In the thesis we learn about the field of artificial intelligence that investigates board games and their program-based solutions. We examine Monte-Carlo tree search algorithm and transfer it to well-known board game Scotland Yard, considering advices from Nijssen and Winands. We focus mainly on the third phase of the algorithm, playout, and decide to implement it in three different ways (from less to more advanced techniques). We compare these three aproaches. We compare the win rates and computation time of simple and advanced methods. We also implement the game to the purpose of automated testing. In this game, detectives play by Monte-Carlo tree search algorithm and Mister X plays in two different ways - random and advanced. We want to test all of these six combinations on a large number of games, compare the results and explain them.
|
