A MiniMax Agent for Playing Ntxuva Game – The Mozambican Variant of Mancala

Since the earliest development of Artificial Intelligence (AI), the AI community has been attracted by games. Because, the task of deciding which move is best to take in a given state of a game, requires the involvement of some strategies which can be seen as a form of intelligence. In games such as Chess, researchers have been able to develop an AI opponent which can defeat a human opponent by using the Tree Search algorithm. The Tree Search algorithm has been also applied in some Mancala's family games, but there is around 200 variant of the game. Even though they are governed by almost the same rules and principle, they differ in terms of configurations and complexity. To our knowledge, Ntxuva is one of the variants of Mancala that has not been explored by the AI community. This paper reports the finding of using AI methods to create intelligent agents to play Ntxuva. Different agents were created namely, a Minimax agent, Alpha-beta pruning agent, as well as a Random agent. Then, they were put into play in a round-robin tournament where the statistics of their winning rate were collected. By applying Minimax Algorithm to the computer agent, it becomes intelligent enough to play Ntxuva as a good opponent like a professional human player.