Formal division | Open division | |
---|---|---|
format | 6-round Swiss | 8-round Swiss |
board size | 13×13 | 9×9 |
rules | Chinese | Chinese |
komi | 7½ | 7½ |
time | 18 minutes absolute | 13 minutes absolute |
The first round started at 08:00 UCT for the Formal and 08:05 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 13×13
|
Open Division 9×9
|
The "real" names of the bots listed above, and of their programmers, are listed here: programs which have registered for KGS Computer Go Tournaments.
YangYun emailed to enter two versions of his program 'break': 'break9' to play in the
Open division, and 'break13' to play in the Formal. However I did not read his email
carefully, and only registered 'break' for the Formal. Then I read it properly, but
found that 'break9' and 'break13' do not have KGS accounts. I informed him of this,
and said that I would register them once they existed.
However, when the tournament was due to start, 'break9' and 'break13' still did not
have accounts, and 'break' was playing a game against a human. I left it in the draw
(for the Formal division), but it never made a move in a tournament game.
I never heard from CrazyStone or MoGo. They are both French, and most French people go on holiday for August.
A few minutes before the tournament was due to start, no version of GNU Go had been registered, so 'woll997' registered his 'wollbot', a straight build of GNU Go 3.7.2, for the Open division. Then Gunnar appeared at the last moment and registered GNU and MonteGNU. We decided that Monte GNU was sufficiently different from straight GNU Go that it and wollbot could both play in the Open division.
IdiotBot was removed from the draw to make the number in the Open division even.
In round 1, break rapidly joined and left its game with FirstGoBot, three times. This was its only involvement in the tournament, it never made a move, and lost all its games by timing out.
So there were only three active participants, and break, which I did not remove as I might as well leave it in the draw as have byes. So each participant played each of the others twice. GNU won all its games, break lost all its games, and FirstGoBot beat AyaBot twice.
HBotSVN joined all its games, but timed out in all of them without playing a single stone. After six rounds of this, its owner Jason House thought he had identified a problem with its time settings, and corrected it, but it continued to time out.
He subsequently explained to me:
HouseBot is a natively multi-threaded application. Work threads ponder continuously as the game goes on. A separate control thread interrupts them at key times in order to respond to genmove and similar commands. Rather than respond immediately, HouseBot's control thread will sleep for a period of time based on the time remaining.Since the previous tournament, HouseBot underwent a port to a new set of D core libraries (Tango instead of Phobos). While the coding of sleeping in Tango was greatly simplified, it appears that prolonged sleeping (> 1 sec?) can actually wait far longer than desired on certain platforms. Development of new features occurred under windows and everything worked well. For the tournament, a Linux machine was used instead.
The issue has been resolved and HBotSVN will do better in next month's tournament.
In round 4 there was an interesting game between agog (goanna) and MonteGNU (GNU Go with some MC enhancements). Agog won this, in the usual MC fashion, by half a point. This was MonteGNU's only loss. Goanna is a new program, and MontGNU is one of the strongest bots at 9x9 Go, so this was an impressive win.
In round 6 agog lost its only game, to ggmcbot.
The tournament ended with MonteGNU and agog each on seven wins from eight games. Agog had played every opponent except dpthought, while MonteGNU had played every opponent except the sleeping HBotSVN which had lost all its games. So MonteGNU won the division on SOS.