Hundred and twentieth KGS Computer Go Tournament

Sunday February 21st, 2016

These results also appear on an official KGS page.


format40-round Swiss
board size9×9
time4 minutes plus 10/30s


The first round started at 16:00 UTC.

Result table

Zen19 abaku WeakB DolBa AyaMC pachi gnugo matil
W144 B082 W0122 B0182 W0222 B12732 W13128 B13510 J2 B15R W19R B116R W126R B129R J34 B037R W0402 B136 W1103 B11540 W1202 B1253 W1302 W111R B114R W121R B124R W136R B139R B11R W117R B123R W128R B132R B1620 W11346 B133R B17R W119R B138F 33840625Winner
2abakus B044 W182 B1122 W1182 B1222 W02732 B03128 W03510
W13R B111R W01448 B121R W130R B127 W091 B01312 W1238 J32 W1337 B134T B17R W110R B119R W126R B140R W15R B116R W02020 J24 W125R J36 W139R B1124 W1156 B12918 W13816 W16R B117R W128F B137F 30½819½538
3WeakBot J2 W05R B09R W016R B026R W029R J34 W137R B1402 B03R W011R B11448 W021R B030R
B08R W1172 B022R W028R B136T B11R W113R B020R W023R B127R W133R W16R B015R W135R J38 W1718 B1128 W11918 B12426 W13230 B13918 W14R B110R W118R B125R W131F 23½822322½
4DolBaram W036 B0103 W01540 B0202 W0253 B0302 W027 B191 W11312 B0238 J32 B0337 W034T W18R B0172 W122R B128R W036T
W14R B16R W116R B129R W038T W17R J11 W112R J19 W0263 W1522 B11846 W12128 B1316 W037T B14034 B11R W114R B124R W127F B035T W139F 22½792305½
5AyaMC B011R W014R B021R W024R B036R W039R W07R B010R W019R B026R W040R W01R B013R W120R B123R W027R B033R B04R W06R B016R W029R B138T
W12R B18R W118R B130R W031R B137R B1312 W198 B11714 W12516 B12842 W1352 W15R B112R W115R B122R W132F B134F 20787233
6pachi W01R B017R W023R B028R W032R B05R W016R B12020 J24 B025R J36 B039R B06R W115R B035R J38 B07R J11 B012R J19 B1263 B02R W08R B018R W030R B131R W037R
B142 W11015 B11416 W1226 B1274 W1342 B13R W19R B113R W121R B129F W133F B140R 19½772½166½
7gnugo3pt8 W0620 B01346 W033R W0124 B0156 W02918 B03816 B0718 W0128 B01918 W02426 B03230 W03918 B0522 W01846 B02128 W0316 B137T W04034 W0312 B098 W01714 B02516 W02842 B0352 W042 B01015 W01416 B0226 W0274 B0342
W12R B18R W11114 B0162 W0208 B0232 W126T B130F W136F 777046½
8matilda W07R B019R W038F B06R W017R B028F W037F B04R W010R B018R W025R B031F W01R B014R W024R B027F W135T B039F B05R W012R B015R W022R B032F W034F W03R B09R W013R B021R W029F B033F W040R B02R W08R B01114 W1162 B1208 W1232 B026T W030F B036F

Excluding forfeited games, Black won 71 games, White won 69, and there were 8 jigoes.


Seven players registered. To make the numbers even, I included gnugo3pt8.


gnugo3pt8 vs matilda
After the game end: the moves
played in the clean-up phase

In round 26, White, then Black, passed in the position (without the numbered stones) shown to the right. They then disagreed about the status of the black stones at the bottom of the board (they are dead, but even if they live somehow, Black has still won). In the clean-up phase, in which the players should demonstrate that any stones they claim are dead really are dead by capturing them, matilda made one move (103 in the figure), and stopped playing. So gnugo3pt8 won on time.

In round 27 and subsequent rounds, matilda was not connected to the server, and so lost its games by default. I assume it has a bug in its handling of clean-ups, which had caused it to crash.

I considered removing both matilda and gnugo3pt8 from the tournament, as neither of them was able beat any of the six stronger players. This would have given more interesting draws for the strong players. But it might also have affected the result in a way I could not predict, and as the tournament was already two-thirds of the way through, I thought it fairer to leave them in the draw.

abakus vs DolBaram
After the game end

In round 33, the game ended in the position shown to the left. The players disagreed about the status of the single white stone on j8, but in the clean-up phase, both passed again, so the game was scored with that stone counted as alive. This increased abakus's winning margin from two to seven.

DolBaram vs abakus
After move 95

In round 34, DolBaram, as White in the position shown to the right against abakus, stopped moving and lost on time.

In round 35 and subsequent rounds, DolBaram did not make any moves, and so lost its games by default. In round 38 I noticed that it was still connected to the KGS server, so I used my admin power to "kick" it, in the hope of waking it up. This usually causes the kicked bot to reurn five minutes later.

DolBaram did indeed reappear very soon after the start of its round 39 game with matilda. But as matilda was black and still absent, Dolbaram did not get a chance to play a stone.

In round 40, DolBaram played normally against gnugo3pt8, winning their game. Also in round 40, matilda reappeared and joined its game with pachi, which it lost.

Annual points

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



Details of processor numbers, power, etc.

abakus, running on 8 nodes each with two Intel Xeon E5-2670, 2.6GHz and 64 GByte main memory (128 cores total) and each with a Tesla K20 (8 GPUs total).
Aya, running on Xeon W3680 3.3GHz 6 cores + GTX 980, 6 threads (no GPUs).
DolBaram, running on Amazon EC2 C4.8xLarge (2.9GHz * 18core)
GNU Go, version 3.8, running one thread on one i5-5200 CPU.
matilda, running on an i7 3930K @ 3.20GHz with 20GiB RAM @ 1.6GHz.
pachi, running on 32x Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz with 128GB RAM.
HiraBot, running on an i7-4790k 4 cores 8 threads. HiraBot uses its own neural network data.
Zen, running on a dual 10-core Xeon E5-2687W v3@3.1 GHz 32 GB RAM.

themongo: you could also mention where dolabarum lost a won game by faulty cleanup