電腦麻將程式ThousandWind的設計與實作

Abstract

近年在科技不斷進步之下，人工智慧電腦對局程式也不斷有新的方法或成果出現，技術方面也越來越成熟，但相對於明確資訊的對局遊戲，不明確且帶有機率性的對局遊戲程式一直以來都不容易跟人類玩家抗衡，相關算法以及論文討論也相對上比較少量，因此在這篇論文之中將會討論到關於電腦麻將程式的人工智慧開發。 這本篇論文中，將會說明電腦麻將程式ThousandWind裡面所使用到的各類算法，包含如何對牌型做評分、利用遊戲過程的統計結果來做動態的權重調整、避免放槍的新策略、以及過去論文沒有提到的關於追求牌分的方法，像是一些牌型比對的策略，以及藉由模擬結果來計算是否該追求更大的牌分。 目前該程式也曾獲得TAAI 2012電腦對局比賽的銀牌，以及TCGA 2013與ICGA 2013電腦對局比賽的銀牌。也期望論文中所提到的各種方法可以對往後不僅是電腦麻將程式的開發，甚至是可以帶給其他不明確資訊且帶機率性遊戲一些啟發。

Because of the advances in science and technology, computer games researchers continue to have new methods and achievements in recent years. Technology has also become increasingly mature. But relative to the perfect information games, programs that play imperfect information games have never been easier to compete with human players. There is less paper dealing with the related algorithms. In this thesis, we will discuss the development of computer AI program for playing mahjong. This thesis will explain all the algorithms which have been used in our mahjong program “ThousandWind”. These algorithms include how to evaluate the scores about the hand patterns, how to dynamically adjust the cards' weights during the game process by using the statistical results, how to find new strategies to avoid letting others win, and how to find new ways (e.g. using the hand patterns matching strategy and simulation results) to win more scores. Our program “ThousandWind” has won the silver medal of TAAI 2012 computer game competitions, and the silver medals of TCGA 2013 and ICGA 2013 computer game competitions. We expect that the methods presented in this thesis cannot only be used for the development of computer mahjong programs, but also be used for the imperfect information games with probability in the future.