Just a thought. Then you have constant arrays of super human things computers are able to do, and even Deep Blue (A chess playing computer) beat Garry Kasparov (The World Champion at the time) in their rematch.
My question is this: Would it be possible to create a computer that's so good at smash bros, that 20xx basically becomes a possibility, and that not even any of the strongest Gods can beat it? Sure, it's radically different to compare a chess playing computer to a smash playing computer, but would that even be possible?
It already exists in a somewhat functional form despite being very unfinished. I'm quite sure that even in its current state the only way to beat this AI would be to abuse bugs which prevent it from functioning properly; fundamentally its strategy of "powershield/shine clank/move away from everything on reaction, and take guaranteed openings to reaction techchase perfectly until a guaranteed kill setup is available" is sound:
SmashBot vs Level 9 CPU
Melee becomes degenerate once you reach a certain execution/reaction threshold, as do many games (maybe even all; I can't say I know enough about it to comment accurately); it's not designed for the players to have 1 frame reactions and perfect techskill. Even games without physical execution barriers still have mental execution barriers which AI are not necessarily restricted by and which can therefore theoretically be played by AI better than any human.
As you can see demonstrated in the video, the AI doesn't require any "mindgames" to speak of, as their strategy leaves no room to be outplayed.
Very unlikely due to the way computers function. A computer cannot think or play dynamically like a human can, it can only think "I must do X because Y" or some variation thereof. It's very easy to catch onto patterns like this and punish them regardless of how frame perfect said computer may be.
To summarise/reiterate what I said above: if the flowchart doesn't provide any room for counterplay then it doesn't matter how predictable it is. You can predict the particular way in which the AI will unpunishably react to whatever you do while never taking risks or otherwise presenting openings, but that information is useless.
In all solvable games, the intuitive/heuristic strategies that humans and imperfect AIs use are a way to compensate for their limitations. Those strategies do not hold any inherent advantage over totally objectively determined "flowchart" strategies. The only times the former are ever better are when the latter are not implemented well enough (due to technical limitations). All solvable games are fundamentally the same in the sense that there is an optimal strategy, and it doesn't matter if you understand the reasoning behind it or not since you can implement it just as effectively regardless. You're not going to claim that you can beat a computer at tic-tac-toe because you possess a sapient understanding of it which the computer doesn't; Melee is the same, just with more possible moves.
