This is not an official KGS page.

Test KGS Computer Go Tournament

Sunday March 6th, 2005

This tournament was held on KGS as a test, preparatory to holding further more serious Computer Go tournaments on KGS. The results are also given on the KGS Tournament Result page.

Rules

9x9 board
Chinese rules, komi 7½
Ten minutes each absolute time

Format

Four-round Swiss

Times

The first round started at 20:30 GMT, subsequent rounds started at 30 minute intervals. The result was announced at 22:20 GMT.

Results

Program	GoGNU	Viking5	botnoid	tlsBot	GoFigure	nio	DumbBot	byes	total	SOS	SoDOS	place	Programmer
GoGNU		1₁ 1₄	1₂	1₃					4	8	8	1st	GNU Go development team
Viking 5	0₁ 0₄		1₃			1₂			2	11	3	2nd	Magnus Persson
botnoid	0₂	0₃				1₄		1₁	2	7	1	3rd	Don Dailey
tlsBot	0₃				0₂	1₁		1₄	2	7	1	4th	Michał Bażyński
GoFigure				1₂			1₁		2	3	3	5th	Evan Daniel
nio		0₂	0₄	0₁			1₃		1	7	1	6th	Christian Nilsson
DumbBot					0₁	0₃		1₂	1	3	0	7th	John Davies

In a program's row, 1 means a win, 0 means a loss. Subscripts indicate the number of the round. The 1s and 0s link to the SGF game records.

The undefeated winner was GoGNU, version 3.7 of GNU Go, written by the GNU Go developers.

DumbBot crashed at the end of round 3, and its operator failed to restart it in time for round 4.
GoFigure quit before round 3, with an unexplained disconnect.
Go81, by Tapani Raiko, also registered, but arrived late because of a network failure.
FStoned, by Chris Fant, also registered, but had a problem with its script and failed to connect.

Comments

Congratulations to GNU Go (playing as 'GoGNU'), the undefeated winner. GNU Go has two advantages in events like this: its many programmers have co-operated to produce a strong engine; and its many volunteer helpers allow it to compete in any event in any timezone.

The event was successful, and was enjoyed by all the participants.

Some learned, from various disconnects and crashes, how to make their systems more robust for future tournaments.

The two strongest players, GoGNU and Viking5, were paired against each other twice. This is not the normal way of running a Swiss event, it is usual to try to avoid repeat pairings. If GoGNU and Viking5 had each won one of these two games, while they had both won all their other games, the winner would have been decided, arbitrarily, by tie-break. This is probably not a good thing. It is likely that the KGS tournament engine will be changed in future, to give a higher priority to avoiding repeat pairings in Swiss events.