Seventy-fourth KGS Computer Go Tournament

Sunday August 14th, 2011

These results also appear on an official KGS page.


format30-round Swiss
board size9×9
time4 minutes plus 10/30s


The first round started at 08:00 UTC.

Result table

pachi2 Zen9 Fuego AyaMC oakfoam Orego12
WJ2 B17R WJ9 B013R W015R BJ19 W1202 B027R W1302 W11R B112R W114R BJ18 W122R B024R W128R B129R W13R B08R W117R B123R W14R B110R W116R B125R B15R W111R B121R W126R 22Winner
2Zen9 BJ2 W07R BJ9 W113R B115R WJ19 B0202 W127R B0302
W05R B18R W116R B023R W14R B110R W114R B121R W0252 B126R W13R B111R W117R B122R W128R W16R B112R W118R B124R W129R 21½
3Fuego B01R W012R B014R WJ18 B022R W124R B028R W029R B15R W08R B016R W123R
W07R B111R W113R B019R W127R W16R B09R W115R B121R W126R W13R B110R W117R B120R W125R B130R 16½
4AyaMC B03R W18R B017R W023R B04R W010R B014R W021R B1252 W026R B17R W011R B013R W119R B027R
W11R B012R W118R B120R W124R B029R WJ30 B12R W09R B115R W116R B122R W028R 12½
5oakfoam B04R W010R B016R W025R B03R W011R B017R W022R B028R B06R W19R B015R W021R B026R B01R W112R B018R W020R B024R W129R BJ30
W17R B08R W013R B114R W119R B123R W127R
6Orego12 W05R B011R W021R B026R B06R W012R B018R W024R B029R B03R W010R B017R W020R B025R W030R W02R B19R W015R B016R W022R B128R B07R W18R B113R W014R B019R W023R B027R

In the table above,
   0 indicates a loss
   1 indicates a win
   J indicates jigo
   a superscript indicates the round in which a 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.

The numbers in the KGS table do not add up as you might expect. This is because they include points for wins against the inactive StoneGrid, and points for byes.

Seven players entered. These included StoneGrid, which suffered from a problem and failed to make any moves. After six rounds I removed it, leaving the six players shown in the cross-table above.


In round 1, StoneGrid and Orego12 played a reasonable game for 24 moves, then StoneGrid stopped moving, and lost on time. Its author John Fan later emailed me "Sorry the deadlock occurred again in StoneGrid. I think I found a suspicious spot. I will let it run on cgos for sometime to see if it is still occurring.".

pachi2vs Fuego
First 17 moves

In rounds 1, 14, 22 and 28, pachi2 was White against Fuego, and the games all started with the same 16 moves, shown to the right. The sequence looks to me bad for White, who has lost two stones for no clear benefit. But White went on to win all four of these games, so I must be wrong.

pachi2 vs Fuego:     Continuations
After move 17 the games diverged. Their continuations are shown to the left.

In round 5, after StoneGrid had lost on time, I used my admin power to kick it from the server, in the hope that this might wake it up.

In round 6 StoneGrid reappeared five minutes after being kicked (I have observed that this five-minute delay is standard, and is somehow a feature of the kgsGtp client). However it still did not make any move. So I removed it from the tournament after it had once more lost on time.

pachi2vs Zen9
After both players passed

In round 9, pachi2 (White) and then Zen9 (Black) passed in the position shown to the right. They then failed to agree on the status of the two dead white stones, with Zen9 claiming that it could kill them and pachi2 claiming that they were alive.
    What should have happened now is for the game to enter the "clean-up phase": play resumes, and continues until there are two more consecutive passes, when all stones remaining on the board are counted as alive.
    What did happen is that the server announced the start of the clean-up phase, and told White to move. White passed. The server then stated "Cleanup mode has ended by passes. It will be assumed that all dead stones have already been removed.", and never gave Black a chance to show that it could capture the two stones. The game was therefore counted as a 12-point win for White. It should have been scored as a jigo, or tie.
    The rest of the pairings were made by the KGS tournament scheduler in the belief that White had won this game, and in KGS's report on the tournament, the game is shown as a win for White and the SOS and SoDOS are calculated accordingly. In this report, the game is shown as a jigo, thus WJ9, and the numbers of wins have been recalculated accordingly. I have not bothered to recalculate the SOS and SoDOS, as they are not needed to break any tie.
    The KGS progammer 'wms' has acknowledged the problem, and will fix it when he next can.

pachi2 vs Fuego
After Black's move 65

In round 14, Fuego played move 65 as shown to the left. This move achieves nothing. It went on to lose the game by resignation (when seven points behind) after its opponent, pachi2, converted its lower left territory to a seki. If instead it had played at 65 at a2 so as to make definite territory there, it would have won the game by two points.

oakfoam vs Orego12
Moves 1-6

In round 19, oakfoam (White) and Orego12 started their game as shown to the right. Move 5 looks to me like a blunder, as it invites the very effective move 6. Indeed, oakfoam went on to win this game. The round 7 game between oakfoam and Orego12 had started with the same six moves, and was also won by oakfoam.

Fuego vs pachi2
After Black's move 49

In round 24, pachi2, as Black, was doing badly against Fuego. In the position shown to the left, it appears to have two options – resign, or hope to win the approach-ko in the top right. Instead, it connected the ko, as shown. By removing the ko, it made its position completely hopeless, in gote.
    But I can say that because I recognise that after capturing the six stones, White will have two eyes in the top right. My guess is that pachi2 was looking at sequences in which White captures, Black throws in, and White, by responding randomly, fails to make two eyes, and loses the whole group.

In round 30, pachi2 and Zen9 both passed in a position in which pachi2 was two points ahead. As in round 8, they then disgreed about the status of some dead stones; and as in round 8, Zen9 never had an opportunity to show that it could kill them, so the game was counted with them alive, giving pachi2 a win by 31 points instead of by two. This time the error made no difference, but it has been corrected in the cross-table above.


85 games were actually played (rather than being byes, or walkovers against an absent opponent) in this event, using komi of 7. White won 47, Black won 34, and there were 4 jigoes.

Annual points

Players receive points for the 2011 Annual KGS Bot Championship as follows:



Details of processor numbers, power, etc.

Aya, running on 6 cores of an i980X 3.3GHz
Fuego, running on 2 x Intel Xeon X5670 Six-Core 2.93GHz (12 cores total, 24 threads)
Oakfoam, running on a quad-core Intel Xeon E5520 with hyperthreading (so 8 threads).
Orego, probably running on one of the five nodes of a custom Linux cluster built 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 64 platforms, each x86 64 bits, 32 GB ram, using 22 cores of each, giving about 1500 playouts/s/core at the beginning of a 19x19 game.
Zen, running on a Mac Pro 8 core, Xeon 2.26GHz