Re: [gnugo-devel] Reliable client?

[Graham Thomson]

Yes, this embarrassingly simple oversight is the problem! All three clients score correctly after manually removing dead stones.


Thanks for the tip,


Graham.

On 01/09/05, Gunnar Farnebäck <address@hidden> wrote:

Graham wrote:
> I have recently starting playing with GNU Go 3.6 - great program! But,
> I started to suspect that the clients recommend on the GNU Go website
> were not scoring the game correctly.
>
> Playing the following game with GNU Go on the command line has the
> result of about W+52 (9x9 board, human black):
>
> Black builds up some territory in the bottom right by playing c3, d2,
> b4, a4, d1, b2, a2, b1. Then, black invades in white's territory, say
> with d9, h4. Black then passes until white does also.
>
> However, playing the same game against GNU Go with Jago 4.6, glGo
> 1.3.1, or qgo 1.0.2 all result in a score of 5 for black, and 0 + komi
> for white.
>
> Am I missing something here?

This is untested but my guess is that those clients expect the user to
mark the dead stones manually as part of the scoring procedure. If
black's invading stones are not marked as dead they will cause white's
territory to score as dame instead, i.e. no points.

GNU Go on the command line on the other hand marks dead stones by
itself.

Is this the problem?

/Gunnar


_______________________________________________
gnugo-devel mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/gnugo-devel