Eighty-seventh KGS Computer Go Tournament

Sunday November 4th, 2012

These results also appear on an official KGS page.


format12-round Swiss
board size19×19
time14 minutes plus 10/30s


The first round started at 08:00 UTC.

Result table

AyaMC ManyF stv pachi nomiB gnugo Orego MCark
B02R W07R B112R W13R B111R W14R B19R W11R B110R W15 W16R W18R 107759Winner
2ManyFaces1 W12R B17R W012R
B15R W09R W11R B010R B16R B1315½ W111 B18R B14R 97853
3stv B03R W011R W05R B19R
W17R B18R B14R B02T W110½ B11R W112R B16R 87340
4pachi B04R W09R B01R W110R B07R W08R
W13R B112R W16½ W15R B111R B12R 77732
5nomiBot2 B01R W010R W06R W04R B03R W012R
W17½ B18R B12R W19R W15R B111R 66716
6gnugo3pt8 B05 W0315½ B011 W12T B010½ B06½ B07½ W08R
W04 W11R B19R W112R 46911
7Orego12 B06R W08R W01R B012R B05R W011R W02R B09R B14
B03R W17R B110R 3686
8MCark B08R W04R W06R W02R B05R W011R B01R W09R B012R W13R B07R W010R

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.

Seven players registered for the tournament. To make the numbers even, I added 'gnugo3pt8', a build of GNU Go running on one processor of my own Windows system.


gnugo3pt8 vs MCark
Moves 86-124

In its round 1 game with gnugo3pt8, MCark misread a ladder against gnugo3pt8, as shown to the right. Initially it was a ladder that could be forced around a corner; then it became a ladder that coiuld be diverted to take advantage of a ladder-breaker. Once the ladder-breaker was reached, MCark realised its stones could not be saved, and played elsewhere.

At the start of round 2, stv failed to connect to the server due to a problem with DNS lookup. When its operator Erik van der Werf saw what had happened, he started another instance on an old machine running rather old versions of Steenvreter and kgsGtp. It joined its game with 12 seconds of main time remaining. He hoped it would play and win the entire game against gnugo in byo yomi. But additional netlag made this impossible and it lost on time.

In round 3 the crippled version of Steenvreter started to play AyaMC. But during the game the original machine came back online, kicking the crippled version; which in turn came back, kicking the good version. This continued until Erik noticed and killed the crippled version, but by then AyaMC had a convincing lead.

After round 9, the results table was pleasingly symmetrical

ManyFaces1 [2d]8
AyaMC [4d]7
stv [2d]6
nomiBot2 [2d]5
pachi [2d]4
gnugo3pt8 [6k]3
Orego12 [-]2
MCark [-]1

and it seemed likely that ManyFaces1 would win the tournament.

In round 10, ManyFaces1 lost its first game, to pachi.

In round 12, ManyFaces1 and AyaMC were leading, with nine wins each. They played each other, and AyaMC won.


Every game not involving gnugo3pt8 ended with a resignation.

For me, this is a big improvement on the days when games often ended with a scoring disagreement, so I might have to count the score myself.

Annual points

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

Many Faces of Go5

Details of processor numbers, power, etc.

Aya, running all 12 threads on an i7-980X 3.3GHz 6cores +HT.
GNU Go, running on one Intel i5-2500 CPU @ 3.30GHz.
Many Faces of Go, running on an Amazon EC2 Cluster Compute Eight Extra Large Instance, 2 x Intel Xeon E5-2670, using 16 cores, 16 threads, and 48 GB of memory.
MC_ark version 6.0, running on Core-i7 3770 3.40GHz*4core (8 threads)
nomitan, running on a machine with 2 Xon E5-2680 processors (total of 16 cores @ 2.7Ghz ).
Orego, running on one node of our custom Linux cluster Fido. The node has two AMD Six Core Opteron 2427 2.2 GHz processors (12 cores total), 8 GB RAM, Centos Linux
pachi, running on an Intel i7-2600 (4 cores, 8 threads) with 16GiB RAM.
Steenvreter, probably running 46 threads each at 2.2 GHz, on a system whose use was generously provided by the Maastricht games and AI group.