This is not an official KGS page.
19x19 board
Chinese rules, komi 7½
58 minutes each absolute time
Round Robin
The first round started at 17:00 GMT for the Formal and 17:10 for the Open division, subsequent rounds started at one-hour 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
|
Open Division
|
The "real" names of the bots listed above, and of their programmers, are listed here: programs which have registered for KGS Computer Go Tournaments.
I hade hoped to make this event welcoming to commercial programs, such as Go++ and Many Faces of Go. The separation into two divisions, and the long time limits, are both things that commercial programers have requested. Go++ did indeed attempt to play; it was left running on Mick Reiss's computer, but unfortunately a carpet-layer who was working in his apartment unplugged the computer, despite explicit instructions not to.
We were able to welcome Bruno Bouzy's program Indigo, a newcomer to these events. It chose to play in the Open division because it was not confident that it would handle the protocol correctly. We hope that it will play in the Formal division in future (if the two divisions are to be retained).
With only four entrants, one of them not actually showing up, there was only one game per round.
In round 1, Dariush beat antbot. Antbot cannot handle the game end, and always claims that all its dead groups are alive, so this game is shown as forfeited.
In round 2, Dariush was drawn against slugGo. SlugGo played worse, and faster, than might be expected in the opening – this was later found to be because it had been set up wrong. This led to a close game.
David Doshay explains: "The problem seems to be that SlugGo did not properly clean up and reset its parameters after the first game in preparation for the second. The 58 minute wait for white's move [while it waited for gopp] resulted in the dropping of SlugGo's playing level, and that state information was carried into game 2. After a number of moves SlugGo calculated that it was enough "ahead of the curve" on time, and its playing level went back up.
In a very close endgame, after all the normal dame had been filled, Dariush filled the last dame by extended into a "Sending two, returning one" position. SlugGo naturally recaptured (and happened to give atari to a group of Dariush's). If Dariush had now connected its ataried group, allowing slugGo to make an eye where the two "sent" stones had been, slugGo would have won by 1½ points (I miscounted this at the time, and made some mistaken comments which appear in the game record). If Dariush had been allowe to recapture, Dariush would have won by 2½ points. In fact, Dariush insisted on trying to recapture, but the recapture was forbidden by the "positional superko" rule, so Dariush failed to move at all and eventually lost on time. Here is the Dariush-slugGo game record.
In round 3, slugGo beat antbot. Again antbot's game is shown as forfeited.
Round 4 was particularly disappointing. Antbot was drawn, for the second time, against the absent gopp. SlugGo was looking forward to a rematch against Dariush, but Dariush had vanished – perhaps its operator thought that after each program had played each opponent, the tournament was over. So no actual Go was played in this round.
GNU was a convincing winner, playing the four strongest of its opponents and beating them all convincingly. IndigoBot and Dar51 tied for second place, losing to GNU and winning all their other games; unfortunately they did not get to play each other.