Optimization of an Evaluation Function of the Four-Sided Dominos Game Using a Genetic Algorithm

摘要

In four-sided Dominos, the popular way of playing Dominos in Amazonas State, in Brazil, the strategies used for the game are more complex than those adopted in the more traditional two-sided Dominos, the most popular domino game played in Brazil. This work presents the optimization of an evaluation function for the best move in four-sided Dominos using a genetic algorithm (GA). The evaluation function comprises terms incorporating game strategies defined as: punctuating, facilitating future moves, and complicating opponents' moves. Coefficients were defined to determine the importance of each term of the evaluation function and a set of parameters and operators for implementing the GA. The players' ability was calculated by the number of wins in 5000 matches. The results obtained during the simulations showed that the team (consisting of two players) using the evaluation function with its coefficients optimized by the GA won in more than 69.18% of the total matches.