Search Articles

Articles

Hungarian Computer Gomoku Tournament 2005

Hungarian Computer “Go-Moku” tournament, 2005

Authors: J. Wágner, I. Virág

We held a “Go-Moku” tournament on 6^th June 2005. It was a small competition with 3 Hungarian programmes, but there were some really interesting games. We decided to make a bigger effort next time to collect more participants. The next tournament will be held in October 2005.

The rules of the tournament

The board size was 19x19. The overline (six or more stones in a row) was considered as a win for both sides. There was a 20-minute time limit for each programmes plus a 25 sec byo-yomi. Games started from positions put up by organizers (first 5 moves). There were no more restriction after the 5th move.

Opening rule was the “Shape Rule”

The organizers prepared several balanced 5 stones (3 black, 2 white) opening positions. There were some additional rules about opening positions. All position contained 5 stones (3 blacks + 2 whites) near the centre of the board (within 5x5 area around the center of the board). One black and one white stones were within 3x3 area around the center of the board. We published a list of opening positions before the tournament. Everybody had an option to say that one or more positions were not suitable because black or white had a big advantage. We randomly selected 2 position for this tournament, and all programmes played in the same conditions, using both starting positions and both colours. So all two programmes played 4 games and winner got 2 points. If the result was a draw then both programmes got 1 point. If two programmes collected the same amount of points then the number of won games decided about the position.

Programmes

Program’s name	Author(s)	Program’s features
Hector for Go-Moku (Hungary)	Csaba Jergler	Alpha-beta searching method with special threat space search. Hash table size: 200 MB Speed: 100000 nodes/sec Opening book: —
Tyson (Hungary)	Gábor Takács	Alpha-beta searching method + VCT engine. Hash table size: varying on position and searching depth. Speed: 20000-50000 nodes/sec Opening book: —
OniX (Hungary)	János Wágner István Virág	Minmax pruning with selective deepening and VCT engine. Hash table size: 250 MB Speed: 10000-50000 nodes/sec Opening book: average
Swine (Czech Republic)	Jiri Fontan	Alpha-beta searching method combined with modified PN search. Speed: 50000-200000 nodes/sec Hash table size: 16 MB Opening book: small
Goro (Russia)	Viktor Barykin	No information

Results

Program	1	2	3	Points	Games
1. OniX	—	3:1	3:1	4	6:2
2. Hector	1:3	—	4:0	2	5:3
3. Tyson	1:3	0:4	—	0	1:7

We wanted the best programmes from the competition in Prague (Goro and Swine) to compete as well, but there were some technical problems. Their competition had slightly different rules. (3 min. time and 20*20 board size.) Board size is usually not very important because most games finish before reaching the board edge. So we used the same versions as in the last Prague competition although the results are not fully comparable.

Goro played quite well in some games, but sometimes it made strange mistakes, maybe because we did not understand the protocol properly (sometimes Goro did not move and the special interface interpreted this as if Goro’s move would have been in the corner: A19), so we decided that Goro could not entered to the tournament. We got a new version of Swine which could play on 19*19 board and used more thinking time but this programme lost all its games. The previous version of Swine was stronger, so we assumed that this was a mistake because of the lack of time. If we can solve this problem we will repeat all games of Swine and maybe we will need to make a post-competition table. There was the same kind of last minute modification in Tyson too which could effect its result.

An interesting fact is that if programmes sensed they had a loosing position then they played out all their fours (or threats) and this way got into a more hopeless position.

Average length of games: 40 moves

Computer hardware: Intel Pentium 4 3.0 GHz with 512 MB RAM

Selected opening patterns

Some selected games

Game No. 5

Hector (black) – OniX (white) 1-0, 45 moves

This opening is not an easy one, both sides had many possibilities to start attacking. Moves from 6^th to 14^th were a possible line in this opening. Somewhere around moves 15, 17 or 19 black lost its confident advantage. 20^th move was a very strong white move, I think it was a victorious move, but Hector found a tricky continuation by 21. L10, 23. H13 and 25. H15, moves 23 and 25 created a VCF threat. 22. M9 was the only correct defense, if 22. i13 then black would have had a brilliant VCT win by sequence 23. K7, 24. L8, 25. L6, 26. i9, 27. K5 etc… 24. J14 was good defense, but 26 H14 was a wrong move (correct 26 would have been i13). After this black had a very deep and hard VCT win. Hector found this win. At move 35, OniX realized that white was in losing position, it seemed that white played out his all fours (36, 38) and played a very poor 40^th move. 40. G11 would have put up better resistance. After the game we analyzed this position with OniX’s strong VCT engine. OniX found the win within 10 minutes by analyzing 5 million positions, the deepest VCT line was 35 moves long. Unfortunately the tournament version of OniX did not analyze so deeply, that was the reason why it lost this game.

Game No. 7

Hector (black) – OniX (white) 0-1, 72 moves

With moves 11, 13 and 15 black built up a strong base for the attack. I think 17. H6 was not the best move, 17. G6 would have been better, or maybe even a winning move. I think 20. J8 was a losing move, 20. G5 would have been better. 21 was a very strong move, it was a winning move, but Hector did not see the victory line well. If 22-J4 then black could have won by 23-G5. If 22-G7 then black could have won by sequence 23. H4, 24. J6, 25. G5. Now Hector saw what it considered a victory line for black, but I think it was a miscalculation, because 23. J6 was a mistake. After this Hector lost its fast winning chance. Hector continued its attack by sequence 25^th-35^th, now with 36. K6 white achieved a favorable position. Black continued his attack by 37^th, 39^th and 41^st. Hector tried to build up new offensive area by 43. K9. OniX understood well the position and ignored Hector’s attempt (43^rd was not VCT threat). 44. M5 was the decisive move. The rest of the game is black’s agony.

