Forty-eighth KGS Computer Go Tournament

Sunday June 21st 2009

These results also appear on an official KGS page which links to the game records.


 Formal division
format9-round Swiss
board size19×19
time18 minutes plus 25/20s


The first round started at 08:00 UTC.



There were fewer entrants than I had hoped for, perhaps because of a clash with the CGF special meeting, held the same weekend in Tokyo. This was particularly unfortunate as this KGS event was held at a time suitable for Japanese entrants but not Americans.

Shortly before this tournament, I decided to reboot the Windows desktop which I use to connect to these events, and for all general purposes such as writing this report. It failed to reboot, its boot loader had gone missing. I was able to connect to KGS using an old laptop with very limited features, but I was not able to devote as much attention as I would have liked to the tournament. Hence this report is shorter than usual. (My desktop is now fixed.)

Zen19 vs Aya
Moves 105-112.


Each program played each other program three times. Zen19 won all its games except for its round 5 game SGF with ManyFaces1. ManyFaces1 beat AyaMC once and lost to it twice. Orego lost all its games.

In round 4, AyaMC played 105 in the position shown to the right, against Zen19 SGF. This threatens to kill the small white group to its right. White can live this group, or Black can kill it, by playing at 112. If this small white group lives, the black group directly to the left of it will die; so a move at 112 appears to be worth over 50 points. However, moves 106-111 were all played elsewhere.

Moves 106-111 were all big, and may have been bigger than 112, I am not strong enough to judge. But I suspect that most strong kyu and weak dan players would have played 112 in response to 105.

However Magnus Persson has responded to the above, his responses are in italics:

In the game white 106 threatens to capture the left side and most importantly avoid the dangers of a huge semeai. If black does not play 111 the game is over without a fight. After white 112 at least black has a chance that white blunders and get a seki or wins a semeai in the center.

Saving the white group with 112 is only big if the black group dies.

Later on in the game white takes the ko at J11 for move 134 and after a ko threat black captures the right side group but white get the big left side for sure and a large endgame move at the bottom, for a clear won position, which illustrates the one sided nature of this game after move 105.

By the way the main meaning of 105 is to threaten the center white group since if black tries to surround the white group white can connect and live with all white stones with 105.

So in my opinion both programs played well moves 106-112. White simply played forcing moves that defended a won position and black tried to keep the game complex enough in order to have a possibility of winning.

But it must die – unless the white central group dies. And the white central group can get out with n13, can't it?

It can get out, but it is not that easy to read out. And this uncertainty is what strong players (and MC programs) try to avoid.

Processor numbers, power, etc.

Aya, running on Opteron 852 2.6GHz x4 (4 cores)
Many Faces of Go, running on a two core, Core2 Duo, 2.1 GHz
running on 2 x 3 GHz Dual-Core Intel Xeon
Zen, unstated but probably running on a quad-core AMD Opteron