It’s not even a well-defined question to ask "what is the best action".
To ask that in the context of a fight (not a bounded game like chess) is already to assume the existence of a complete utility function on which to measure it. That’s:
1. Philosophically, putting the cart before the horse.
2. Computationally, asking for the function that is the entire universe. Any utility function you define, an adversary (say, God) can find edge cases it doesn’t account for, endlessly. Chess has a finite state space; a fight doesn’t. Formalizing this hits the usual incompleteness and undecidability limits.
You’re claiming a perfect map exists (Platonist position); I’m saying that if such a thing exists, it’s just the territory itself, which isn’t a map (Nominalist position).
I'm not saying you can get a perfect map, I'm saying the territory exists.
I don't think we really disagree.
Other than that if you compare 2 possible future timelines, you can either pick a favourite in which case that one has more utility, or you can't in which case they have equal utility.
Yes your opponent can make a move and you don't know what move they'll make. Chess is like that too.
I'm not saying this is in fact a good way to win a physical fight.
I'm also not saying the optimal move is unique. If 2 moves have the same utility then they can both be optimal.
What I'm saying is that just because you don't know what is the best move doesn't mean a best move exists.