[Computer-go] kgs-genmove_cleanup again

William Shubert wms at igoweb.org
Mon Mar 29 00:20:03 PDT 2010


On Sun, 2010-03-28 at 16:59 +0200, Erik van der Werf wrote:

> On Sun, Mar 28, 2010 at 4:18 PM, Alain Baeckeroot
> <alain.baeckeroot at laposte.net> wrote:
> >
> > Le 28/03/2010 à 01:19, Petr Baudis a écrit :
> >>
> >>   Hi!
> >>
> >>   I have yet another query about kgs-genmove_cleanup - I think its
> >> behavior now is rather user-unfriendly and also a bit confusing.
> >>
> >>   The genmove cleanup command is defined "do not pass until there are no
> >> dead groups on the board".
> >
> > This is done in gnugo with the "capture_all_dead".There are some pitfalls,
> > but it is not that hard to implement, and it is clean, no possibility
> > of mistakes on scoring after this.
> >
> > I think conforming to the request of the server is better than modifying it
> > to afford an half-baked solution, and i bet wms will tell you to fix your
> > program :-)
> 
> 
> Ok, but why then doesn't the same apply to the human player? Looking
> at the game Petr mentioned at the (2nd) end the human player was again
> allowed to mark groups dead *without* explicitly capturing them!
> 
> I had always assumed that in the cleanup phase the rules are changed
> so that finally all remaining stones will be assumed to be alive.
> However, apparently the human player is allowed to play under a
> different set of rules, which enables him to cheat.

The difference is that humans can talk with each other to resolve
differences. Computers can't. So they must follow a stricter set of
rules.

Alain has it right. :) I don't see what will go wrong if you change your
program to implement the cleanup command correctly. The original email
game a log of what happened in a game that ended up wrongly scored, but
admits that the program hadn't followed the cleanup protocol fully. If
it did, then the game would have been correctly scored. So the problems
look to me to be in the program, not in the protocol.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20100329/ed491604/attachment.html>


More information about the Computer-go mailing list