This has been similar to 2012's presidential election with one exception -- the last month of 2012 was a dead heat -- a problem we don't seem to have this year. If we assume the sine wave pattern will continue through election day (Nov. 8), Donald Trump is running out of time. Looking at the RealClearPolitics polls for both 2016 and 2012, or even 2008, we see a consistent sine wave.
The distance between close race numbers (wavelength) was about 30-40 days for 2012, and a decreasing wavelength in 2016 from 90 to 54 days. Extrapolating the 2016 trend indicates that we'll be back to a close race within 30-52 days -- which is exactly how much time we have left before the election. I think it's possible we may end up with close polls on election day, just like we did in 2012 (but not in 2008), even though the current trends and news don't support that. You can even see the sine wav in the Predict it gambling website, though the data doesn't go far enough back to find a wavelength.
As I stated last time, I do believe this could be a year that the polls don't match the voting -- but I can't tell whether that will favor Clinton or Trump camps -- both may come out in high numbers to "vote against the other candidate." Add that to who stays home because they hate both candidates, and this could get interesting.