2012 Autumn Slow KGS Computer Go Tournament

Sunday September 2nd-5th, 2012

These results also appear on an official KGS page.


format12-round Swiss
board size19×19
time3 hours each, absolute


The first round started at 22:00 UTC.

Result table

Zen19 AyaMC Crazy nomiB pachi ManyF gomor NiceG MCark Orego
B11R W19R B03R W18R B17R W04F B110R W12R W15R W112R W16F B111R 107459Winner
2AyaMC W01R B09R
B15R W111R B18R W06R B112R W14R B110R B13R W1768½ W12R 97650
3CrazyStone W13R B08R W05R B011R
B11R B12R W16R B112R B14R B17R W19T W110R 97648
4nomiBot W07R W08R W01R
W19R B111R W03R B010R W12R B112R B16R B14T B15R 77131
5pachi B14F W010R B16R W012R W02R B09R W011R
B18R B03R W15R W11228½ W17R 68032
6ManyFaces2 B02R B04R B06R W012R B13R W110R W08R
W17R B09 B111R W15F B11R 67627
7gomorra32 B05R W010R W04R B02R W012R W13R B07R W19
B11R W18R W111196½ B16R 67123
8NiceGo19N B012R W03R W07R W06R B05R W011R W01R B08R
B1283½ W11082½ B14R W19R 4656
9MCark B06F B0768½ B09T W04T B01228½ B05F B011196½ W0283½ B01082½
W13R B18R W112R 3610
10Orego12 W011R B02R B010R W05R B07R W01R W06R W04R B09R B03R W08R B012R

In the table above,
   0 is a loss
   1 is a win
   J is jigo
   left superscript is the player's colour
   right superscript is the round in which the game was played
   a subscript shows how the result was determined:
      R for resignation
      T for time
      F for forfeit
      a number for the points difference after counting.
All the 0s, 1s and Js are links to the game record.

AyaMC vs Zen19S
Moves 1-11.


In round 1, AyaMC and Zen19S began their game as shown to the right. Moves 1 to 7 look good. Move 8 does not look right to me – but I am not strong enough to question it. But move 11 is surprising, and I do not think it can be a good move.
      Zen19S's operator Hideki Kato has explained to me that during this game he found a NaN in its log file, leading him to believe that one of the machines in the cluster (the 6-core Xeon W3680) was not working correctly. He removed this machine from the cluster before the next round. He believes that Zen19S was lucky to win this game.

Zen19S vs ManyFaces2
Moves 11-24.
In round 2, Zen19S and ManyFaces2 began their game as shown to the left. Move 14 (again by Zen19S) is bad; move 24 is clearly bad. Hideki Kato now began to suspect that his cluster was suffering from low voltage in the cores of the two Xeon X5680s (which needs to be adjusted to compenstae for the overclocking). So he raised the voltage; and again (with some luck, he says) Zen19S was able to beat ManyFaces2.
Zen19S vs ManyFaces2
Position at move 94.
Later in this game, in the position shown to the right, there is a semeai in the lower left. My understanding is that Black can kill, starting at a5, while White can make seki, starting at d2. In fact Black played at a5. But at the end of the game, it had become seki again. It is possible that ManyFaces2 misevaluated this semeai.

In round 3, CrazyStone beat Zen19S, with Zen19S making some more moves that Hideki Kato thought strange. After the game, he checked the cluster again, and fell asleep before switching it on again.

In round 4, Zen19S did not appear to play its game with pachi, and lost on time.
      Also in round 4, MCark obtained a totally lost position against nomiBot, but continued making purposeless moves until it lost on time. It then lost its next two games without making a move.

In round 5, CrazyStone lost its first game, to AyaMC.

In round 7, Zen19S played no strange-looking moves against nomiBot, so Hideki Kato reattached to its cluster the machine removed after round 2.

Zen19S vs CrazyStone
Move 32.

In its round 8 game with CrazyStone, Zen19S played its move 32 as shown to the left. This received some praise from the kibitzers. Zen19S eventually won the game.

In round 11, gomorra3 did not move in its game with MCark until its operator restarted it, after two hours. Its cluster resources had expired, and it was restarted with only 16 cluster nodes.
      Also in round 11, AyaMC again beat CrazyStone.


I have two requests to make, following this event. Both are inspired by discussions with kibitzers.


Several people said it was annoying that MCark continued to play on in clearly lost positions instead of resigning. I tend to agree. MCark has some understanding of the game, it is not like IdiotBot. It must know when its position is hopeless, and it ought to resign. By resigning, it would allow its opponent to switch off his machine and save electricity, or to use it for something else, or to fix a bug. It would also save electricity itself, and stop contributing to global warming for no purpose.

Therefore I urge MCark's programmer to make it resign in completely hopeless positions. In future tournaments, when it is completely clear that it (or any other program) has no chance of winning, I shall use my KGS admin powers to end the game and assign a result.

In view of what people have said, I have changed my mind about this.

Power of hardware

Two users, OldGeezer 2d and RayTomes 1d, said that it was a pity that most bots do not state in their info what platform they are running on.

I agree with them. This information is of interest to kibitzers, and is available to operators. I require operators to provide me with this information, and I list it in my reports. But it is also of interest during the event, before my report has appeared.

Therefore, in future, I shall request entrants to add details of processor numbers, power, etc. to their bot's info. If they forget, I shall use my KGS admin power to add it myself.

Annual points

Players receive points for the 2012 Annual KGS Bot Championship as follows:

Crazy Stone4

Details of processor numbers, power, etc.

Aya, running on 6 cores of an i980X, at 3.3GHz.
Crazy Stone, running on 24x2.2GHz AMD Opteron.
Gomorra, running on an Infiniband cluster using 32×12cores running at 2.67 Ghz.
Many Faces of Go, running on an EC2 high memory instance, 4 cores, 8 threads, 40 GB of memory. "In the first few rounds a bug limited memory usage to 3GB, perhaps contributing to its loss against Zen" - DF.
MC_ark, running on a Core-i7 2600K 3.40GHz*4core (8 threads)
oakfoam, running on an : i7-2600K
nomitan, probably running on a machine with 2 Xeon 5680 processors (total of 12 cores).
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.
pachi, running on 4 cores i7-920.
Zen, running on a mini computer cluster consisting of a dual 6-core Xeon X5680@4.2 GHz 24 GB RAM, a 6-core i7 3930K@4.2 GHz 16 GB RAM, a 6-core Xeon W3680@4 GHz12 GB RAM, and a 6-core i7 980X@4 GHz 6 GB RAM PCs connected via a GbELAN. 4 PCs (30 cores) total.