretty much given up turning his analysis into election predictions, but had some insight into the 'low information' voter.
('low info' being voters that really do not follow much media but reliably vote, unlike 'no info' voters that falsely claim not to follow media when adverse news appears in low bias sources)

nate claims this low info group is likely to be swayed by whatever final media casually passes them (not literally, but proximally) on the way to vote. e.g., this may be a conservative voter whose final impression is that 'trump is a convict', and then atypically, decides not to vote.

its all plausible, but nate admits he has no idea how many low info voters exist, how many would encounter and retain recent media, and how many it would impact. regardless, expect a lot of ad money to be spent in the final 2 weeks in case this has punch.