Game No. 8

OniX (black) – Hector (white) 1-0, 71 moves

The opening sequence was very sharp, Hector played first moves very aggressively (6-16). 10. H10 was very strong move. 16. K11 was tricky move, 17. i13 was not enough safe. OniX did not understand meaning of the 16^th. 18. M9 was the victory move for white, both programmes saw that this was white’s game. But later we amazed that Hector missed the winning line: 24. J11 was bad, 24. M7 would have been correct move. Hector tried to win in the lower side of the board by sequence 26-38, but OniX defended it well and later won in the upper area.

All games

Game No. 1

Hector (black) – Tyson (white) 1-0, 39 moves

12^th and 14^th were very nice moves, Tyson developed a very strong position. Hector tried to attack with moves 15. N6 and 17. L6. 18. H11 was a terrible mistake, after that Hector found a very nice VCT victory.

Game No. 2

Tyson (black) – Hector (white) 0-1, 26 moves

7. K11 was a disaster, it was a typical Tyson mistake (see games No. 13 and No. 17). 8. H10 solved the problem, after this white could win easily anywhere.

Game No. 3

Hector (black) – Tyson (white) 1-0, 71 moves

The opening sequence was well played (6-21). 22^nd move was unnecessary. 25^th was a tricky move, only 26. K11 could defend it. 29^th , 30^th , and 32^nd were strong moves. White started an attack by 38. F9 and 40. G9, but this attack was no use (black got some extra stones). 42^nd, 44^th, 50^th and 52^nd were pointless moves, black received four extra stones for no expense. Tyson gave full point to Hector.

Game No. 4

Tyson (black) – Hector (white) 0-1 (timeout), 74 moves

The opening sequence was normal (6-20). 29^th was a nice pause move, Hector started to fill in left side of the board by attacking. It was a very nice strategy from program Hector. 43. D7 was too safe, 43. D11 would have been better, and then black would have had a real winning chance. After that there were some unhelpful moves: 53, 55, 57 and 59. Anyway it was an interesting game, unfortunately Tyson lost it by timeout.

Game No. 6

OniX (black) – Hector (white) 1-0, 33 moves

9. i9 was a nice move. 10. J11 was a blunder, Hector saw what it considered a victory line for white, but it was a miscalculation. 11. J11 was the best defense and it was enough for black to win the game.

Game No. 9

Hector (black) – Swine (white) 1-0, 39 moves

In the opening, black player increased his advantage progressively by white player’s small inaccuracy. After 21^st move Hector had big advantage and won the game after careless 22. F10.

Game No. 10

Swine (black) – Hector (white) 0-1, 48 moves

I was an interesting game. The opening sequence was balanced (6-13). The 14^th move was very nice, 15. G12 was the strongest defense. After this, Hector continued his attack on the right side. 21. N7 was too careful, 21. J11 would have been the best, after that black could win easily. Next similar cases: 23^rd and 25^th were also very cautious, 23. H13 and 25. J11would have been enough to win the game by Swine. 33^rd and 35^th were last mistakes, Hector won the game by brilliant 38. O8.

Game No. 11

Hector (black) – Swine (white) 1-0, 29 moves

8. i12 was the losing move (8. J11 or H11 would have been better), 9. i13 was not the most optimal move (9. i10 would have been the best move), but it could be enough to win the game by Hector.

Game No. 12

Swine (black) – Hector (white) 0-1, 24 moves

7. K10 was a weak move. Hector found nice winning move: 8. H10.

Game No. 13

Tyson (black) – Swine (white) 1-0, 37 moves

7. K11 was a disaster. If 8. H10 then white could win easily anywhere, but Swine played 8. G11 and 10. G12. After 12. K10 black had a winning position, but Tyson also missed the most optimal win (13. H9 would have been the correct move).

Game No. 14

Swine (black) – Tyson (white) 0-1, 26 moves

10^th and 12^th were interesting moves. It is hard to understand why black defended the three in the 17^th move on the left side (17. F11), why not on the right side (17. J11). 17. J11 would have been a winning move. 19. H10 was not the best move either, after 18. H9 black player had a nice VCT victory by sequence: 19. H12, 20. K12, 21. F12, 22. G12, 23. H14, 24. i13, 25. G13 etc… 21. F10 is an unbelievably poor move! The correct move would have been 21. K9 which would have given a black advantage.

Game No. 15

Tyson (black) – Swine (white) 1-0, 29 moves

8. i12 was a losing move (8. J11 or H11 would have been better), it was a typical Swine mistake (see game No. 11), moves 9^th to 13^th were all logical, Tyson found the direct win.

Game No. 16

Swine (black) – Tyson (white) 0-1, 28 moves

7. i12 was a mistake, white had a direct attack and after 15. i10 Tyson had a VCT sequence on the lower side of the board.

Game No. 17

Tyson (black) – OniX (white) 0-1, 28 moves

7. K11 was disaster, it was a typical Tyson mistake (see games No. 2 and No. 13). 8. H10 solved the problem, after this white could win easily anywhere.

Game No. 18

OniX (black) – Tyson (white) 1-0, 39 moves

11. i9 was a very strong move. Move 20^th was poor. 20. H10 was better. After 20. J12 black had a very hard VCT victory. 21^st and 25^th were very nice fukumi moves.

Game No. 19

Tyson (black) – OniX (white) 1-0, 41 moves

Both programmes played the opening phase of the game well. Later OniX made a mistake at move 18, 18. K10 would have been better. 19^th and 21^st were very nice moves! After 26. K10 Tyson found the VCT sequence.