I basically agree; but it also depends on expectations and information sampling by trading crowds.

There is a superabundance of information, not a scarcity. Just that most people do not bother to go through it all. Too costly. So much of it and most useless. Therefore exists a low probability of covering search costs. In that case the Bitcoin position as "First" and "Most Popular" (echoing Ries, Trout, and recently, McAfee) are sufficient for to individuals buy it — even when they hear news of other coins going up. If the high price does not motivate individuals to bother to learn what coin is most interesting, then another two patterns are possible for a while.

(1) If crowds are mostly uninformed: people buy, price rises, they sell to lock in gains, price falls somewhat, seeing this, more such people buy including those who sold, price goes even higher, they sell, for the same reason, prices falls percent wise even more but to a floor higher than it was, and so on. This repeats at multiple time scales. Not all individuals buy and sell at once but in random clumps and some make decisions more or less slowly depending on their risk preferences.

(2) If the uninformed crowds also have too high expectations: they buy, then wait and then sell even if the price goes up but not so quickly as they wanted, and then it goes down.

Such patterns can be superposed on your pattern depending on the detailed composition of trading crowds.

Positions are surprising resilient when information is too plentiful (Ries Trout, 1987, Positioning). In my estimation it could go for double or triple what the recent peak was. The way to know if that is happening is to look at whether volatility is following a power law. (Is scale invariant or not.) That would indicate whether the first process is occurring or not and what is the mode composition of the crowd.

This is why cycling and fractal like behavior occurs in waves. (Elliot waves in triangle occur when corrections arise from dampening by expections (process 2) following a rapid but increasing volatile rise (process 1). Crowds typically are underinformed by choice and therefore behave in cellular automaton fashion more often than not. Results in some order scale invariant behavior, feedback, and dampening over several time intervals superposed.