But I pretty much include "either side might win" in my understanding of a good game.
The supercompetent character is a common dramatic conceit. Mu current fave example is Abbey from NCIS – the lab tech who can competently use any piece of technology as an expert. The reason we have this conceit is that it is boring and not relevant to the story moving forward that her performance would be covered by a team of different people. We don't want to invest (learn about, track, spend time on idiosyncrasies) in a bunch of different characters, so writers give us one Uber character.
Same thing goes for dying – We don't want to be introduced to a slew of new people all the time. So they survive. This also leads to branding … "The Lone Ranger" is identifiable vice the list of twenty vigilantes who die every other month.
But, as pointed out above (and for JCM, Conan, Sackett, etc.) Having a character not die doesn't equal them winning every scenario. Many stories are built on "bad guys win; we have to set things right". Even more have the looming "one (two, three) that got away" story arch. And good stories have outcomes where a win is just the least heinous long-term consequences.
And winning doesn't have to be all or nothing. Did you catch the bank robbers? Did some of them get away? Did the mastermind? Did you recover all the money? Part of it? Did you discover the deputy was in on it, but you can't prove it (because you had to kill (or let the gang kill) the only witness)?
As long as there are consequences, and best continuing consequences game to game, either side can win. This is how the archetype ubercharacter maintains interest in their exploits and also becomes a viable side for a wargame. In this sense Batman is the Lone Ranger is the Scarlet Pimpernel.