Hmm I'm not convinced. See Buena uses it in a more elaborate fashion, obviously not aware of all the misunderstandings he is provoking by using the term 'OP' which really everybody seems to have its one defenition(s) of.
He surely doesn't intend to say "whatever the last team who beat me plays". Hmm but I recall a broadly shared opinion from this thread that 99% of all viable 5v5 setups include one paladin, so what you say might as well be right since the chance for the last team who succeeded over Buena's 5s running one paladin is very high. Things just got more complex..
As a more serious answer, the issue at hand is that research has shown that it is 12 times as hard to change perception than it is to confirm it. That is, for each instance where your preformed opinion is confirmed you need 12 instances where it is proven wrong. So if you belive that class X is overpowered you need 12 instances of the class underperforming to counter one instance where the class outperformed.
That is why I have been harping on numbers. They are more reliable than your perceptions. Example (this is an example to make a point not a statement of game balance).
12 games of warrior/palls vs warrior/druid. Warrior/druid tries the "CC of doom" all 12 times. 11 times it fails and the druid goes splat. One time it works. Is that overpowered? To the uninterested observer the answer is clearly no. But to someone who believes that the CC of doom is overpowered, this is actually confirmation that it is, because in our minds, eleven counterexamples are not enough to outweigh one example that supports our beliefs.
Because humans think that way you cannot use perception and opinions to judge balance. The only opinions you can use are those of neutral observers. Problem is, a neutral observer is unlikely to know enough about arenas to balance them. So the only thing you are left with are numbers.