[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnubg] Resign bug

From: Joern Thyssen
Subject: Re: [Bug-gnubg] Resign bug
Date: Mon, 18 Aug 2003 14:39:40 +0000
User-agent: Mutt/1.4.1i

On Mon, Aug 18, 2003 at 10:44:10AM -0300, Albert Silver wrote
> I'd suggest you guys try to relax a bit. I'll add that similar problems
> occur in chess programs though a bit differently, and this has also been
> subject to debate. The most obvious one is when the program sees a very
> strong, and probably deep threat, and tries to minimize the damage. For
> example, it will suddenly see that you have a strong threat of mate in 8
> (which you aren't even close to having seen) so the program starts to do
> anything within its power to delay this disaster and begins sacrificing
> massive amounts of material, leading to an equally lost position of
> course. In the user's eyes, it looks like the program suddenly went
> crazy for no good reason and out of the blue began sacrificing its
> rooks, pieces, etc. The program's behavior will be identical whether
> this is a fairly obvious threat of mate in 3 or a monstrous mate in 23.
> The result was that many many users have complained about this over the
> years for 95% of the programs released. What has the developers'
> reaction been? A few, maybe 2-3 programs, have implemented special code
> dealing with these situations to humanize them. The vast majority of
> others continue to have the 'problem'. The other practical side is that
> when users win such games, they more often don't complain, since they
> soon realize the reason, and just show their friends how them forced the
> program to capitulate due to their mate in 23 threat (which they never
> saw). :-)

You can actually have the same kind of problems in backgammon: the
resignation offered by gnubg and the accept/reject of resignations by
gnubg is based on the assumption that the opponent plays perfect.

However, even the most basic player knows how to bear off chequers,
although I guess you can find some who don't :-)


Attachment: pgpT146erlfyT.pgp
Description: PGP signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]