Twenty-fifth KGS Computer Go Tournament

Sunday April 8th, 2007

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


 Formal divisionOpen division
board size9×913×13
time13 minutes absolute28 minutes absolute


Formal division: eight-round Swiss.
Open division: four-round Swiss.


The first round started at 09:00 UCT for the Formal and 09:10 for the Open division.


As usual, the tournament was held in two divisions, Formal and Open, with more restrictive entry conditions for the Formal division.

Formal Division   9×9


Open Division   13×13


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


There were two newcomers to this event. Dpthought, by Cai Qiang, has been playing on KGS since January 2006, and is rated 6k there, but this was its first tournament. Jiango is a MC program by Peigang Zhang, and only registered at KGS this February.

hb05 has played previously from the account HBotSVN, and is version 0.5 of Jason House's HouseBot.

Formal division

Dpthought finished at the bottom of the table, with two wins. But with slightly better time management, it would have scored four wins: in round three against jiango, and again in round five against GoComputer, it ran out of time in the very late endgame, when its opponents' groups had all been reduced to one liberty.

CrazyStone has achieved an implausible 1k rating on KGS. In its round two game against MoGoBot, it did badly early in the game, and then applied the UCT principle that if all moves lead to a loss you may as well play at random. By the end of the game it had lost all its stones.
      They were drawn together again in round eight, and this time CrazyStone won after a messy-looking game. So MoGoBot and CrazyStone each ended with seven wins, and MoGoBot was declared winner on SOS.

It seems odd to have an eight-round tournament with only seven players. A seven-round round-robin, with each player receiving one bye, would be better. But the KGS automated tournament system requires the format of the tournament, and the number of rounds if it is Swiss, to be specified in advance, before the number of players is known.

I found the round five game between MoGoBot and MonteGNU remarkable, for looking like a game between human single-digit-kyu players. MonteGNU's move 11 was poor, d3 would have been better. However MonteGNU managed to play the important e2 first, leaving MoGoBot with a dead corner and little compensation. But when MoGoBot played a4 in the hope of swindling its dead corner back to life, MonteGNU failed to see the danger, and tenukied, losing the game as a result.
      I used, years ago, to praise HandTalk as the only program that actively played swindles – sequences that a dan-player might play, knowing that they weren't good Go but could work against a kyu-rated opponent. I don't know how HandTalk managed this, it can't really have been doing it deliberately, but it sometimes looked that way. It's easier to see how the same effect can be achieved by a MC program.

Open division

In round three, StoneCrazy surprised the onlookers by beating MoGoBot13. CrazyStone/StoneCrazy was playing noticeably better than in previous KGS tournaments, and its author Rémi Coulom agreed that he has made a recent improvement. He will reveal this after the ICGA Tournament in Amsterdam this June, which he will be competing in.