Sixty-eighth KGS Computer Go Tournament

Sunday February 6th 2011

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


format16-round Swiss
board size19×19
time14 minutes plus 25/60s


The first round started at 08:00 UTC.

Result table

Zen19 Erica pachi AyaMC valky ManyF Stone Orego PueGo gomor PNUGo oakfo Czech Simpl WeakB Idiot
B0519½ W114R B01R W112R B14R W18R B17R W116R W13R B111R B115R B16R B113R B110R W12F W19236½ 14145120Winner
2EricaBot W1519½ B014R
W02R W113R B116R W09R W13R W115R W16R B18R W112R W14R W17R B11033½ B111R W11268½ 13152115
3pachi2 W11R B012R B12R B013R W016R
W15R B115R W06R B14R B111R W110R W19R B13R B114 B18R B17R 12159109
4AyaMC W04R B19R B05R W015R
W11R B112R W110R B16R W02R W07R B113R B13R B111R W114R W116110½ B18353½ 1113583
5valkyria19 B08R B03R B015R B16R B01R W012R
B19R B17R W15R B110R B116R W011R W14R B12141½ W113186½ B114353½ 1013667
6ManyFaces1 W07R B016R B06R W04R B010R W09R
B11R W114R B13R W111R B15R W12T W18R B115183½ W112174½ W113274½ 1013460
7StoneGrid B03R W08R W011R W06R W07R W01R B014R
B113R W116R W15119½ W115R B14R W112R W110193½ B19185½ B12289½ 912949
8Orego12 W011R B012R B010R B12R B05R W03R W013R B016R
W04R B18 B115 B114R W06R W1124½ W17100½ W1954½ 713040
9PueGo W015R B04R B09R B17R W010R B011R B05119½
B02F B114R B012R W016R B13T B113R B18141½ B16105½ W11224½ 712839
10gomorra3 W06R B07R W03R W013R W016R W05R B015R B14R W12F W014R
W11T B112R W09R B11177½ B1824½ B110209½ 712836
11PNUGo W013R W01033½ W014 W03R B111R W04R W08 W015 W112R B116R B01T
B19R B05R W17368½ W12300½ B16353½ 712539
12oakfoam W010R W08R W011R B04R B02T W014R W03T W012R W09R
W11F B16 W113 B1542½ W11617½ B17231½ W115256½ 710722
13CzechBot B02F W011R W07R B014R B08R B012R B16R W013R B19R W15R B01F
B1431½ W01023½ W01511½ W1362½ B116353½ 611625
14SimpleBot B09236½ B016110½ W02141½ W015183½ B010193½ B0124½ W08141½ W01177½ B07368½ W06 B013 W0431½
B13329½ B114243½ W15368½ W112368½ 41108
15WeakBot50k B01268½ B013186½ B012174½ W09185½ B07100½ W06105½ W0824½ B02300½ W0542½ B01617½ B11023½ B11511½ W03329½ W014243½
B14225½ W111274½ 410412
16IdiotBot W08353½ W014353½ B013274½ W02289½ B0954½ B01224½ W010209½ W06353½ W07231½ B015256½ B0362½ W016353½ B05368½ B012368½ W04225½ B011274½

Sixteen players registered. Apart from CzechBot, PueGo and PNUGo, which missed part or all of their games in the first two rounds, all 16 players played all their 16 games.

We welcomed a new player, Oakfoam. It is the creation of Francois van Niekerk, a student of engineering in South Africa. It uses RAVE, with Mogo-style 3×3 patterns, and is open source.


In round 1, StoneGrid chased ManyFaces1's stones in a ladder for nine moves, then noticed that ManyFaces1 had a ladder-breaker, and tenukied. ManyFaces1 won the game.

The bots operated by Petr Baudiš were not behaving well. PueGo joined its game several minutes late; PNUGo joined its game towards the end of its first overtime period, and found that it had one second left to make 25 moves; and CzechBot, which was connected to KGS but was not in the Computer Go room, did not join its game at all. Petr's own bot, pachi2, was unaffected; it was being operated by Jean-loup Gailly.

In round 2, oakfoam (which had had a walkover against CzechBot in round 1) played very slowly, and lost on time. I suggested to its operator that he should adjust its time setting.

PNUGo played normally, but PueGo and CzechBot did not play at all, they were both involved in games with other KGS users.

At the end of round 2 I telephoned Petr Baudiš, and told him what was happening. He realised that his bots were not set to tournament mode, so they were joining non-tournament games whenever a tournament game was not available. He changed their configuration files to correct this.

I ought not to have done this. As tournament director, I ought to be impartial, and not give help to any of the participants; even if I try to give help impartially I cannot, as I do not know all the participants' telephone numbers, and I do not share a common language with some. But Petr is running these three bots, which he did not create, in the common interest, and I judged that alerting him was better than removing them from the tournament.

oakfoam vs PueGo
Moves 33-80.

In the round 3 game between oakfoam and PueGo SGF, PueGo tried to catch one of oakfoam's stones in a ladder, as shown to the right. It steered the ladder away from a white ladder-breaker, but this left a black stone in atari, and the white stones eventually escaped. However oakfoam's time management was still poor, and it eventually lost on time.

