[thirtyseven]

Not a game dev but I don't think that's really feasible in a world where network latency is a thing, and it's necessary to send game state to a client that may not currently be visible on screen, unless you'd like us to move to a Stadia-like model of game streaming and dumb terminals.

[devwastaken]

You don't need game state. Good cheaters don't need to look behind walls, and it's obvious when they do so. You can theoretically build your own bot simply from visual recognition that plays pretty well.

[anonymousab]

And if a cheater makes it non obvious by being a bit worse sometimes, and has become indistinguishable from a regular or just really good player, then... Mission Accomplished? At that point, it's no worse for the other players than playing with someone who is organically good, unless the sanctity of some matchmaking algorithm is more important to them than the gameplay.

Which is fine too, but the solutions for that problem may ultimately need to extend our be situated beyond the game software itself.

[smegger001]

I wonder if you could train an ai using screen grab and the mouse/keyboard input of players as a training set. so that it would look like a natural player rather than a aimbot

[inykt]

Well, Open AI's Dota 2 bot beats top players. It does not use any extra info thats not available to any dota player + lower reaction time than humans.

https://www.techspot.com/news/70552-openai-bot-beats-world-b...