Computer Chess Club Archives


Search

Terms

Messages

Subject: KQQKQQ etc: a proposed way forward

Author: guy haworth

Date: 07:56:12 09/14/99

Go up one level in this thread


Michel

Thank you for your very positive response.

Robert Hyatt is quite right to suspect that those asking for this piece of work
underestimate its difficulty and size and to 'caveat' on the intrinsic problems.

The best hope is that an existing and proven code can be run on your machine.
The possible sources would appear to be:

    Eugene Nalimov's, on the CRAFTY server
        don't know if this exploits parallel-processor architectures

    Wirth/Nievergelt's, not on any server I know but ....
        which exploits parallel-processor architectures to effect
        which checks positions' values/depths against those of their successors

By running a proven program, especially one with 'redundant' self-checking like
W/N's, your results would have some inherited credibility.

Hopefully, these programs can also take advantage of the fact that KQQKQQ has
only a quarter of the distinct positions of KRBKRN for example.

The likely Kasparov-v-WorldTeam position is 'Position D' in the game-analysis,
namely:

    7Q/1p6/3p2K1/6P1/8/8/8/1k1q4, wtm

ie wP on g5 aiming for g7/g8 and bPs on b7 and d6 aiming for b2/1 and d2/1.

Let's assume a 7-man endgame database is too difficult.  The relevant 6-man
databases are:

    KQP(g5-g7)KQP(b7-b2 or d6-d2) ... 33 databases

Those with the wP on 'g8' and g7 need to be calculated before those with the wP
on g6 and g5 etc.  This is of course a bit unfortunate as the World Team needs
the 'g5' database before it needs the 'g6' database.

With some small loss of accuracy, it's worth assuming (simplifying heuristic)
that Pawns convert to Queens and that KQQKQP is a win for White unless the P
converts immediately

The easiest db to work on is KQQKQQ and I have a call out to Eugene Nalimov
asking for help.  Hopefully, you two can come together.  Presumably, Nalimov's
code works back, standard retrograd analysis style, from 'wins in 0' (not
necessarily mates) to 'wins in 1' etc.  The deepest wins are 44m deep but there
are lots of shallow wins where the 2nd player has a Q en prise or is skewered
through the K.

Even if you don't get back to depth 44, it would be VERY useful to know what the
shallower wins are to avoid them - and of course, the more the merrier.

Although KQQKQQ is not the endgame profile right now, it would be worth knowing
whether it was worth the World Team playing to achieve.

Regards, Guy




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.