Fifty-third KGS Computer Go Tournament

Sunday November 8th 2009

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


format10-round Swiss
board size19×19
time14½ minutes plus 25/20s


The first round started at 16:00 UTC.

Result table


The numbers in these tables do not add up as you
might expect. This is because a player that does
not show up for a round still receives half a point
for the "bye", while its opponent received a full point.

MoGoBot1 was MoGo, not MoGoTW as last month.

Aloril entered GnuGoAl, an unofficial versions of GNU Go. He also entered his usual collection of weaker bots, and allowed me to remove IdiotBot if it would make the numbers even.


Before round 1, there were 11 entrants: the nine listed above, IdiotBot, and break. I removed IdiotBot to make the numbers even.

In round 1, break did not appear. A version of it called 'break9' was present, but it was 'break' that was registered for the event, and it is not in my power to change the names of registrants after the event has started (nor, I think, for the previous two minutes). Break lost on time against GnuGoAl.

GnuGoAl vs Fuego
After move 13.

In round 2, played a high and random-looking fuseki against GnuGoAl, which was to be typical of its games in this event SGF. The position after 13 moves is shown to the right. Fuego eventually won this game.
       Some observers compared Fuego's opening moves with those of Robert Jasiek, a German 5-dan player and rules specialist, who plays on KGS as 'sum'. I beieve that the comparison is misleading. Fuego plays like this because its MC analysis finds that these moves are as good as any other. Robert plays like this (but with more attention to the relationships between his stones) because he knows that such moves are not much worse than "normal" ones and he has much more experience than most opponents in handling the games that they lead to.
       In round three, Fuego's author Martin Müller explained that its opening book had not been loaded.

Break still did not appear. Its author admitted that he had forgotten its password; and when he tried to get a new password for it, he found that it had not connected to KGS for more than six months, and another user now controlled the account of that name. I removed it from the draw; unfortunately, this left 11 players, so that byes were needed for subsequent rounds.

Fuego vs MoGoBot1
After move 16.

In round 3, Fuego had its only loss of the tournament, against MoGoBot1 SGF. As usual Fuego started with a structureless high fuseki, as shown to the left.

Fuego vs pachi
Move 138.

In its round 4 game with pachi SGF, Fuego made the move shown to the right. This is a large gote move, but it is not as big as a move at a. It seems that Fuego "missed" the need to defend here, perhaps not "seeing" the shortage of liberties on its 13-stone group around o6.

In round 5, Fuego played WeakBot50k and easily got a won position SGF. However WeakBot50k did not resign. Now, I consider a purpose of these events is to encourage programmers of strong programs to make them robust against various behaviour by their opponents, and it is with my encouragement that WeakBot50k does not resign, as a test for its oponents. (If I ever split these events into two divisions again, I may make the ability to beat WeakBot50k into a requirement for playing in the upper division.)
      Fuego ran out of its main time on move 359, and entered the severe Canadian overtime of 25 moves in 30 seconds. It had nothing to think about that required any calculation, but it used almost all its overtime. On move 405 it had two seconds left to play the last of its 25 stones, before entering the second overtime period; and when the game ended, it had four seconds left for five stones. It got away with cutting things very fine; this could cause a loss if there were significant lag.

In round 6, Fuego played ManyFaces2, and took 3½ minutes to make its first move SGF. After that it played normally, and won.

In round 8, Pachi achieved a won endgame against Fuego SGF. But it got into overtime, and when required to play 25 moves in 30 seconds, it managed only 8 of them, and so lost on time.

Processor numbers, power, etc.

Aya, running on Opteron 852 2.6GHz x4 (4 cores)
Fuego, running on an 8-core Intel Xeon 2.5Ghz, 8GB of memory.
GNU Go, running on Aloril's Linux-based laptop, 2GB RAM, Intel Celeron M CPU 530 @ 1.73GHz
Many Faces of Go, running on a quad core Q8200, 2.3 GHz.
MoGo, unspecified but probably running on R900 16-cores Intel(R) Xeon(R) CPU X7350 @ 2.93GHz
Orego, unspecified but probably running on 2 x 3 GHz Dual-Core Intel Xeon
pachi, running on four cores of i7 920 @ 2.67GHz.
running on one processor of a 4GiB RAM, AMD Athlon(tm) 64 X2 Dual Core Processor 4000+, shared with WeakBot50k
running on one processor of a 4GiB RAM, AMD Athlon(tm) 64 X2 Dual Core Processor 4000+, shared with SimpleBot