ManyFaces1 vs pachi2
At point where White resigned.

The round 4 game betwen ManyFaces1 and pachi2 was interesting. SGF They divided the board into quarters, each having two opposite quarters of the board. But pachi2 had put a large dent into ManyFaces1's quarter on the right edge, and won the game by resignation in the position shown to the left.

After round 4, pachi2 was the only undefeated player.

Stonegrid vs PueGo
After final passes.

In round 5, Zen19 suffered its second loss, to EricaBot SGF.

Also in round 5, the game between StoneGrid and PueGo SGF ended in the position shown to the right. The game was now scored with all the stones counted as alive. This is clearly wrong. I think the large white group in the left side is dead, and I am sure the small white group in the bottom left corner is dead. If I am correct about the left side, then Black, PueGo, should have won.

In round 6, oakfoam showed that it had learned to play fast enough, at least once into overtime. It got into overtime in its game with SimpleBot SGF, played badly by filling its own eyes, and failed to handle the clean-up, treating all SimpleBot's dead groups as alive; but it still won.

Also in round 6, pachi2 suffered its first loss, to valkyria19 SGF.

In round 7 CzechBot was slow to make its first move in its game with pachi2. Petr found that this was because it has a memory leak, and so should be restarted between games if it is to play well.

After round 7, pachi2 and EricaBot were tied for first place, with six wins each.

In round 8 gomorra3 failed to clean up WeakBot50k's dead stones, and won anyway.

In round 9 AyaMC beat EricaBot, leaving pachi2 in the lead with eight wins.

In round 10 CzechBot played slowly in its game with WeakBot50k. When it had only three seconds of main time left, it started passing, and made no further moves. WeakBot50k continued to put stones on the board, saving three dead groups and winning the game. It seems that CzechBot was, or became, unaware that there would be overtime.

In round 12 Zen19 beat pachi2 SGF. This put them both on ten wins, along with EricaBot.

valkyria19 vs WeakBot50k
Moves 55-84.

In round 13 valkyria19 played out a losing ladder against WeakBot50k SGF, as shown to the left. However valkria19 was still able to win the game: WeakBot50k is reliable at spotting atari, but can do little else.

EricaBot vs pachi2, top right corner
Move 175.

Also in round 13, EricaBot beat pachi2 SGF. One of pachi2's moves is shown to the right. This move is definitely sente, it threatens to win the semeai in the corner; but if answered, it achieves nothing, in fact it is a small loss as Black now has one fewer forcing move on the outside. I have observed that when an MC-based bot makes such a move, it is desperate and is going to lose the game, even if the statistics it reports don't reflect this.

EricaBot and Zen19 were now leading with 11 wins from 13 games, with pachi2 one point behind.

In round 14 Zen19 beat EricaBot, while pachi2 had an easier win against PNUGo. This left Zen19 in the lead with 12 wins, and EricaBot and pachi2 a point behind it with 11 wins.

In round 15 as in round 10, CzechBot played slowly in its game with WeakBot50k, started passing when it had only three seconds of main time left, and lost when it allowed WeakBot50k to make many undeserved gains.

Zen19, EricaBot and pachi2 all won their games in this round, leaving them on 13, 12 and 12 wins respectively.

In the final round 16, EricaBot played and beat pachi2, while Zen19 played and beat ManyFaces1 to win the tournament.

The players receive points for the 2011 Annual KGS Bot Championship as follows:



Details of processor numbers, power, etc.

Aya, running on 6 cores of an i980X 3.3GHz
MoGo, running on an 8-thread i7 920, 6GiB RAM
Erica, running on a i7 950, 4 cores, each 2.7 GHz, Windows 7.0
Gomorra, running on a cluster of 16 nodes with 6 cores each, thus 96 cores. Cores are running at 2.67 Ghz. Nodes are connected through 4xSDR InfiniBand.
running on Linux, 2GiB RAM, Intel(R) Celeron(R) M CPU 530 @ 1.73GHz
Many Faces of Go, running on 4 cores.
Oakfoam, probably running on an AMD Athlon(tm) 64 Processor 3000+.
Orego, running on one of the five nodes of a custom Linux cluster build by PSSC Labs:
pachi, running on 32 unspecified 20-core platforms.
Fuego (not the latest version), running on an 8-thread i7 920, 6GiB RAM
GNU Go, running on a 2-core system, probably a Core2 Duo E7200 2.53GHz
running on Linux, 8GiB RAM, AMD Athlon(tm) 64 X2 Dual Core Processor 4000+
StoneGrid, running on an Intel Core2 Quad Q9400 (2.66GHz, 6MB L2, 1333Mhz FSB)
running on Linux, 2GiB RAM, Intel(R) Celeron(R) M CPU 530 @ 1.73GHz
valkyria, running on a i7-860 4 Core processor at 2.80 Ghz.
Zen, running on a Mac Pro 8 core, Xeon 2.26GHz