Having not read the article and going solely on the information presented in your post:
I feel like this may come down to a bit of a schism between what’s good and exciting for spectators vs what’s good and exciting for players. One would hope that players of a sport, especially a game like Yomi that’s played for fun/bragging rights rather than material gain, would be happy that the tournament scene has a healthy population of excellent players. After all, that should theoretically improve the overall quality of tournament games for players (if we assume that people enjoy playing both at a high skill level and against evenly-skilled opponents).
For spectators, though, unless you have clear top dogs and under dogs, it’s very difficult to create interesting, or even coherent, narratives about a tournament season. People want to be able to trace a neat story, and you can only get a neat story to sell if the players are lop-sided enough in skill to produce that.
There’s also the problem of perceptual biases. For example, even if Player A had an 7-3 advantage over Player B, it’s not that unlikely that Player A might lose twice in a row to B. But casual spectators might feel like that can’t possibly be correct because Player A is the “better” player and therefore ought to win. Even losing 3 times in a row has a healthy 2.7% chance of happening.
Lastly, I think it’s important to make a distinction between two different kinds of random. True randomness would be complete luck (e.g.: we flip a coin, and whoever guesses right, wins). True randomness means that the outcome is impossible to know ahead of time /and/ we have no ability to modify the outcome. However, Yomi is not /truly/ random because players are making decisions that affect the outcome. Thus, if every match is 5-5, the outcome is “random” only in the sense that we cannot definitively predict which outcome will occur. The outcome is mostly determined by the players’ choices. To call that “random” removes the human component of playing games and regards the players as mere machines executing the expected probability range.