You don't need that many matches to iron out statistical deviations. There are plenty of tournament structures that have been used to deal with that exact issue, such as the Swiss system. Besides, I am still not talking about competitive tournaments with my arguments. I have little expectation that the competitive Smash scene is going to change based off my opinions.
I just finished addressing that exact scenario. That scenario is a) statistically unbiased (it doesn't favor either player, and even if one player benefits that benefit will even out over multiple matches) and b) rather uncommon. That exact event isn't skill based, but it is the only scenario I have seen someone put forward that doesn't involve player skill. The existence of that possible event doesn't preclude player skill from determining the winner of the contest, either.