Blatant Plagiarism

Subject:

Comments:: In a move worthy of Rex Langman, (who is still incarcerated for his involvement in the Circle of Five/Keepers of the Naktrah/Disappearance of Rosh fiasco) I am here reprinting virtually wholecloth some information I found in a link that the late Roshambollah left on his web browser. The info is from the First Annual Roshambo Programming Competition, and much of the strategic info applies to both human and computer players. Fascinating stuff, especially considering recent discussions re: optimal strategy. : Bill Helfer : Curator, Master Roshambollah Memorial Library : Myth: Rock-Paper-Scissors is a trivial game. : Sure, the game has a simple optimal strategy (choose a move uniformly at random), but that has little bearing on the problem at hand. First, not all the players are optimal. This changes everything. To win a tournament where some players are known to be sub-optimal, it is absolutely essential to try to detect patterns and tendencies in the play of the opponent, and then employ an appropriate counter-strategy. A match consists of several turns, and this changes the nature of the game, as was seen in the famous Iterated Prisoner's Dilemma problem. : RoShamBo (and it's even simpler cousin, the Penny-Matching game) is an example of a pure prediction game. The difficulty lies in everything else that is associated with opponent modeling, or trying to outwit an adversary. : There is a lot of theory that can be brought to bear on the problem, including but not limited to advanced game theory (the "best-response dynamic in fictitious play"), prediction models, information theory, statistics, encryption, and even philosophical meta-theory. : Myth: Random (Optimal) can't be beat. : The optimal strategy won't lose a match by a statistically significant margin, but it also won't win a match, regardless of how predictable the opponent is. Try winning a chess tournament by drawing every game! : Moreover, the statement isn't even true in a more fundamental sense. Opportunistic strategies can be theoretically better, having positive expectation under more realistic assumptions. People interested in advanced game theory may enjoy the recent book "The Theory of Learning in Games" by Fudenberg and Levine. : Myth: Since all non-optimal strategies can, in theory, be exploited, the result of a tournament will be a crapshoot. At the very least, the outcome will be highly sensitive to the exact composition of players (algorithms) in the tournament. : The premise is true, but the conclusion is false. Any non-optimal algorithm can be beaten, just by employing the same algorithm and adding one to the action (r -> p -> s -> r). But complex algorithms are not vulnerable to such an attack. In general, they can only be beaten by an opponent who does a superior job of analysis. : There are many levels of complexity for playing algorithms, which can differ in the way they use history (context), in their perceptiveness of the opponent strategy, and in their defensive ability (hiding their own strategy). By in large, the more information a program processes, the better it will play the game. : (Note: I concur. Regardless of whether a player is using the "Great Eight" gambits, reverse profiling, inclusive strategies, kinetic and muscular clues, etc, one fact jumps out: the more levels of strategy one is capable of using at once is directly related to one's skill as a player. -Bill)

Optional Link URL:
Link Title: