Human error does determine match up ratios... specifically how likely it is to be made and the effects of it happening and so on. This includes spacing errors by the way. If human error doesn't determine match up ratios then every match up should be 100-0 or 0-100, except for dittos, since the only variation in the winner of a match comes from either human error or random elements, and the random elements in Brawl are too insignificant to account for matches not being 100-0. (We are assuming that both players are at the same overall "skill level", or propensity for making errors, for match up ratios in general, so that isn't a factor.)
A more formal approach might be something like this. What do we really mean when we say both players are at the same skill level? What we mean is that their skill levels, namely A and B, can be represented as two probability distributions with a characteristic function of the same family describing a distribution Q that depends on your character and the opponent, let's say A ~ Q(Ness, Game & Watch), B ~ Q(Game & Watch, Ness), and the match up ratio is P(B > A). In other words, the random variable A gives you a particular value for players A's skill in that match up, and the random variable B gives you a particular value for player B's skill in that match up, The match up ratio is the chance of a random value of random variable A being greater than random variable B.
It's plainly apparent from this formalisation that "human error" is the factor determining match up ratios. Interestingly, this formalisation also allows us to explain some related ideas. For example, in the later rounds of a tournament we might expect the variance in Q to be lower, since you are "warmed up", giving rise to more extreme match up ratios (closer to 100-0).