Sixty-seventh KGS Computer Go Tournament

Sunday January 9th 2011

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


format8-round Swiss
board size19×19
time29 minutes plus 25/60s


The first round started at 16:00 UTC.

Result table

Zen19 Erica ManyF pachi Czech AyaMC Stone PNUGo PueGo Simpl break WeakB Idiot Orego
W15R B16R W17R B14R W13R W11R W18R B12159½ 83636Winner
2EricaBot B05R
W12R W11T B03R B16R W14R W17R W1853½ 63724
3ManyFaces1 W06R B02R
B13R B15R B14R B18R W17228½ B11R 63319
4pachi2 B07R B01T W03R
W18R B16R B14R B1573½ B12111½ 53313
5CzechBot W04R W13R W05R
B07R B16 W11F W12108½ B18139½ 53214
6AyaMC B03R W06R W04R B08R W17R
W15R B1211½ W11156½ 43813
7StoneGrid B01R B04R W06R B05R
B1319½ B12R W18280½ W17161½ 4329
8PNUGo B08R W06 W0211½ W0319½
B11R B14353½ W17254½ W15280½ 4287
9PueGo B07R W08R W04R W02R W01R
B15125½ W16138½ W13186½ 3294
10SimpleBot W02159½ B0853½ B07228½ B01F W04353½ W05125½
W13142½ B16353½ 2342
11break19 W0573½ B02108½ B08280½ B07254½ B06138½ B03142½
W1435½ B11F 2230
12WeakBot50k W02111½ W08139½ B01156½ B07161½ B05280½ B03186½ W06353½ B0435½
13IdiotBot W01R
060Quit before round 2
14Orego12 W01F
020Quit before round 2

Fourteen players registered. However only twelve were present at the start of the tournament, Orego12 and SimpleBot were missing.


In round 1, pachi2 played out a losing ladder agaisnt EricaBot SGF. After this, it had little chance of catching up, and eventually lost on time. It was found later that although pachi2 was running on 64 20-core systems, it was only using one core of each, so its play was much worse than it should have been.

Pachi2 continued to play in its single-core configuration for rounds 2 and 3, before the error was noticed and corrected.

On the day after the tournament, pachi2 and EricaBot played another game on KGS, with pachi2 using its correct 20-core configuration SGF. Pachi2 won – not that it is relevant to this tournament.

Also in round 1, StoneGrid started well in its game with Zen19, SGF but then went to pieces and lost. Or at least, that is how it appeared to me and to some of the kibitzers. But Zen19 is now rated 4d, much stronger than me and comparable with the strongest of the kibitzers, so it is possible that it was winning all along, and we just didn't understand the game.

At the end of round 1, Orego12's owner Peter Drake appeared, and apologised for its absence. I removed it from the draw. Then, to make the numbers even, I could have removed the absent SimpleBot; but as its owner Aloril was present and had started his SimpleBot, I removed his weaker IdiotBot instead.

In round 2, ManyFaces1 appeared to be winning against EricaBot SGF: EricaBot's owner Shih-Chieh Huang commented "mfgo is leading for sure", and he is rated 6d on KGS, so I won't question his judgement. However, EricaBot was able to trash ManyFaces1's central territory, and won the game.

In round 3, CzechBot beat EricaBot SGF. This left CzechBot and Zen19 as the only undefeated players.

In round 4, Zen was drawn against CzechBot SGF, and beat it.

From the start of round 4, pachi2 was playing with its full 64×20 cores instead of just one per machine. It played noticeably better.

In round 6, Zen19 played against ManyFaces1 SGF again. Stronger players than me thought that Manyfaces1 was ahead: "Warfreak2 [4d]: zen could actually lose this game". However, ManyFaces1 made some strange moves; then after a very confusing yose, Zen19 turned out to be ahead. This put Zen19 on six wins from six games, with two rounds to play and its nearest rival on four wins.

Zen19 vs pachi2
Moves 1-17.

In round 7, Zen19 played against pachi2 SGF. The fuseki is shown to the right. I am not able to comment on the merits of these moves; I will just say that things have changed greatly since the days when programs like TurboGo could win games by making solid third-line territory and ignoring the centre until later in the game.

The players receive points for the 2011 Annual KGS Bot Championship as follows:
Many Faces of Go4

Appendix I: Late Entries

After the tournament had started, someone told me that he would like to enter his bot. I had to refuse. Regardless of my views on accepting late entries, the KGS tournament scheduler makes it impossible to accept an entry after play has started.

Attempts to enter after a tournament has started are, therefore, not a problem for me. But attempts to enter in the twenty minutes before play starts are inconvenient for me. In the last twenty minutes before the start of play, I am busy, checking for late entrants, checking whether those who have entered have in fact appeared, chatting to various people, counting the entrants and arranging to make the numbers even, etc. I try to accept entries up until a few minutes before play starts, but I discourage it.

Therefore, entries which I receive less than thirty minutes before play starts are not guaranteed to be accepted. If you try to enter your bot in the last thirty minutes before play starts, I will probably try to register it, but I will accept no responsibility if I fail.

Appendix II: Accidental Tournament on January 8th

When I announced the date of the Sixty-seventh KGS Computer Go Tournament, I gave it as Sunday January 8th. I meant Sunday January 9th, and this was soon pointed out and corrected. However I had also accidentally configured the tournament to take place on January 8th.

Unfortunately I was away from home on Saturday, from before this accidental tournament started until after it finished, so I was not aware that it was happening. Most of the players also were not present. But Zen19 was present for the whole tournament, and three others were present for some of it and managed to play at least one game each. The "result" was that Zen19 was the undefeated winner, winning three of its games by playing them and all the others by default.

This tournament should not have happened, and will not be counted for any purposes. It is recorded on KGS, at

On the Sunday morning, I set up the tournament correctly, and registered the players for it.


Details of processor numbers, power, etc.

Aya, running on 6 cores of an i980X 3.3GHz
break, probably running on a single processor Intel(R) Celeron(R), 1.7Ghz
MoGo, running on an 8-thread i7 920, 6GiB RAM
Erica (with no opening database), running on a i7 950, 2 cores with 4 threads, each 2.7 GHz, Windows 7.0
Many Faces of Go, running on a 12-core 3.2GHz Xeon provided by Microsoft.
pachi, running on one core of each of 64 unspecified 20-core platforms for the first three rounds, and on all 20 cores of them, 1280 total, for the last five rounds.
Fuego, 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
Zen, running on a Mac Pro 8 core, Xeon 2.26GHz