The problem as I see it is the extrapolation based on a limited amount of data. We're trying to make predictions based on a cycle that has completed only three times.
A simple alternative, for instance, could be a repeating A-B-C pattern. 3 part patterns are commonly seen across various markets.
Or the pattern could be more complex. Here's another alternative analysis of the (well, a similar set of) data:
That's actually the entire encapsulation of the TA problem with Bitcoin. Extremely constrained data history. How long's the open market been in existence, 7 years at maximum? With today's kind of volume, 2 years maximum? 3 cycles is all it's ever had, 3 bubbles, and the first two were easily manipulable so should be discarded.
The case is even worse with altcoins. Patterns repeat, until they don't. And they only show up if you're looking for them.