Autumn 2013 Slow KGS Computer Go Tournament

September 8th - 10th, 2013

These results also appear on an official KGS page.

Rules

format6-round Swiss
board size19×19
rulesChinese
komi
timeThree hours each, sudden death

Times

The first round started at 22:00 UTC on September 8th

Results table


PlaceNamecross-tableWinsSOSSoDOSNotes
Zen19 AyaMC pachi NiceG gnugo Orego
1Zen19S
X
W12R B16R W11R B13R W14R W15R 61616Winner
2AyaMC B02R W06R
X
W13R B15R B11R W14R 4208
3pachi B01R B03R
X
W14R B16R W15R W12R 4166
4NiceGo19N W03R W05R B04R W06R
X
W12R B11R 2202
5gnugo3pt8 B04R W01R B05R B02R
X
W13R B16R 2160
6Orego12 B05R B04R B02R W01R B03R W06R
X
0200

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.

Black won 7 games and White won 11.

Five players registered for the tournament. I therefore added GNU Go, for the same reasons as in the August tournament.

The format, Swiss with six players and five rounds, was unfortunate. The players played each opponent once (producing entirely consistent results, with no intransitivities); then there was a final round, in which the second-best player, AyaMC, played its second game against the best player, Zen19S. This resulting in the third-best player, pachi, achieving as many wins as AyaMC, and (as it worked out) as many "annual points". To avoid such outcomes, a Swiss tournament with an even number of players should not have the same number of rounds and players (and one with an odd number of players should not have one more rounds than players).

Howver, I specify the formats of these events in advance, and only learn the number of players shortly before play starts. So I cannot avoid such outcomes.

Results

In round 5, the game between Zen19S and AyaMC attracted 50 kibitzers at one time. Some of them said complimentary things about the level of play.

Comments

When I announced this tournament, I said that it might be the last Slow tournament on KGS. I thought there was little purpose in having slow tournaments, now that it is easy to hire multiple processors in the cloud.

However three people emailed me, urging me to continue running them. I had failed to appreciate the differences between designing a program to run on a cluster, and to run for longer on a single system. I will continue to run Slow bot tournaments on KGS, probably twice a year. I will schedule them to run for longer than the 36 hours that this one ran for.


Annual points

Players receive points for the 2013 Annual KGS Bot Championship as follows:

Zen8
Aya4
pachi4
oakfoam2
GNU Go1

Details of processor numbers, power, etc.

AyaMC
Aya, running on 6 cores of an i7-980X 3.3GHz 6cores +HT.
gnugo3pt8
GNU Go version 3.8, running at its "level 21" on a single 3.3GHz Intel i5-2500 CPU. For rounds 1 and 2 it was configured to level 19, so that it played too fast.
NiceGo19N
oakfoam, running on a mini-cluster: i7-2600K and i7-920
Orego12
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
pachi running on 2x Opteron 6134 (15 threads) with 64GiB RAM.
Zen19S
Zen, running on a mini cluster of a dual 6-core Xeon X5680@4 GHz 24 GB RAM, a 6-core i7 3930K@4 GHz 16 GB RAM, a 6-core Xeon W3680@4 GHz 12 GB RAM, and a 6-core i7 980X@4 GHz 10 GB RAM computers connected via a GbE LAN. 30 cores total.