Click to follow
The Independent Culture
ONE OF the most interesting aspects of the sometimes symbiotic, sometimes frankly antagonistic relationship between chess and computers, has been the development of endgame databases, encapsulating perfect play for some specific material balance.

It was initially developed for just four pieces - queen vs rook was one of the very first - but advances in computer hardware have now made five- piece endings readily accessible.

Indeed, the four-CD-rom set of these, developed by Ken Thomson of AT & T Bell Laboratories and marketed by ChessBase of Hamburg, has been extant for some time. Incorporating all endings without any or including just a single pawn, it's available from either Chess and Bridge (0171-388-2404) or the BCM Shop (0171-603-2877) for about pounds 100 including postage.

For all the joy of "perfect play", however, many of these endgames, especially the ones without pawns, look like nothing on earth. We mere mortals are used to pawns giving a position structure, so the new development of endgame databases with more than one pawn is particularly interesting.

I touched on the Dane Lars Rasmussen's pawn ending engine recently when discussing an ending "discovered" by John Nunn and am now delighted to have been able to experiment with it myself. (He's working on a final commercial version, to be released anon.)

Running on 32-bit windows (so you require Win95. 98 or NT), it has three parts: a database generation program (pawngen.exe), a database reader (pawn.exe) and a small file (queen.dat) pertaining to queen vs pawn. It can't assess positions with queen (and pawn/s) vs several pawns, but this is a small blemish compared to the vistas opened up.

To examine a particular pawn structure you key it into pawngen using a simple interface. I tried f2, g2, h2 vs g7, h7 which required 106,348,544 bytes - ie 8,192 per position x 12,982 pawn structures. Running on my elderly Pentium 200 it took just under 14 minutes to generate the database (pawn, dat). You use this with play.exe and can discard it later.

For me, perhaps the greatest interest was testing my intuition of what is and isn't winning. With a little work you can also find striking positions such as this, similar to ones I've met before but still distinct, which arose in one line (to save space I'm just listing it).]

Black: King Ke6, Pawn h7.

White: King Kf4, Pawns f3, h5.

The obvious 1 Kg5? fails after Kf7 2 Kf5 Ke7 3 Ke5 Kf7 4 f4 Ke7 5 h6 Kf7 6 Kd6 Kf6 when White is in zugzwang and if 7 Kd7 Kf7!

Instead White must play 1 Kg4! Kf6 2 f4 Kg7 3 Kg5 Kf7 4 Kf5 Ke7 5 Ke5 Kf7 6 Kd6 Kf6 7 h6! when Black is in zugzwang:

a) 7 ...Kf5 8 Ke7 Kxf4 9 Kf6! Ke4 10 Kg7 Kf5 11 Kxh7 Kf6 12.Kg8

b) or 7 ...Kf7 8 Kd7 Kf6 9 Ke8 Ke6 10 Kf8 Kf6 11 Kg8 Kg6 12 f5+! etc.