gnugo-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] gnugo


From: Gunnar Farnebäck
Subject: Re: [gnugo-devel] gnugo
Date: Thu, 10 Jun 2004 16:42:25 +0200
User-agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/21.3 (sparc-sun-solaris2.9) MULE/5.0 (SAKAKI)

Johanna Tuominen wrote:
> I'm running gnugo bot in KGS (brainless). I have been using versions 
> from 3.5.4 to the current 3.5.8 and this same problem encounters in each 
> of them. I'm not even sure is this  problem related to the gnugo or to 
> the KGS client. But after both players have passed and death stones are 
> marked. There is a possibility to cheat by asking undo. It's hard to 
> explain but after requesting undo and human has played some meaningless 
> stone into the bots area. Then the bot starts to play inside it's own 
> area and losing points. I've attached a sgf file from one of SweGnuBots 
> game. This has happened atleast for my bot and few others aswell, but 
> this is the worst example of it.

What you see is GNU Go playing moves generated by the
kgs-genmove_cleanup command. The kgsGtp documentation says this about
the command:

kgs-genmove_cleanup: Optional, recommended for playing ranked games
    This is a kgsGtp-specific command. Engines that support it should
    treat it the same as the genmove command, except that they should
    not pass until all dead stones are removed from the board. This is
    used by kgsGtp when the engine and the opponent disagree on which
    stones are dead at the end of the game.

Later in the documentation about playing ranked games it also says
this: 

7 If the engine and the human disagree on the status of living and
  dead stones at the end of the game, one of two things will happen:
  If the game uses non-Japanese rules and the engine supports the
  kgs-genmove_cleanup command, then the engine will recommend that the
  human press "undo", after which the engine will get
  kgs-genmove_cleanup commands until it passes. If the game uses
  Japanese rules or the engine does not support kgs-genmove_cleanup,
  then the engine will simply refuse to score the game unless the
  human agrees.

Since the game was played with Japanese rules I think it's pretty safe
to say that there is a bug in kgsGtp.

Dan wrote:
> One step that you might consider is to maintain a list
> of players from which you don't accept matches.

No such option in kgsGtp and since it's distributed without source
you can't add it yourself either.

Paul wrote:
> If you don't know how to produce a log of GTP session in KGS client,
> ask William Schubert (at least I hope there's a way).

There is. The option "verbose" logs all GTP traffic and some of the
communication with the server. It is also possible to dump the GTP
commands in the GNU Go end with the --gtp-dump-commands option.

/Gunnar




reply via email to

[Prev in Thread] Current Thread [Next in Thread]