Diseño e implementación de una estrategia para un juego de conexión

Abstract

The purpose of this project is developing an automatic strategy for playing Selfo, a connection game whose goal is to get arranged all the friendly pieces into a single connected group. The idea is to copy some kind of swarm conduct, like the observed in an ant colony or in a flock of birds, to play this game. The swarm behaviors consist on few simple inborn rules. But these simple rules applied in group usually bring complex dynamics. The ultimate goal is to check whether a computer player based on these principles can compete against a thinking trained human. We first built a simple computer version of Selfo in order to deeply examine the dynamics of the game by running human-human and human-dummycomputer games. From this experience we proposed and developed a swarm behavior based strategy. We built the program over the Zillions of Games platform. Zillions is a popular website where people can download and play a multitude of board games. This is a great tool to get the players experiences. And it also allowed us to easily define the user interface and game rules, letting us focusing on the game engine. To evaluate the quality of the proposed strategy, we run several simulations of computer-computer games and played human-computer games. We noted that the computer opponent can defeat many times an average player and sometimes beat an experienced player. We also learned that it is very important the initial distribution of pieces and sometimes the strength of the first move. Finally we could observe the speed of the computer player, despite the computational complexity of the game. We conclude that we have achieved the initial aims: building an easy to use computer version of Selfo with an on-line game mode and developing a fast strategy inspired by nature which could face up a human player.