科技演進使得原本許多的人力工作改讓機械取代，而人工智慧已經成為現代科技演進中重要的一環。人工智慧能讓機械式的工作自動化更為簡便且快速，甚至可以消除管理階層的負擔與成本。現今的人工智慧中，AlphaGo是最常使用且最容易套用的棋類模型，程式的編譯只需要基礎的遊戲規則，就可讓其產生神經網路來進行評估與落子，再讓其透過自身對下的方式不斷優化神經網路、強化自身棋力。2016年AlphaGo與李世石的圍棋競賽中，最後AlphaGo以4:1戰勝李世石(AlphaGo李世乭五番棋)，證明AlphaGo的強度保證。
一般的樹狀搜尋法會透過快速且大量的展開支節點，從中找出最佳的結果，但此方式會需要龐大的硬體資源與運算時間去做支撐。AlphaGo所使用的搜尋法為MCTS，此搜尋法方式優點在於避免搜尋樹的完全展開，透過選擇評估值較高的節點向下搜尋，雖然在單一節點需要額外資源，但能相對減少過多低回報的節點生成，使得在有限的資源下產生更有效率的分支。七國象棋為一種走步選擇相當龐大的棋類遊戲，本論文將利用MCTS於七國象棋之上，透過七國象棋的複雜度， MCTS能否在對局中展現一定的棋力。

The evolution of technology has made many jobs use machines, which makes artificial intelligence an important part of the evolution of modern technology. AI can make automating mechanical tasks easier and faster, and can even eliminate the burden and cost of management. In modern artificial intelligence, AlphaGo is the most commonly used and easiest to apply chess model. The compilation of the program only needs the basic game rules, and it can generate a neural network to evaluate and make moves. The next way is to continuously optimize the neural network and strengthen its own chess skills. In the Go competition between AlphaGo and 李世石in 2016, AlphaGo finally defeated 李世石 by 4:1(AlphaGo李世乭五番棋), proving the strength of AlphaGo.
The general tree search method will find the best result by expanding branch nodes quickly and in large numbers, but this method will require huge hardware resources and computing time to support. The search method used by AlphaGo is MCTS. The advantage of this search method is that it avoids the complete expansion of the search tree. By selecting a node with a higher evaluation value to search downwards, although additional resources are required at a single node, it can relatively reduce the generation of too many low returns. node. MCTS produces more efficient branches with limited resources. Seven States Chess is a chess game with a large selection of moves. In this paper, MCTS will be used on Seven States Chess. Through the complexity of Seven States Chess, whether MCTS can show a certain chess power in the game.

摘要 ii
ABSTRACT iii
致謝 iv
目錄 v
圖目錄 vii
表目錄 viii
第一章、緒論 1
1.1 七國象棋 1
1.2 研究動機 2
1.3 論文概述 2
第二章、文獻探討 3
2.1 Bitboard 3
2.2 UCT 演算法 4
2.3 MCTS演算法 5
第三章、研究方法 7
3.1 棋盤基礎編譯 7
3.2 盤面走步建構 9
3.3 走步生成範例 13
3.4 棋盤盤面評分 16
3.5 MCTS的應用 17
第四章、實驗結果 19
4.1 環境配置 19
4.2 棋盤設定 19
4.3 實驗目標 20
4.4 實驗數據與分析 21
4.4.1 全隨機走步 21
4.4.2 全MCTS預測 22
4.4.3 MCTS作為最後手對隨機走步 22
4.4.4 MCTS執行時間變化 23
第五章、結論與未來展望 24
參考資料 25

七國象棋【維基百科】。取自https://zh.wikipedia.org/zh-tw/%E4%B8%83%E5%9B%BD%E8%B1%A1%E6%A3%8B
蒙地卡羅樹搜尋【維基百科】。取自https://zh.wikipedia.org/zh-tw/%E8%92%99%E7%89%B9%E5%8D%A1%E6%B4%9B%E6%A0%91%E6%90%9C%E7%B4%A2
AlphaGo李世乭五番棋【維基百科】。取自https://zh.wikipedia.org/wiki/AlphaGo%E6%9D%8E%E4%B8%96%E4%B9%AD%E4%BA%94%E7%95%AA%E6%A3%8B
Chang-Shing Lee, Mei-Hui Wang, Guillaume Chaslot, Jean-Baptiste Hoock, Arpad Rimmel, Olivier Teytaud, Shang-Rong Tsai, Shun-Chin Hsu, Tzung-Pei Hong (2009). The Computational Intelligence of MoGo Revealed in Taiwan’s Computer Go Tournaments. IEEE Transactions on Computational Intelligence and AI in Games. doi:10.1109
Introducing Bitboard-with a little Othello implementation-. Retrieved from https://linuxtut.com/en/09f114799319c7ba19dc/
Levente Kocsis & Csaba Szepesvári (2006). Bandit based Monte-Carlo Planning. Machine Learning: ECML 2006. doi:10.1007/11871842_29
Pablo San Segundo, Ramon Galan, Fernando Matia, Diego Rodriguez-Losada & Agustin Jimenez (2006). Efficient Search Using Bitboard Models.2006 18th IEEE International Conference on Tools with Artificial Intelligence (ICTAI'06). Retrieved from https://ieeexplore.ieee.org/abstract/document/4031890
Rémi Coulom (2006). Efficient Selectivity and Backup Operators in Monte-Carlo Tree Search. Computers and Games, 5th International Conference, CG 2006. doi:10.1007/978-3-540-75538-8_7