770 likes | 951 Views
C onnect6 / 六子棋 A N ew C hallenging G ame and its Latest Developments 一種新的挑戰性遊戲及其最新發展. I-Chen Wu / 吳毅成 Department of Computer Science, National Chiao Tung University, Hsinchu, Taiwan 台灣新竹國立交通大學資訊工程系 October 28, 2008. 人腦戰電腦 紅面棋王敗北 ! (08/24/2007).
E N D
Connect6 / 六子棋A New Challenging Game and its Latest Developments一種新的挑戰性遊戲及其最新發展 I-Chen Wu / 吳毅成 Department of Computer Science, National Chiao Tung University, Hsinchu, Taiwan 台灣新竹國立交通大學資訊工程系 October 28, 2008
人腦戰電腦 紅面棋王敗北! (08/24/2007) 民視新聞(http://www.connect6.org/2007824L09M1.wmv )
個人資料 • 現職: • 國立交通大學資訊工程系教授 • 鈊象交大聯合研發中心主任 • 台灣六子棋協會理事長 • 學歷: • 台灣大學電機工程學系學士(1982) • 台灣大學資訊工程研究所碩士(1984) • Ph.D., Computer Science of Carnegie Mellon University (1993)
研究兩大主軸 • 網路遊戲 • 發展網路遊戲平台: • 研發高速的Java AWT系統[Wang and Wu 2007] • 研發P2P傳輸下載系統。 • 鈊象交大聯合研發中心:並有其他技術移轉! • 棋牌類人工智慧 • 發表新的棋種:六子棋 • 發展六子棋程式:交大六號(NCTU6) • 兩度獲得奧林匹亞競賽金牌 • 發展其他人工智慧程式:如電腦象棋、數獨驗證系統等。
Outline • Introduction to Connect6 • Characteristics of Connect6. • Fairness of Connect6 (六子棋) • Compared with Gomoku (五子棋) or Renju (連珠棋). • Threat-based winning strategies for Connect6 players or programs. • Impact of Our Research • Report of some Tournaments • New challenges in the future.
三大特點 • 規則簡單 • 變化複雜 • 遊戲公平
規則簡單 • 黑先下1子 • 之後各下2子 • 連成6子者勝 一般化K子棋: • Connect(k,p,q) • 六子棋為:Connect(6,2,1)
Generalized Connection Games • K-in-a-row (K子棋): Connect(m,n,k,p,q) • Rules in Chinese: • 有黑白兩玩家,黑先下q 子,之後輪流下 p 子。 • 誰最先在棋盤橫向、豎向、斜向形成連續的同色 k子者為勝。 • 棋盤大小是mxn。 • When m= & n=, also called Connect(k,p,q) • Example: • 井字遊戲:Connect(3,3,3,1,1) • 五子棋: Connect(15,15,5,1,1) • 六子棋: Connect(19,19,6,2,1) 或 Connect(59,59,6,2,1)
變化複雜 (Game Complexity) • 六子棋的複雜度非常高。可從目前常用的兩個複雜度標準來看。 • State-Space Complexity: (所有盤面狀態數) • 六子棋複雜度:10172 • 與圍棋複雜度相同 • Game-Tree Complexity: (遊戲樹展開的大小) • 圍棋複雜度: 10360 • 日本將棋複雜度: 10226 • 象棋複雜度: 10150 • 六子棋複雜度:( 10140~ 10188)象棋與日本將棋之間 • 最早估計一般走30步,則 (300*300/2)30 ~= 10140 • 但目前高段棋士,常走到 40步以上。則約為 ~= 10188
西洋棋一般公認複雜度第四 象棋一般公認複雜度第三 資料來源: Herik的論文[2002] 圍棋一般公認複雜度最高 比五子棋複雜度低 百年前日本即有職業連珠棋院 日本將棋一般公認複雜度第二高 Connect6 10172 10140 ~ 10188
遊戲公平 • Key: Balanced (平衡) • One player always has one more stone than the other after making each move. • 每當一方下出一步(兩子),該方一定比對方多出一顆子。 • B: #1, W: #0 • B: #1, W: #2 • B: #3, W: #2 • B: #3, W: #4 • B: #5, W: #4 • …
Outline • Introduction to Connect6 • Characteristics of Connect6. • Fairness of Connect6 (六子棋) • Compared with Gomoku (五子棋) or Renju (連珠棋). • Threat-based winning strategies for Connect6 players or programs. • Impact of Our Research • Report of some Tournaments • New challenges in the future.
