This is a zero-sum game with the same 16 strategies for each player. If player 1 does damage A1 (for attack), S1 times per (say) hour, and player 2 has armour that cancels D2 points from a successful attack and gives E2 (evade) percent change of dodging, the rate of damage by player 1 on player 2 is S1*(1-E2/100)*(A1-D2). Subtract from this the same expression with 1 and 2 swapped and that is the differential rate. Given the conditions of the problem, the only thing that matters about this rate is its sign. So there's a 16*16 payoff matrix with entries 1, 0, and -1.
I've computed it, and it turns out that no strategy dominates any other: every row contains at least one -1 in a column containing at least one 1. However, I forget how to compute a minimax mixed strategy for such games. I think it's a linear programming problem, and for 16 variables it can't be done manually unless the data have some regularity to be exploited. Eyeballing the matrix, I don't see any.
Note: this image does not belong to me; I found it on 4chan. It presents an interesting exercise, though, so I'm posting it here for the enjoyment of the Less Wrong community.
For the sake of this thought experiment, assume that all characters have the same amount of HP, which is sufficiently large that random effects can be treated as being equal to their expected values. There are no NPC monsters, critical hits, or other mechanics; gameplay consists of two PCs getting into a duel, and fighting until one or the other loses. The winner is fully healed afterwards.
Which sword and armor combination do you choose, and why?