Monte Carlo Methods for the Game Kingdomino Magnus Gedda, Mikael Z. Lagerkvist, and Martin Butler Tomologic AB Stockholm, Sweden Email: fi
[email protected] Abstract—Kingdomino is introduced as an interesting game for enhancements have been presented for MCTS, both general studying game playing: the game is multiplayer (4 independent and domain-dependent, increasing its performance even further players per game); it has a limited game depth (13 moves per for various games [14], [15], [16], [17], [18]. For shallow game player); and it has limited but not insignificant interaction among players. trees it is still unclear which Monte Carlo method performs Several strategies based on locally greedy players, Monte Carlo best since available recommendations only concern games Evaluation (MCE), and Monte Carlo Tree Search (MCTS) are with deep trees. presented with variants. We examine a variation of UCT called Kingdomino [19] is a new board game which won the progressive win bias and a playout policy (Player-greedy) focused prestigious Spiel des Jahres award 2017. Like many other on selecting good moves for the player. A thorough evaluation is done showing how the strategies perform and how to choose eurogames it has a high branching factor but differs from parameters given specific time constraints. The evaluation shows the general eurogame with its shallow game tree (only 13 that surprisingly MCE is stronger than MCTS for a game like rounds). It has frequent elements of nondeterminism and Kingdomino. differs from zero sum games in that the choices a player makes All experiments use a cloud-native design, with a game server generally have limited effect on its opponents.