Fairness of Gomoku • Rules in the free style (no restriction). • After B move, B has one more stone than W. • 每當黑方下出一步後,比白方盤面多一顆子 • After W move, W only has the same number of stones as B. • 然而每當白方下出一步後,盤面子數卻只能與黑方打平。 • Allis [1994] proved that B wins. • Allis [1994]等人證明出先下必勝 。
Fairness of Renju • Japanese Professional Renju Association (1903) imposed some new rules to restrict the play of B for professionals. • No double 3, double 4, and no overline. • 限制雙活三、雙死四、長連 • Called Renju (連珠棋) • Problem still. • Professionals still said that B wins. • Wagner、Virag [2001] proved this.
Side Effect One • More and more complex rules. • RIF: Renju International Federation. (國際連珠棋協會 ) • New Rules by RIF [1998]. • New restriction in opening. (限制許多開局規則 ) • Call for new rules in 2003. • Vote new rules in 2006. • But, vetoed by Japanese RIF.
Side Effect Two • Smaller board [Sakata and Ikawa, 1981] (棋盤變小 ) • Sakata 及 Ikawa said: • Increasing the board size raised B advantage. Restrict the board size to 15x15 • But, a smaller board lowers the complexity of the game and thus makes it easier to solve the game.
More about Fairness • Definitions of fairness • Strategy-Stealing Argument • Breakaway (脫離戰場 ) and fairness • Fairness of Connect6 • Fairness of Connect games
Definitions of Fairness • Herik et al. 2002: • A game is considered fair if it is a draw and both players have a roughly equal probability of making a mistake.” • “一個遊戲是公平的話,那麼它必須是個平手的遊戲,且雙方有相同的犯錯機率。” • Problem: • hard to have a perfect model for calculating the probability of making a mistake • On the contrary, it is relatively easy and possible to show when a game is unfair.
Unfairness • Definitely unfair, • if it has been proved that some player wins the game. • For example, Go-Moku (in the free style) • Monotonically unfair, • if it has been proved that one player does not win the game. • For example, for Connect(k,p,p) or Connect(m,n,k,p,p), based on the so-called strategy-stealing argument. • Empirically unfair, • if most players, in particular professionals, have claimed that the game favours some player. • For example, before Go-Moku was solved, Go-Moku was empirically unfair.
Potential Fairness • Potentially fair, • if it has not yet been shown or claimed to be definitely unfair, monotonically unfair, or empirically unfair. • Properties: • A potentially fair game for the time being may not remain potentially fair in the future. • If a game remains potentially fair any longer, it could have a higher chance to be fair.
Strategy Stealing Argument • Raised by Nash in 1949 • The argument shows that W does not have a winning strategy in Connect(m,n,k,p,p). • Proof: • Assume by contradictory that W has a winning strategy, say S. • B makes the first move at random. • Then, B simply follows S (stealing W’s strategy) which leads to a win. • If the strategy S requires B to place on the squares where B placed earlier, B chooses other squares at random again, instead. • Thus, B still remains following the strategy S to win the game, contradictory to the assumption.
