[computer-go] End game in CGOS

[Aloril]

On Sun, 2006-09-24 at 16:17 -0400, Don Dailey wrote:

> KGS tried to work this out by have a protocol that continued the play
> when there was a disagreement on what was dead.   It did not work
> because many program just continued to PASS and disagree.   I don't know
> the current status of KGS in this regard - but of course it's possible
> to fix this with some  arbitrary rule such as "score as is after the 2nd
> disagreement, etc."   But it's not addressing the basic issue of program
> cooperation which cannot be enforced.

I believe that in KGS3 dead dispute resolving between bots in tournament
games works correctly, at least it seemed so in latest test tournament
in beta KGS3 server. There is though 'feature' that allows you to mark
stones in a way that is advantageous to opponent: You can mark opponent
dead stones as alive or your alive stones as dead and opponent bot won't
contest it. This will make onlookers to wonder what is happening but
should not make things worse for bot that is behaving correctly.

There is no disagreement after cleanup phase: all stones are assumed
alive except for 'feature' mentioned earlier. Properly working bot would
have removed all opponent dead stones no matter what opponent does. If
opponent passes and fails to remove dead in its territory, too bad for
it ;-)

Cleanup phase is only needed when bots disagree on stone status and then
its actually equivalent to CGOS rules, except for that 'feature'.


(I think maproom will score manually as needed for a while until most
bots work correctly with above?)

-- 
Aloril <aloril at iki.fi>