Twenty-seventh KGS Computer Go Tournament

Sunday June 3rd, 2007

These results also appear on official KGS pages: Formal Division, Open Division which link to the game records.


 Formal divisionOpen division
board size19×1919×19
time58 minutes absolute58 minutes absolute


Each division four-round Swiss.


The first round started at 08:00 UCT for the Formal and 08:10 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   19×19


Open Division   19×19


The "real" names of the bots listed above, and of their programmers, are listed here: programs which have registered for KGS Computer Go Tournaments.


The UCT algorithm, which has proved so effective on smaller boards since its publication last September, is less effective on full-sized 19x19 boards. Nevertheless, in the March KGS bot tournament, the last one to be held on 19x19 boards, the UCT-based MoGo won both divisions without loss, beating the non-UCT programs GNU Go, Many Faces of Go, and Aya. That event used fast (18 minutes) time limits, which are said to be a disadvantage to MC and UCT programs: a UCT program can productively use all the time you give it to search the game tree more thoroughly, while a conventional program like GNU Go does not really know what to do with time after some point. So I had anticipated MoGo, or StoneCrazy which also uses UCT, winning this event, which used slow time limits of 58 minutes each.

I was wrong. GNU Go won the formal division, beating CrazyStone. Unfortunately it was never drawn against MoGo. In the Open division, a modified version of GNU Go (MonteGNU, which includes some MC code) beat MoGo, and lost against StoneCrazy, and all three programs ended on three wins.

Formal division

In round 1 FirstGoBot passed early in its game against MoGoBot. This indicates some kind of bug, which we hope can be found and removed.

Open division

Before the event started, I removed IdiotBot, to make the numbers even and avoid byes.

In round 3, when MoGoBot19 played the circled move K18 (see diagram at right) against StoneCrazy, StoneCrazy failed to notice that it could save the eight triangled stones by capturing five white ones, and instead played at L10. Had it made the F18 capture instead it would probably have won the game.
      StoneCrazy's auther Rémi Coulom commented at the time "The capture at F18 was pruned... Should not be too difficult to fix." However he later wrote to me, "In fact, the capture at F18 was not pruned. But White extension at F18 after Black L10 was. The reason is that F18 is a Ko shape."
      So there is still room for significant improvement in this UCT program. And despite its making mistakes like this, it has (playing as CrazyStone) a rating of 2k on KGS, and has had a 1k rating.

Also in round 3, Orego2 was winning a close game against SimpleBot when it vanished, and eventually lost on time to SimpleBot.

In round 4 Orego2 was still missing, so its opponent StoneCrazy won by default.