Breakaway and Fairness • A breakaway move: (脫離戰場 ) • place stones far away from the major battle field • An initial breakaway move: • The first move by W (after the first move by B) is also a breakaway move. • Fairness and Breakaway: • If W makes an initial breakaway move without a penalty, then the game is played like Connect(k,p,p) with W playing first. Such games are monotonically unfair or
Connect(9,6,4) • Breakaway
Empirical Analysis • For most δ = k–p = 3 games, most are empirical unfair according to our experiments. • B (W): Informally proved. FB (FW): Favors Black (White)
Maker-Breaker • Strategy stealing argument: [Csirmaz 1980, Pluhar 1994] • Connect(m,n,k,p,p), monotonically unfair (白不會贏) • So, for combinatorial analysis, some researchers proposed Maker-Breaker Model: • W is not allowed to win. • Theorems: • Let values k and p satisfy a condition, roughly like δ = k − p = O(log2p / log2 log2p). For all q, 1 ≤ q ≤ p, B wins Connect(k,p,q). • Let values k and p satisfy a condition, roughly like δ = k − p = Ω(log2p). For all q, 1 ≤ q ≤ p, both B and W tie for Connect(k,p,q).
(=k-p) 我們證明必為平手 = (log2 p) 其他潛在公平的遊戲 六子棋 = 3 我們分析為對某一方有利 p K子棋的公平性分析 • K子棋的公平性分析 • 尚未解的其他潛在公平K子棋遊戲。
Outline • Introduction to Connect6 • Characteristics of Connect6. • Fairness of Connect6 (六子棋) • Compared with Gomoku (五子棋) or Renju (連珠棋). • Threat-based winning strategies for Connect6 players or programs. • Impact of Our Research • Report of some Tournaments • New challenges in the future.
Threat-based Strategy (迫著戰術 ) • One has t threats (迫著) If the other needs t stones to defend from connecting 6. • B wins for 3 threats.
List of Single Threats (單迫著) Single defense point. Further Attacking.
死三活二 單一子形成「單迫著」 還保留另一「單迫著」 。 也可成活二 (兩子成「雙迫著」)
Threat-based Search Strategy迫著為主的搜尋策略 • VCF (Victory by Continuous Fours): 所謂的追四勝 • All double threats. • Well known since Connect6 was introduced. • Double threats mixed with single threats. • New developments.
Winning Moves • Winning moves withsingle threats. • Winning moves withdouble threats.
Null Move Heuristics • Like null moves in Chinese Chess or Weichi, • Need to find threats. • Need to build a relevance zone. • R1-Zone: For one null move. • R2-Zone: For two null moves. • R3-Zone: For three null moves. • A powerful technique to solve game positions. • Theorem 1: Black wins in Connect(6,2,3). White cannot make a breakaway move. (不能脫離戰場)
Winning Openings for Connect6 • Theorem: B wins for the following openings. (17th popular opening)
Key of the Winning Move • This move uses a special null-move technique.
More Winning Openings for Connect6 • We have found • At least 5 openings with B winning. • 4 of them are popular openings. • We are verifying them before announcing.
Outline • Introduction to Connect6 • Characteristics of Connect6. • Fairness of Connect6 (六子棋) • Compared with Gomoku (五子棋) or Renju (連珠棋). • Threat-based winning strategies for Connect6 players or programs. • Impact of Our Research • Report of some Tournaments • New challenges in the future.
六子棋提出後之後續發展(1) • 被登錄到國際線上維基百科全書。 • 成為電腦賽局競賽項目之一 • 奧林匹亞電腦賽局(Computer Olympiad) • 中國象棋電腦博弈錦標賽暨機器博弈學術研討會(Chinese Computer Games Conferences-CCGC 2006) • 六子棋公開賽之舉辦 • 交通大學盃: • 第一屆(2006):人數86人! • 第二屆(2007):人數92人! • 群想盃: • 第一屆(2006):人數約60人! • 俄羅斯六子棋大賽:Andrey主辦
後續發展(2) • 六子棋遊戲網站:如 • www.cycgame.com (群想遊戲網站;中文語系;即時;有許多台灣中國高手) • littlegolem.net (英文語系;非即時;有許多東歐高手) • pente.com (英文語系;即時) • www.brainking.com (多國語系;即時) • 總人次 > 100,000