This is not an official KGS page.
Chinese rules, komi 7½
28 minutes each absolute time
Chinese rules, komi 7½
28 minutes each absolute time
Both six-round Swiss
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.
As usual, the tournament was held in two divisions, Formal and Open, with more restrictive entry conditions for the Formal division.
Formal Division 19x19
Open Division 19x19
The "real" names of the bots listed above, and of their programmers, are listed here: programs which have registered for KGS Computer Go Tournaments.
Only two games ended in arguments between the bots. Records these "forfeited" games are now available below.
In round 1, SlugGo had a won position against botnoid (over 170 points ahead), but blundered it away, and finally lost by half a point. There was some dicussion of the result half a point to White (btnoid) is correct under the Chinese rules that are used in these events. SlugGo had passed while botnoid had been playing inside its territory, and Japanese rules would have given the win to Black.
Also in round 1, GoWind had an overwhelmingly won position against PyBot, with ten minutes on its clock while PyBot had only eleven seconds. But after playing inside its own territory and removing dead white stones for a while, it stopped playing and lost on time.
The third round 1 game was one of the two which ended with the players disagreeing on the status of groups. IndigoBot claimed that three of Ayabot's groups were dead, despite two of them each having three rock-solid eyes. But indigoBot had won this game anyway.
In round 2, PyBot resigned its game against botnoid without playing a move. Its operator had accidentally left its resign flag set.
In round 3, indigoBot had an overwhelmingly won position against botnoid, but lost on time while clearing up the dead stones inside its territory.
In round 4, I was looking at the position shown on the right, and wondering if the programs would understand this seki at the end of the game, or if I would have to adjudicate it. Botnoid, playing White, resolved this by playing in the corner, a move that cost it 26 points.
In round 5, I was watching the game between indigoBot and AyaBot, and thinking
"surely Ayabot will hane underneath, why is it taking so long?", when I realised
that AyaBot was not present. After eight minutes, AyaBot's operator reconnected it,
and it played the expected hane. But indigoBot was also missing, and eventually
lost on time. I assume something happened at the server, which disconnected both
players from this game.
An incident like this gives an ethical problem to me, the organiser. On one hand, I would like the games to proceed, and would like to notify the players if I see that their bot needs attention. On the other hand, I ought to be impartial among the players, whhich is difficult when I know the telephone numbers of some and not of others. Also, I do not watch all games with equal attention. My policy will be, to watch all games at least to some extent, and if I happen to notice that action by an operator is required, I will try to contact them by the facilities of KGS but not by other means. I shall use this list of names of KGS bot operators; if anyone wants an operator's name added to that list, they should tell me.
Also in round 5, PyBot got another win on time from a lost position, this time against GoWind.
SlugGo, playing at a faster time setting than it has used previously, was the clear winner, despite its half-point loss to botnoid.
Several of the bots in the Open division showed an inability to read out ladders. In rouind 3, spectators could enjoy watching two ladders being played out simultaneously, in different games.
In round 5, indigoBot2 had a won position against SimpleBot, and lost on time
while clearing up the dead stones.
Some programs are able to clear up dead stones impressively fast, and others aren't. This makes me wonder what will happen when the KGS server is upgraded and the game-end protocol is used. Will clocks be used during the clean-up phase, making it possible to lose on time after the "game end", and giving bots an incentive to be awkward if their opponent is short of time? If clocks are not to be used in the cleanup phase, how much time will I need to allow for it in the tournament schedule?
GNU Go was the undefeated winner.
I am fortunate that none of these KGS events has yet involved an end-of-game dispute which has been difficult to resolve.
Sometimes, when there is an end-of-game dispute, a bot makes a claim which it obviously does not "believe". For instance, maybe a Black group which is agreed to be alive borders a White group whose status is disagreed, and there are unfilled dame between them; or even, unclaimed territory between them. Black's play has been consistent with Black's claim that the white group is dead, but White's play has not been consistent with White's claim that the white group is alive. (I know that White is not really "lying" it is just using different sections of code for move-generation and for game-end claims.)
If something like this happens, in a position whose resolution is not obvious, I shall be inclined to rule in favour of the bot whose claims are consistent with its previous play.
Here is a list of the "forfeited" games. You can download any game listed here by clicking on it. Other game records are available from the KGS pages: Formal Division. Open Division.
|round||Formal division||Open division|