Re: [Bug-gnubg] please have a look at

[Joseph Heled]

Ian Shaw wrote:
> 
> > -----Original Message-----
> > From: address@hidden
> > [mailto:address@hidden Behalf
> > Of Joseph Heled
> > Sent: 30 July 2002 03:12
> > To: gnubg
> > Subject: [Bug-gnubg] please have a look at
> >
> >
> >
> >
> > 'http://pages.quicksilver.net.nz/pepe/gb/top.html#A first look'
> >
> > I need you input!!
> >
> > -Joseph
> >
> Joseph wants examples where GnuBg is playing a containment position badly.
> How does one best determine this? I'm not an expert who can judge it by
> experience.
> I found a position where the odd-ply evaluations were very different from
> the even-ply evaluations, which made it obvious.
> I'm looking for a numerical
> value I could use as a test condition that something funny is happening.

(Unfortunately) this is a good test. The thing I look for is the move made.
Regardless of level of expertise, in many cases most humans will recognize a
doubefull/awful move by the bot.

-Joseph

> Another test would be if the rollouts differed from the evaluations by a
> lot, but that's a much slower test to implement.
> 
> Ian
> 
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg