everything is subject to some sort of calculation. If you don't use math the guys that critique your design will. In any game where there are successes and failure's math can be used to measure it.
I did not say that math was irrelevant.
I said that it could not be simply reduced to a brute calculation: pick the thing with the biggest number and go.
Good, well-balanced games offer multiple choices where the calculations
cannot account for the differences, and thus fail to provide a compelling reason to choose. Instead, a well-balanced game forces value judgments. E.g., "You could dual wield, which lets you make a lot of attacks but results in you dealing less damage with each attack. Or you could go two-handed weapon, which will hit hard, but more rarely, so it's high risk, high reward. They both end up doing essentially the same average damage, so it's more a matter of what you prefer in terms of feat support and items. Dual wielding is pretty versatile but expensive and not very focused. Big two-handers are the reverse, specialized and cheap, but inflexible."
That's a situation where mathematical analysis cannot guide you to one obvious singular correct choice. The player must make a value judgment about what they like better, what sounds more appropriate, what makes more sense in the context the player faces. Etc.
Believe me, I'm a huge advocate for doing the hard, tedious work of actual statistical analysis of game design. You do that hard math work
so that the resulting game offers choices that differ primarily in intangible or incommensurate ways, rather than in simplistic "X does 20% more damage, always choose it and ignore Y and Z."
Unbalanced games encourage such simplistic, brute calculations. Choose the thing which performs best and ignore the rest. There is a dominant strategy: exploit it. Etc.