Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hsu Presents a Paper at

Author: Steven J. Edwards

Date: 23:52:30 06/25/98

Go up one level in this thread


On June 26, 1998 at 01:13:38, Robert Hyatt wrote:

>On June 26, 1998 at 00:44:08, Don Dailey wrote:
>>One  more  example.   Endgame   databases are an   example  of PERFECT
>>knowledge  and  most databases contain  thousands   if not millions of
>>terms or  parameters or whatever  you  choose to  call them.   And yet
>>despite this massive amount of knowledge,  they do not contribute more
>>than a few rating points to a chess program.  And this is with PERFECT
>>knowledge.   Most of the strength   they do contribute is concentrated
>>among 2   or 3 common  endings.  It's  a  whole lot of  knowledge, but
>>rarely used.

>I disagree here.  KRP vs KR is a big winner.  I win many pawns, against
>GM's and more particularly against other programs, and many of these games
>end up KRP vs KR where I win when ahead, and draw when behind.  A couple
>of weeks ago, Crafty was playing DiepX, and in 4 successive games, 3 were
>won by this endgame database.  Because I won a pawn, held on to it, trading
>into a won position at just the right moment.  3 of 4 there.  There is hardly
>a day (playing other computers) where I don't see several of these.  Against
>humans, no, because they break badly when they break and the game doesn't
>drag on...

A couple of things:

1) It would be difficult to state precisely the rating gain by having perfect
endgame knowledge.  It would depend upon: 1) the form (DTC/DTML) of the
databases, the number of the databases, the distribution of the databases, the
access time of the databases, and very importantly, the conditions which enable
database probes (i.e., which nodes).

2) One must also consider the alternative to endgame databases; this being hand
coded routines with lots of complexity and special cases.  Will these be faster?
 Maybe.  Will they be correct?  Maybe.  Will they be precise?  Maybe.  Will they
be more correct, more precise, and simple as endgame databases?  Never.

3) Opportunity cost: the time saved by incorporating endgame databases can be
spent on other parts of program development.

4) Cost trend: disk capacity per unit cost increases at about 50% per year and
access speed increases at about 10% per year.  Having endgame databases helps a
program to share in this beneficial trend.

5) Sharing: not too many programmers have offered their hand coded specialized
endgame routine sources to the public.  But Thompson's endgame databases are
easily accessible and the tablebases are free as is their generator program.

6) If a program makes it to the endgame, then having tablebases really helps.
With probes being done on interior full width nodes, a root position with a
dozen men will often hammer the five and four man classes even with less than a
minute per search.  Every time that a successful probe is made, then not only is
that entire subtree (with its effectively unlimited depth) cut off, but greater
depth is obtained by having more time to spend on other nodes.  Also, a probe
that returns a mate/loss score will often cause some serious a/b cutoffs.

7) Against human opposition, endgame databases offer a psychological advantage
of sorts when opponents know they don't have a chance in the endgame unless they
have both a winning advantage along with extremely accurate technique.
Therefore, some humans may be intimidated into trying to force a win in a
tactically-rich middlegame -- not always a good idea.  So having the databases
may contribute rating points even if they are never probed.

-- Steven (sje@mv.mv.com)





This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.