The computer Connect6 tournament was held as part of the Computer Olympiad 2019, which took place in Macau, China on August 10–16. A total of four teams participated in the tournament. The participating teams and the final standings are listed in (Table 1). All four of the team’s programs were executed on Windows-based PCs. Each team obtained a total score of 10, 8, 4, or 2 points. The participants and the cross table are listed in Table 1. The game Connect6, a kind of six-in-a-row game, was first introduced by Wu and Huang (2005) and then described in more detail by Wu et al. (2006).
BIT_Connect6 won the gold medal at the 2019 Computer Olympiad Connect6 tournament by winning 5 of the 6 games with their only 1 loss to Kavalan (Yang and Yen, 2019; Yen and Yang, 2011; Yen et al., 2012). Kavalan won the silver medal. Kavalan respectively lost one game to BIT_Connect6 and Explorer and achieved total score of 8 points. Zeta won the bronze medal with a total score of 4 points. Table 2 shows the final standings. Each of the four programs had an opening database for the opening game. As for the search algorithm, Kavalan’s was developed based on Monte-Carlo tree Search (MCTS) using a convolutional neural network, while the other programs were based on the alpha-beta search method. Kavalan used a pre-trained convolutional neural network to perform predictions of promising moves generated from the MCTS leaf nodes instead of using the play random game (or playout) method for their predictions.
The participants and the cross table
Program
BIT_Connect6
Kavalan
Zeta
Explorer
Total score
BIT_Connect6
–
2
4
4
10
Kavalan
2
–
4
2
8
Zeta
0
0
–
4
4
Explorer
0
2
0
–
2
In analyzing the positions that Kavalan lost in the Connect6 tournament, the position that the Opponent team played on the fourth move was the same, as shown in (Fig. 1). Although the move sequence and relative position of Black and White stones look different in Fig. 1, it was the same after rotating either side of the Board position. In the two games, Kavalan played the first move (Black stones) and the Opponent (White stones) played the same second move. Because Kavalan played the same third move to defend against the Opponent second move, it can naturally play the same fourth move. Therefore, the Opponent took the advantage of Board position in the fourth move. A good opening move has a great impact on the game outcome.
The final standings
Ranking
Program
Author
Country
Points
1
BIT_Connect6
Chai Zenghao, Han Youfang, Gong Kaixiong, Wang Yuhan, Zhang Qian
China
10
2
Kavalan
Jung-Kuei Yang and Shi-Jim Yen
Taiwan
8
3
Zeta
Chen Zhang
China
4
4
Explorer
Li Xiaorui, Zhang Yunpeng, Li Yaguang, Zhou Ke
China
2
Pposition analysis of Kavalan v.s. BIT_Connect6 and Explorer.
This report provides commentary on three games, Kavalan (Silver) versus BIT_Connect6 (Gold), Zeta (Bronze) and Explorer. The first game, Kavalan (Black) vs. BIT_Connect6 (White), shown in Fig. 2 below, is the only game in this tournament in which BIT_Connect6 was defeated by Kavalan. KAVALAN achieved this victory on 12th move (win-position) over BIT_Connect6. According to the above discussion, when BIT_Connect6 play the fourth move as shown in Fig. 2, it took the advantage of the Board position. Finally, BIT_Connect6 found the win-position in 12th move (double-threats move). Figure 3 shows the Kavalan v.s. Zeta game. The 2nd move for Zeta was different from BIT_Connect6 and Explorer. In this game, the key move was made on the 7th move of Kavalan where Zeta did the wrong blocking move to defend against Kavalan’s attack as shown in Fig. 3. Therefore, Kavalan found the wining position on the 9th move. The development of the Board position shown in Fig. 4 is roughly the same as Fig. 2. The difference is that Kavalan’s 7th move used one stone to block the single-threat move of White and the other stone to generate the single-threat move. In the 6th move, White as shown in Fig. 2 and Fig. 4 already took the advantage of the Board position. Therefore, no matter how Black played their remaining two stones to block the attack it would have virtually no effect on the ultimate result of the game.
Black: Kavalan, White: BIT_Connect6, Moves 1–12.
Black: Kavalan, White: Zeta, Moves 1–9.
Black: Kavalan, White: Explorer, Moves 1–10.
From the right to left: Jung-Kuei Yang (Kavalan, silver medals), Chai Zenghao (BIT_Connect6, golden medals), Zhang (Zeta, bronze medals).
Footnotes
Acknowledgement
This work was supported in part by the MOST of Taiwan under contracts 108-2634-F-259-001- through Pervasive Artificial Intelligence Research (PAIR) Labs, Taiwan.
References
1.
Wu, I.-C. & Huang, D.-Y. (2005). A new family of k-in-a-row games. In The 11thAdvances in Computer Games (ACG11) Conference, Taipei, Taiwan.
Yen, S.-J. & Yang, J.-K. (2011). Two-stage Monte Carlo tree search for Connect6. IEEE Transactions on Computational Intelligence and AI in Games, 3(2), 100–118. doi:10.1109/TCIAIG.2011.2134097.
5.
Yen, S.-J., Yang, J.-K., Kao, K.-Y. & Yang, T.-N. (2012). Bitboard knowledge base system and elegant search architectures for Connect6. Knowledge-Based Systems, 34, 43–54. doi:10.1016/j.knosys.2012.05.002.