Monte-Carlo Tree Search for the Game of "7Wonders"

Monte-Carlo Tree Search, and in particular with the Upper Confidence Bounds formula, has provided large improvements for AI in numerous games, particularly in Go, Hex, Havannah, Amazons and Breakthrough. In this work we study this algorithm on a more complex game, the game of "7Wonders". This card game gathers together several known challenging properties, such as hidden information, multi-player and stochasticity. It also includes an inter-player trading system that induces a combinatorial search to decide which decisions are legal. Moreover, it is difficult to hand-craft an efficient evaluation function since the card values are heavily dependent upon the stage of the game and upon the other player decisions. We show that, in spite of the fact that "7 Wonders" is apparently not so related to classic abstract games, many known results still hold.