Anything's possible, but I think the best chance for Trump right now:
1. A period of deep voter skepticism of Clinton - these windows happen for two weeks at a time, max
2. Extreme voter forgiveness of Trump - these are also two-week windows
3. These two coinciding at the exact same time
4. Election Day falling during this coincidence.
Number one has happened three times now - Comey's remarks, the wikileaks DNC emails, and the September slump. I calculate this represents about 20% of this election cycle since April. Which means if we have another such "event," it represents a probable 20% of the remaining days.
Number two happened three times: when Trump officially won the GOP nomination, when he got his convention bounce, and when he "went teleprompter." I calculate this also represents about 20% of this election cycle since April, and another such "event" taking up a probable 20% of the remaining days.
Coinciding events would take up 0.2 * 0.2 = 4% of the remaining days of the cycle. There are 33 days left. Which means basically one day left, probabilistically, for this to occur. And that one day left would have to be November 8. Which means the probability of this happening is 1/33 or about 3%.
Even if this happens, it's probably only enough to give Trump a slight edge. It gives him Iowa, and Ohio, and shifts the battleground to CO, FL, NC, PA, MI, NH, and possibly WI. Clinton still has many paths to 270, especially if she wins Florida or North Carolina.
Your statistics are incorrect.
If each candidate's event causes a change in the polls lasting 14 days and 20% of time is during the boost after such an event, then the probability of an event on an arbitrary day is (1/(14*5)) = (1/70). The probability of an event occurring in the last 14 days before the election (when the polling boost would matter) is 1 - (69/70)^14, which is 18.2%. The probability of a good event for Trump and a bad event for Clinton coinciding is .182^2, which is 3.3%. This is the chance that Trump wins, given your assumptions.
(Note that this assumes that Clinton can have only bad events, Trump can have only good, and that Trump loses unless a good event for him and a bad one for Clinton coincide.)
Poll
