A common argument that Bitcoin will remain the dominant crypto is given here by Adam Back.
if a constant stream of altcoins over took each other, then there is no digital scarcity, so no store of value, end of experiment […] I view it as more definitional, B overtook A, then C can overtake B, and D in turn will be in everyones minds who was burned by hoping for a store of value. That becomes a timing game, and far weaker value proposition
The claim is that B overtaking A (in this case A == Bitcoin) is analogous to C overtaking B. If B does overtake A, then we should expect it to be much more likely that some other coin will eventually overtake B.
If we don’t have any other context then knowing that B overtook A should cause us to update our probability estimates significantly in the direction that B will be overtaken.
However, we do have other context. In the real world things are complicated – there are a lot of factors that can play into whether a coin will be overtaken by another.
For instance, if we know that A had a particular glaring flaw that B does not have then we should update in the direction of B being more secure. If our main reason for previously thinking that A would be stable was due to a theory of network effects, but we then learn more about network effects and discover that A’s network effect strength was 100x weaker than B’s network effect strength becomes later, then B overtaking A should no longer fill us with so much concern. Or if we already have information about specific competitors to A and know that they have gotten somewhat close to overtaking it in the past, this should increase our estimates of A being overtaken.
The root of the problem
Bitcoin maximalist reasoning almost never looks like Bayesian reasoning. Bayesian reasoning involves taking every new piece of information as evidence which should almost always shift our estimates at least slightly.
Maximalists generally prefer deductive reasoning to probabilistic reasoning. They often have a series of deductive arguments based on very simple models of the world.
For instance in the above argument the only information we’re considering about a cryptocurrency’s staying power is a couple bits of information about whether the top coin has ever been overtaken, and whether it is the current top coin. The model that generates the famous “money must always evolve in 4 stages, starting with SoV” belief which Vijay Boyapati describes in The Bullish Case for Bitcoin is also incredibly simple.
The problem with relying heavily on these simple models is that it presumes that we know exactly which factors we can treat as irrelevant (and thus as OK to completely exclude from our model). In the real world, almost nothing is completely irrelevant (see Bayesianism) and even picking out the most relevant few factors in a given situation can be incredibly difficult.