When one is not able to understand what another says, assuming the other person is necessarily incorrect/silly may not be optimal gameplay (depending upon what variable(s)[1] one is optimizing for of course, or if one never even thought about that aspect of it in the first place).

Less snarkily: the 1/2/3 watches example is a single, ~deterministic, ~knowable data point; it is roughly the simplest type of problem space you can deal with, and even here there is uncertainty. Now consider a much more complex problem space, reality (better known as "the reality"), and things get much worse. But paradoxically, with all the added complexity, "correct" answers often become simpler, not more difficult. (Here I am referring to the conversations you may find on social media (or anywhere) about complex matters such as war.)

[1] Having fun, fitting in, just joking, etc.