[Computer-go] Reminder: Human vs Computer Go competition in Barcelona tomorrow

Erik van der Werf erikvanderwerf at gmail.com
Fri Jul 23 05:02:21 PDT 2010

On Fri, Jul 23, 2010 at 1:30 AM, Jeff Nowakowski <jeff at dilacero.org> wrote:
> On 07/22/2010 11:49 AM, Erik van der Werf wrote:
>> IMO it has been clear for quite some time now that White has a
>> significant advantage at high levels. It seems the main reason for
>> still using 7.5 komi in 9x9 Go is just that KGS cannot handle other
>> values under Chinese rules...
> I know that you have mentioned in the past that it's extremely likely that
> the perfect-player komi is odd under area-scoring rules, so what would you
> propose instead?

Well, first of all I think in normal games players should be free to
chose any komi they like. Of course for rated games there may be some
restrictions but on KGS those don't apply to 9x9 anyway (9x9 games are
not rated).

For 9x9 computer go tournaments I would propose to use either komi of
6.5 or 7.0.
That way at least it's not a clear disadvantage to play first.

> Is the problem with the KGS implementation or Chinese (area-scoring) rules
> in general?

KGS implementation, area rules in general are ok, especially for
computer go tournaments because they make it easy to resolve disputes.

> Are you saying that Japanese (territory-scoring) rules are more
> fine-grained, and that 6.5 komi would result in a winning percentage closer
> to 50% among top programs?

Well, I may have made an argument like that before, but I'm not sure
about the winning percentage of 6.5 vs 7.5 especially under territory
scoring. I do however think that, especially once you get to the level
where it is often about getting the last point, territory scoring is
more interesting than area scoring.

> Or do you have in mind some other form of area-scoring rules?

Maybe add a button (see http://senseis.xmp.net/?ButtonGo)


More information about the Computer-go mailing list