Some AI techniques for RPS game
Returns a specific move depending on a given probability distribution. Seem to lose to any other type of agent that counts probabilities and can detect the bias
Returns the move depending on the probability distribution of sequences of moves. Sequence length is a parameter for the agent. Obviously, is susceptible to sequences of bluffs that will shift the probabilities, beats the biased agent
Returns the move based on a simple HNN with a few layers and inputs given by the last X observed moves. Beats Markov Agent and Biased Agent
Find the longest sequence in all the moves made by the adversary that matches the last X moves. Returns the opposite of the next item in that sequence If the sequence is not found, it uses a Markov Agent as a proxy for the answer. Beasts all of the above