Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: 6-man Chess solved

Author: Vincent Diepeveen

Date: 05:42:23 08/09/05

Go up one level in this thread


On August 09, 2005 at 06:06:01, GuyHaworth wrote:

>
>I infer, from the fact that Eugene Nalimov sent me the KPPKPP stats file this
>morning, that 6-man chess is fully solved.  This is on the reasonable assumption
>that the EGTs have verified correctly as aok.

Looks like to me the assumption is correct for 3 vs 3.

I'm not so sure he has already generated 5vs1 though.
obviously 5 vs 1 is not so useful from 6 men viewpoint seen,
just like 6 vs 1 isn't from 7 men viewpoint seen.

Yet obviously 3vs3 is the biggest hurdle in the 6 men and Nalimov completed
his task there.

Note 5vs1 is needed to generate for example a 5vs2 stone.

Of course the small 7 stone are not very interesting and relative easy to
generate (like KQQQKQQ), but in whose format will all the 'useful' 7 men get
generated first?

>My congratulations to Eugene on this achievement, a tribute to a professional,
>determined and robust approach to the challenge.
>
>My thanks also to Rob Hyatt for making the data accessible, and to those who
>have linked their chess engines to the EGTs and mined them for greater
>knowledge.

>This milestone is passed on the 48th Anniversary of the 1956 'Artificial
>Intelligence' conference at Dartmouth which targetted world-class computer chess
>as an AI aim.
>
>guy

From objective viewpoint seen, it's pretty frightening slow at which the
development of the EGTBs happens. There is programs like hydra which search
a few hundred of million nodes per seconds and in this world champ nearly all
participants will hit far over 1 million nps.

With the 200+ million nps Hydra gets, searching till a few trillion nodes
(assuming the american notation : 1 trillion = 1000 billion),
doesn't take *that* long.

In fact it's seeing a few trillion nodes per game.

The total size of the EGTBs so far is just a few tera-entry. I'm not sure
about Nalimov, as his indexing scheme is a bit more efficient than that of
dieps, but diep needs roughly 5 Tera-entry. Uncompressed all 6 men are exactly
1 TB.

If we compare that with the search speeds of the fastest program,
then the progress in terms of EGTBs hasn't been so big yet.

Of course the 2 reasons for that is its relative importance (instead of
generating 7 men EGTBs at my machines i'm playing testmatches at my machines in
order to improve the engine) and the limited capacity of harddrives (for a
single giant 7 men you'll need nearly 2 TB storage space for diep's generator,
to generate it) and it'll run for months to generate just one.

In that respect EGTBs are moving forward slowly.






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.