Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 7-man endgames (long)

Author: Marc Bourzutschky

Date: 08:35:00 08/29/05

Go up one level in this thread


On August 29, 2005 at 10:45:38, Duncan Roberts wrote:

>On August 29, 2005 at 03:29:12, Marc Bourzutschky wrote:
>
>>...snip

>Great achievement. well done.
>
>
>how long would a dtm program have taken to solve the  6-man krnknn endgame on
>your machine. ?
>

We don't have a DTM version, so I don't know.  In this particular case, probably
not that much longer.  The real benefit of DTZ is in endings such as kbbknn,
where the bishops may capture to a very long kbbkn ending, artificially creating
very long lines for kbbknn.  In DTZ all you care is whether the subgame is won
or not.  In pawn endings the benfit is even more dramatic.

>have you ever tested the elo increase in strength for a complete 6 man endgame ?
>

No.  I don't believe 6 and even 7-man tablebases will improve playing strength
significantly.  First of all, few games ever get that far into the ending before
being decided, and secondly it is very difficult to have an ending avoid wasting
time probing in positions it could handle perfectly well on its own.  My own
work right now is purely focused on the "search for truth" rather than improving
computer playing strength.

>how many computer months roughly do you reckon it would take all the pawnless 7
>man database to complete as well as the whole 7 man database. and how many
>positions are there ?
>

It would take several computer years, but I'm not even thinking about starting
such a project.  The vast majority of endings would be of no interest.  Our next
focus is probably on endings such as kqppkqp, krppkrp, kbppknp, perhaps only
considering promotions to queen to avoid having to generate massive amounts of
strange tablebases.

>is there any computer knowledge you can extract from these tests that could be
>put in an evaluate for those particular positions.

I'm sure there is, but extracting such knowledge even from 5 or 6 man endings
has not really been done successfully so far.
>
>
>thank you
>
>duncan

-Marc



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.