This is not an official KGS page.

Second KGS Computer Go Tournament

Sunday May 8th, 2005

Rules

9x9 board
Chinese rules, komi 7½
Twenty-eight minutes each absolute time

Format

Round Robin

Times

The first round started at 08:00 GMT, subsequent rounds started at one-hour intervals. The result was announced soon before 19:00 GMT.

Results

The winner was AyaBot, a version of Aya, written by Hiroshi Yamashita. It lost only one game, on time, to SimpleBot. It was then replaced by an older and faster version (which did not know how to score a game, with consequences described below), which won all its subsequent games, including one against GNU, the version of GNU Go which was playing.

Comments

Congratulations to Ayabot, which played in its first KGS tournament and won convincingly.

Again the tournament was successful, with a record number of participants. It showed up what appears to be a serious bug in the KGS tournament mechanism, as described below.

In round one, LeGoBot ended the game with a dead group. But LeGoBot refused to admit it was dead. Therefore the Tournament Director adjudicated this game as a loss for LeGoBot. This result was initially acknowledged by the tournament system, but soon spontaneously "flipped" to a win for LeGoBot.

In round two, Neuron ended the game with a dead group. But Neuron refused to admit it was dead. Therefore the Tournament Director adjudicated this game as a loss for Neuron. This result appeared on the results page as a win for Neuron.

In the first two rounds, AyaBot did not reach the end of a game. It had a bye, then lost a game on time. It was then replaced by an old version of Ayabot, which did not understand the game result agreement procedure. In round three, Ayabot was ahaead at the game end, but as it could not perform the agreement procedure, the Tournament Director assigned it a win. This result was initially acknowledged by the tournament system, but soon spontaneously "flipped" to a loss for AyaBot.

In round four and all subsequent rounds, AyaBot's owner disconnected it at the end of each game and manually did the game end procedure. In every case, this was initially reported (correctly) as a win for AyaBot, but then appeared on the web page as a loss for AyaBot.

Some recommendations and suggestions, in order of priority:

• To wms: fix the bug that "flips" results.
• To the authors of LeGoBot, Neuron, and AyaBot: fix your bots so that they handle the game end procedure correctly.
• To all bot programmers: arrange that your bot actually removes all its opponent's dead stones before passing. I am not going to insist on this, but it does help when your opponent won't admit that its one-liberty group is dead. We are using Chinese scoring, so it costs nothing.
• To wms: there was a bug which caused a "round 12" to appear on and disappear from the schedule. I think it appeared shortly before each round started.
• To wms: I would like to see a half-point awarded for a bye. This obviously does not affect the result in a round-robin, but it does make it easier for the onlookers to understand how well each bot is doing, part way through.
• To wms: I would like there to be a final tie-break method, I suggest SoDOS if this is easy to implement.
• To wms: a game whose result has been determined by the Tournament Director is listed as "forfeited". This suggests that the loser has been penalised for doing something wrong. I wopuld prefer "adjudicated".