This is not an official KGS page.

Fourth KGS Computer Go Tournament

Sunday July 10th, 2005

These results also appear on official KGS pages: Formal Division, Open Division which link to all the game records.


9x9 board
Chinese rules, komi 7½
Fifteen minutes each absolute time


Round Robin


The first round started at 16:00 GMT, subsequent rounds started at 35-minute intervals.


The tournament was again held in two divisions, Formal and Open, with more restrictive entry conditions for the Formal division. Essentially, an entrant to the Formal division must not contain move-generating code used by any other such entrant, and its author's real identity must be known. A program may enter both divisions simultaneously (though it will have to use different names for each).

Formal Division

1stAyaBot5 wins
2ndbotnoid4 wins
3rdgnugo3pt63 wins
4thtlsBot2 wins
4thNeuron1 win
4thgonzoBot0 wins

Open Division

1sttlsBotExp16 wins
2ndviking55 wins
3rdNeuronC4 wins
4thgonzoBot23 wins
5thSimpleBot2 wins
6thGo811 win
7thIdiotBot0 wins

The winner of the Formal division was AyaBot, Hiroshi Yamashita's Aya.

The winner of the Open division was tlsBotExp1, by Michał Bażyński and Karol Golab.

The "real" names of the bots listed above, and of their programmers, are listed here: programs which have registered for KGS Computer Go Tournaments.

Description, Formal division

Round 1 produced two close games. Botnoid's central strategy defeated Neuron's staregy of making small groups with territory in them, while AyaBot was happy to take more than half the board without fighting, against gnugo3pt6. (No recent version of GNU Go had been entered for either division, so the tournament organiser had entered his own out-of-date build of it, with permission.) GonzoBot did not show up in time for the draw for the first round (which is two minutes before the announced start time), so it missed its first-round game, to the annoyance of its New Zealand author Peter Mackenzie, who had got up at 4 a.m. to watch it.

In round 2, gnugo3pt6 beat tlsBot, which showed little interest in territory. Botnoid's central strategy was effective against gonzoBot, which made two small live groups and one dead one; but at the end of the game, while they agreed on the status of all the groups, and even on the score, they failed to agree on who had won; so the tournament organiser had to assign the win. AyaBot beat Neuron, in a game that appears not to have been recorded properly.

In round 3, the eventual winner, AyaBot, was drawn against the eventually last-placed gonzoBot. But gonzoBot came up with some intelligent moves, and had some chances of winning. Neuron had some chances against gnugo3pt6, but squandered them with tenukis. The botnoid-tlsBot involved fighting right from the start, and botnoid got the best of this.

In round 4, AyaBot made territory while its opponent tlsBot didn't, and won accordingly. Neuron beat gonzoBot, and as usual, gonzoBot accepted the score but refused to acknowledge that it had lost. However the tournament organiser did not need to assign the result this time - gonzoBot's owner disconnected it, logged in using its account, and resigned for it. (This did not work quite as expected, though no-one realised this at the time.) Botnoid, rated on KGS as 21k, beat gnugo3pt6 which is rated as 13k; this despite gnugo3pt6 making no blunders.

In round 5, AyaBot outplayed botnoid from the start, killing all its stones. Gnugo3pt6 did the same to gonzoBot, which as usual refused to acknowledge that it had lost. The tlsBot-Neuron game might have been close, but Neuron allowed a corner to die.

With six entrants, there should not have been a round 6. But gonzoBot has missed its first round game, causing tlsBot to have a bye there; so the automated tournament scheduler arranged a sixth round, with just these two bots playing in it. TlsBot won comfortably.

Description, Open division

In round 1, Go81 killed all of IdiotBot's stones. IdiotBot plays badly, but as tournament organiser, I appreciate the way that it behaves, never causing any trouble at the end of the game. It never claims its opponents' stones are dead until it has removed them from the board, which helps. Viking5 defeated SimpleBot, despite letting a totally dead group come back to life - I think this happens because viking5 thinks "this move is good enough to win, and therefore as good as any other". NeuronC (a version of Neuron) defeated gonzoBot2 (a version of Gonzo) in a game which, like all Gonzo's games, had a problem at the end, and this one has not been recorded properly. TlsBotExp1, another version of tlsBot, had a bye. Botnoidx, another version of botnoid, was registered to play in this division, but it connected using botnoid's account, so it never got to play a game, but sometimes announced its presence during botnoid's games, causing amusement.

In round 2, tlsBotExp1 achieved a won position against viking5, which was polite enough to recognise this, and resign. NeuronC beat SimpleBot in a close game. GonzoBot2 had no trouble in killing all of IdiotBot's stones, in a game that lasted 158 moves.

In round 3 gonzoBot2 had an easy win against SimpleBot. NeuronC killed all of Go81's stones. TlsBot1 had an easy win against IdiotBot.

In round 4 tlsBot1 began its game with Go81 by building a wall down the middle of the board, and then killed all the white stones to the left of this and some of those to the right. Viking5 had no trouble beating gonzoBot2, which had its usual game end problem. SimpleBot showed that even it can kill all of IdiotBot's stones.

In round 5 it was NeuronC's turn to kill all IdiotBot's stones. TlsBot1's central wall strategy was effective against SimpleBot. Viking5 used some understanding of nakade to defeat Go81.

In round 6 SimpleBot's simple strategy of capturing stones so as to make eyes was effective against Go81. TlsBotExp1's central strategy was effective against gonzoBot1. Viking5 won a close game against NeuronC.

In round 7 Viking5 could easily have killed all of IdiotBot's stones. But its belief that "if any move is good enough to win, you may as well pass or play at random" produced some amusing effects, in a game lasting 226 moves. GonzoBot1 killed all of Go81's stones, but as usual it had a problem at the game end. To help resolve this, its operator killed it and connected using its account. This was detected by the automated tournament system, which asssumed that he was cheating, and assigned him a loss. Normally (I think) the human tournament organiser would have been able to override this; but as (I think) this was the final round, he found that he did not have this power, so this game is still shown on the KGS results page as a win for Go81, whereas the table above shows it as a win for gonzoBot1. TlsBot1's central strategy was less effective against NeuronC than against most opponents, but was still good enough for a close win.


Some technical problems are mentioned below, mainly to remind the persons responsible to fix them.

If you are entering bots for both tournaments at once, it can be confusing. You must get all the following right:
      Have a KGS account for each
      Tell the organiser which one is entering which division
      Arrange for each to log in to its correct account

For your bot to be included in a round, it must be present two minutes before the round starts.

For some reason that is unclear to me, I (as organiser) cannot use the tournEditor program to change the Go81 - gonzoBot2 result in the final round of the Open division.

Gonzo has a problem with acknowledging the game result, even when it has agreed the score.

Don't log in using your bot's account during a tournament. If you do, the system will assume you are cheating and assign it a loss. If you need a result assigned in a game, ask the tournament organiser.