Sixty-sixth KGS Computer Go Tournament

Sunday December 5th 2010

These results also appear on an official KGS page which links to the records of all the games.


format20-round Swiss
board size9×9
time9 minutes plus 25/60s


The first round started at 08:00 UTC.

Result table


The numbers in this table may not add up
as you might expect. This is because some
entrants played no games, and games
against absent opponents treated as wins.

Fifteen players registered. One of these overslept and did not launch his program. The others all played in almost every round.

Fuego tried to register, but was prevented by a problem with a firewall. So it did not play, and PueGo remained registered.

We welcomed two new players to these events. 'GoCoo', by Yoshie Kato, is based on UCT, with some Go knowledge. (Yoshie is the wife of Hideki Kato.) And 'sf9x9bot' is 'Stone Fighter', by Fuming Wang.

Some observers thought that John Fan, with his bot 'StoneGrid', was also a newcomer. But StoneGrid has taken part before: in June 2008 it won the 9×9 KGS bot tournament.


In round 1, pachi arrived late for its game with PueGo because of a technical problem. With only 72 seconds left for all its moves, pachi was outplayed, and lost.

OOhikaruO2 had not appeared by the end of round 1, so I removed it from the draw.

In round 4, sf9x9bot resigned right at the start of its game with coldmilk, before a stone had been played. Its operator was trying to end its round-3 game with PueGo, which in fact it had already resigned.

In round 5, sf9x9bot obtained a won position against GoCoo SGF, with all GoCoo's stones dead. However, they were unable to handle the clean-up phase correctly. Sf9x9bot repeatedly disconnected, reappeared (with an extra minute on its clock somehow), both passed, and then disagreed about the status of GoCoo's dead groups. Eventually sf9x9bot lost on time. But the extra minutes caaused the round to end about four minutes late. Fortunately, the KGS tournament scheduler can handle this: it does not start a round until all the games in the previous one have ended, and starts each round as soon as they have all ended and the scheduled start time has been reached. So the start times for the next three rounds were all delayed, until round 9 was back on schedule. I would like to congratulate wms on getting this right!

In round 6, sf9x9bot timed out in a won position against break9. It did not know how to handle resumption.

In round 7, CzechBot and StoneGrid played an interesting game SGF. CzechBot won: this was StoneGrid's first loss.

Zen9 vs StoneGrid
Position after move 56.

In round 8, Zen and StoneGrid SGF achieved the position shown to the right. Here, it is conceivable (at least to a kyu player like me) that Black can either convert the whole upper left half of the board to a seki, or get to play enough useful stones while forcing White to capture in the corner that he catches up in territory. If I were playing, I would start at b in the diagram, hoping for something like the variation given in SGF. But StoneGrid played the meaningless move a, and went on to lose. I do not understand this. I accept that StoneGrid can "read" better than me, and may have realised that b does not work; but then nor does a. Either b or resignation must make more sense than a.

StoneGrid vs PueGo
Position at the game end.

In round 10, StoneGrid and PueGo SGF both passed in the position shown to the left, and scored it with all the stones alive, giving StoneGrid the win by 9½ points. The white group in the upper left is of corurse dead, and PueGo should have won by 17½ points.

In round 11, break9 lost on time in a won position against GoCoo.

In round 12 break9 was missing, and timed out in its game with PNUGo.

After round 13, ManyFaces1, valkyria and Zen9 were all tied for first place on 10 wins.

In round 14, Zen9 beat valkyria, while ManyFaces1 beat sf9x9bot, leaving Zen9 and ManyFaces1 tied for first place on 11 wins.

Zen9 and ManyFaces1 both won their round 15 games.

In round 16, Zen9 beat coldmilk while ManyFaces1 lost to StoneGrid, and valkyria lost to PueGo. This left Zen9 in the lead with 13 wins.

In round 17, Zen9 beat ManyFaces1 in a good game SGF. I understand little of this game, and will not try to comment on it. After this, Zen9 had 14 wins, two points ahead of its rivals, and so was almost certain to win th tournament.

In round 18 GoCoo beat sf9x9bot on time, in a totally lost position.

In round 19 break9 beat sf9x9bot when sf9x9bot failed, in clean-up, to show that break9's one-eyed group was dead.

In round 20 sf9x9bot lost yet another won game, this time against PNUGo, by its failure to support clean-up.

Sf9x9bot lost many won games by failing to support the KGS clean-up procedure, either allowing the game to be mis-scored or losing on time. If this can be fixed, I expect to see it doing well.


Details of processor numbers, power, etc.

Aya, running on 8 cores of a Xeon 5355 2.6GHz
break, probably running on a single processor Intel(R) Celeron(R), 1.7Ghz
coldmilk, running on a 12-core Xeon, 2.66Ghz
MoGo, running on double-core AMD Athlon(tm) 64 X2 Dual Core Processor 4800+ (2.5GHz).
MoGo, running on double-core AMD Athlon(tm) 64 X2 Dual Core Processor 4800+ (2.5GHz).
pachi, running on a 40-thread cluster
Fuego, running on two threads on a Core2Duo E7200
GNU Go, unspecified platform
GoCoo, running on one thread, on an IBM Think Pad T42, CPU Intel Pentium processor 1.7GHz 1.25 GB
Many Faces of Go, running on a 12-core Xeon, 3.2 GHz.
Orego, running on one of the five nodes of a custom Linux cluster build by PSSC Labs: the node has two AMD Six Core Dual Opteron 2427 2.2 GHz (12 cores total), 8 GB RAM, Centos Linux.
was running on a cluster of two Intel i5 750 computers with 8 cores
StoneGrid, running on an Intel Core2 Quad Q9400 (2.66GHz, 6MB L2, 1333Mhz FSB)
valkyria, running on a i7-860 4 Core processor at 2.80 Ghz.
Zen, running on a Mac Pro 8 core, Xeon 2.26GHz