Hundred and thirtieth KGS Computer Go Tournament

Sunday May 7th, 2017

These results also appear on an official KGS page. (During the tournament, that page did not update and showed no results.)

Rules

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

Time

The first round started at 16:00 UTC.

Results table

PlaceNamecross-tableWinsSOSSoDOSNotes
AyaMC Zen19 Cyber gnugo Julie EasyB
1AyaMC
X
B14F W08R B011R B1119½ W1620½ B112105½ W13R B110R W12 B17T B15197½ W19204½ 106751Winner
2Zen19X W04F B18R W111R
X
W1514½ B1715½ W02F B19R W112R W01F B110F W03F B16T 87049
3CyberGoBot W0119½ B0620½ W012105½ B0514½ W0715½
X
W14R B18R B13 W19F B111F B12321½ W110216½ 77226
4gnugo3pt8 B03R W010R B12F W09R B012R B04R W08R
X
B05R W06R B11261½ W17T B111353½ 47517
5JulieBot B02 W07T B11F W010F W03 B09F W011F W15R B16R
X
B14217½ W08F B012F 47419
6EasyBot W05197½ B09204½ B13F W06T W02321½ B010216½ W01261½ B07T W011353½ W04217½ B18F W112F
X
37416

Excluding forfeited games, Black won 16 games and White won 11.

Players

We welcomed a new player to these events: CyberGo, playing as CyberGoBot, is the work of Austrian programmer Paul Hepperger.

When I checked in the Computer Go room whether there were any last-minute entries, I was asked if EasyBot could enter. I asked how strong it is, and was immediately disconnected from KGS, so I did not see the response (I suffered several unexplained disconnections around that time). When I reconnected, time was short so I gave it the benefit of the doubt and registered it to play. I now regret this: it is very weak (maybe 27k), very reluctant to pass, never resigns, and cannot score correctly so all its games go into the clean-up phase.

Results

In round 1, Zen19X was not connected to KGS. I learned later that its operator had forgotten to launch it before he went to sleep. It therefore lost its first four games on time. He woke up and launched it during round 5.

In round 2, EasyBot was late to join its game with CyberGoBot, because its operator had failed to put mode=tournament in its configuration file, and it was playing a human KGS user when the round started.

In round 5, Zen19X appeared, four minutes late, in its game with CyberGoBot, which it won.

JulieBot vs AyaMC
Moves 306,307.

In its round 7 game with AyaMC, JulieBot played move 306 as shown to the right, putting its own two stones into atari. AyaBot captured with 307. JulieBot presumably tried to recapture at A18, which is illegal according to the positional superko rule as implemented on KGS for games using Chinese rules. Anyway, JulieBot disconnected from KGS, and did not reconnect, losing this and all its subsequent games on time.

Also in round 7, EasyBot was twelve minutes late joining its game with gnugu3pt8. It then passed repeatedly and rather slowly, until gnugo3pt8 had 44 stones on the board and it had none, when it lost on time. Its operator explained that this was the result of a bugfix intended to cause it to pass during the clean-up phase, which in fact caused it to pass in all circumstances. He then fixed this behaviour.

In round 8, and again in round 11, Zen19X played against AyaMC, and won. These two games were AyaMC's only losses, so it finished with ten wins. Zen19X had lost its first four games, including one against AyaMC, on time before it connected to the server; so it finished with eight wins.

Annual points

Players receive points for the 2016 Annual KGS Bot Championship as follows:

Aya8
Zen5
CyberGo3
GNU Go2
Julie1

Details of processor numbers, power, etc.

AyaMC
Aya, running on a W3680 3.3GHz 6 cores with a GTX 980
CyberGoBot
CyberGo, running 1 core 1 thread on Phenom II X4 965 3.4 GHz plus Nvidia GTX 970
EasyBot
unstated
gnugo3pt8
GNU Go, version 3.8, running one thread on one i5-5200 CPU.
JulieBot
Julie, running on a 4.0 GHz Intel Core i7, 4 cores; no gpu
Zen19X
Zen, running 6 threads on a dual 10-core Xeon E5-2687W v3@3.1 GHz 32 GB RAM with an nVidia GTX-1080.