[Computer-go] Operators for Frisbee Go Simulation

Josef Moudrik j.moudrik at gmail.com
Sat Apr 16 07:19:05 PDT 2016


Hi, I had nothing to do this morning, so I hacked together a very
rudimentary version of frisbee-twogtp. The GTP support is done by
introducing three new commands: frisbee-play, frisbee-reg_genmove and
frisbee-epsilon. I chose to add new commands for play and reg_genmove
because of the shift in semantics, I think it is better than subtly change
semantics of std. play and reg_genmove commands. I hope to make my own
frisbee bot soon :-)

https://github.com/jmoudrik/frisbee-twogtp

The purpose of frisbee-play is to support another special value - apart
from PASS - and that is SKIP (meaning involuntary pass as has been used
here). I feel that calling it something else than pass makes actually sense
(and simplifies the terminology imo).

The frisbee-reg_genmove behaves just like a normal genmove, except that
bots might return incorrect moves (of course, hoping that it will land
somewhere nice). To add to the discussion above, I think that this makes
sense to allow, since e.g. for some rather high values of epsilon, the
probability of landing on the center stone is small and "what if" the four
neighbors would be some good moves - the condition to only return valid
moves seems as having no real purpose. Of course, "return incorrect moves"
 at the beginning of the paragraph still means "on board", because adding
support for deliberately aiming outside of board would require some
coordinate-redefinition-fun. (But again, aiming outside of board seems to
make some sense in some positions, one might want to play e.g. A1 which
might have very high utility (killing a group), while landing at A2 might
make the group alive, so expected utility of throwing at A0 might easily be
better than throwing at A1 itself or A2).


Regards,
Josef


On Thu, Apr 14, 2016 at 7:16 PM Gonçalo Mendes Ferreira <gonmf at sapo.pt>
wrote:

> Hm interesting question.
>
> komi
> 8.0   49%
> 7.5   50%
> 7.0   51%
> 3.5   56%
> 1.5   58%
> 1.0   60%
> 0.5   61%
> 0.0   61%
>
> Also these win rates do not include the probability of drawing.
>
> Gonçalo
>
> On 14/04/2016 18:08, "Ingo Althöfer" wrote:
> > Hi Goncalo,
> >
> >>      accuracy p
> >> komi  0.5  0.2
> >> 7.5   31%  22%
> >> 3.5   43%  36%
> >> 1.5   48%  45%
> >> 1.0   49%  47%
> >> 0.5   51%  49%
> >> 0.0   52%  51%
> >
> > Interesting.
> >
> > Concerning your bot in "normal" 9x9-Go:
> > Which win rates do you get there for different komi values?
> >
> > Ingo.
> > _______________________________________________
> > Computer-go mailing list
> > Computer-go at computer-go.org
> > http://computer-go.org/mailman/listinfo/computer-go
> >
> _______________________________________________
> Computer-go mailing list
> Computer-go at computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://computer-go.org/pipermail/computer-go/attachments/20160416/5a1c0e27/attachment.html>


More information about the Computer-go mailing list