I saw this article in the NY Times and was inspired to write my own Rock Paper Scissors implementation.
The AI uses something very similar to a Markov Chain to determine it’s next move, with the exception that the transition matrix is updated after each decision by the player. Currently, I just use a simple weighting scheme to change the conditional probabilities between rounds, although hopefully in the near future I’ll be able to find the time to develop an updating scheme based on Bayes Theorem.
Quick after thought: I beat the NYTimes/Facebook ai 8-2-2 with the strategy of determining my next selection by rolling a die. 