Mancuna››                             Atari User Vol.1 No.12››This game-playing program is basically an exercise in artificial intelligence.›Unlike most game-playing programs however, this one is different in that the›computer actually learns as it goes along.››The program partly simulates human learning as the computer remembers positions›which led it to lose, and then, given the opportunity, plays these back against›its opponent.››The game is a simplified version of an Arabic game called Mancuna. I chose this›game for the exercise because it is fairly challenging while not being too›complex and it may also be new to many people.››This simple version is played by two players using eight counters on a 2 x 2›square board. The game begins with two counters on each square, and the players›face each other across the board.››Taking turns, each player picks up the counters from one square on his/her side›of the board and moves them anticlockwise round the board putting one counter on›to each square in turn until they are exhausted.››Thus if one player picks up three counters from the bottom left square he must›put one counter on the bottom right, one on the top right, and the last one on›the top left, leaving no counters on the original square. The winner is the›first player to have all eight counters on his/her side of the board.››The program allows you to play against the computer. On your turn move the›joystick left or right to choose the square you want to move from, and then›press the fire button.››At the start the computer is totally naive, knowing only the rules of the game.›Each time it loses, the computer remembers the position which led it into a›losing sequence of moves. In future it will avoid getting into this position›itself, but will put you into it if it can.››Humans also learn like this - but unlike a human the computer learns nothing›from winning, nor can it formulate general rules of strategy to guide its play.››Against this however, the computer does have an advantage that it never forgets›positions which it knows must lead to a loss, or those which must lead to a win.›It learns rapidly from its mistakes.››After five losses it is quite a fair player, and after 10 it can be difficult to›beat. Try playing it 'the first one to 10 wins', and see who learns quickest.›