Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Generating endgame tablebases

Author: Gareth McCaughan

Date: 15:15:07 10/11/01

Go up one level in this thread


Robert Hyatt wrote:

> I believe Lewis published some of them quite some while back.  But the
> thing he published was mainly "deepest mate" and the like.  I didn't follow
> this very closely, and only remember Wendroff asking for "my contact" up at
> Cray in trying to set the time up...

The book "Games of No Chance" (edited by Richard Nowakowski,
published by Cambridge University Press in the UK and maybe
by MSRI Publications in the US) contains an article by
Lewis Stiller entitled "Multilinear algebra and chess
endgames". It contains a table giving, for each of 41
pawnless 6-piece endgames, the number and percentage of
positions won for White, the maximum distance-to-win of
any position with those pieces (among those that are
wins for White), and the number of mutual zugzwang positions
(i.e., positions where WTM draws and BTM loses).

The class with the longest DTM is KRN v KNN. 78% of
these positions are won for White; the longest DTM is
243 moves; there are 18176 mutual zugswangs. One position
(perhaps the only one, up to symmetry; it's not clear)
that takes the maximal number of moves is:
[D]6N1/5KR1/2n5/8/8/8/2n5/1k6 w - -

There is a position in KQR v KQR which is won for White
but requires 92 moves to win.

KRB v KNN, unlike KRN v KNN, is almost always won
(96% of positions, the others presumably being ones
where White has a piece en prise or something of
the sort), but its longest DTM is 223 moves.

(I use "won" to mean "won if the 50-move rule is
temporarily repealed".)

The book also contains some interesting mathematical
analysis of a few chess endings. It's of more interest
for mathematicians than for chess players, though. :-)
As an example, they explain why this position
[D]8/1p5p/p7/4k3/4Pp2/5K1P/PP5/8 w
is a win for whichever player moves first.

--
g



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.