[computer-go] Bot ratings and strength

[Łukasz Lew]

ZG1bot-200k is online for some time.
Łukasz

On 9/14/06, Don Dailey <drd at mit.edu> wrote:
> I have come to the same conclusion - 1500 is about as strong as you can
> get on CGOS without doing the tree based stuff or other tricks.
>
> GenericMC_200K had time control problems and I changed it 2 or 3 times
> to deal with it.   It  should be ignored and doesn't properly represent
> 200,000 simulations.
>
> - Don
>
>
> On Wed, 2006-09-13 at 14:56 -0700, Christoph Birk wrote:
> > On Mon, 4 Sep 2006, Don Dailey wrote:
> > > I just posted "GenericMC_10000 on CGOS.    This is a test to see what a
> > > very generic monte/carlo player should be able to achieve doing 10000
> > > simulations - less than a second per move.
> > >
> > > Others should be able to duplicate this bot exactly from my description
> > > and it would be nice if someone would try just as a check.
> >
> > I posted very similar bots:
> >
> > 'myCtest-2' runs 10,000 simulations (like GenericMC_10000, ELO=1143) and
> > settled at ELO=1126, very close to Don's program.
> >
> > 'myCtest-5' is identical except it runs 50,000 sims and scores 1373.
> >
> > 'myCtest-25' starts with 250,000 simulations (reduces this number
> > during the game due to time problems) and scores 1460.
> >
> > At 10000 simulations my program appears to score very close to
> > Don's and the 50k/250k versions indicate an upper limit of about 1500
> > for simple simulations; duplicating Don's tests.
> > I assume GenericMC_200K scored so poorly because of time problems?
> >
> > Christoph
> >
> > ps. ignore 'myCtest' and 'myCtest-1b' as they were (failed) experiments
> > _______________________________________________
> > computer-go mailing list
> > computer-go at computer-go.org
> > http://www.computer-go.org/mailman/listinfo/computer-go/
>
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>