Computer Chess Club Archives


Search

Terms

Messages

Subject: Tablebases generation: help request

Author: Guido

Date: 06:35:58 08/30/99


I downloaded the Nalimov's tablebases generator
(ftp://ftp.cis.uab.edu/pub/hyatt/TB/tbgen.zip) and tried to compile and execute,
that happened without problems under an old Linux version  (Slackware 1.2.13),
installed on an Intel P133.

By the way I observe that compiling with gcc option  -fomit-frame-pointer
saves about 10%  in execution time since this option adds an extra register for
local (register) variables.

As I wrote a similar program that generates tablebases, I was interested in
comparing results.

Analyzing randomly single positions I found coincident answers in terms of
number of moves to mate or to lose, but statistics differ between tbgen and my
program, except for kpk and  kppk endings.

For all the other endings I obtain the same maximum number of moves for victory
or defeat, but the number of positions, won or lost, versus number of moves, are
a little different, always with larger values for  tbgen  in respect to my
program, and therefore the total number of valid positions (no. bytes of the
file - broken positions?) differs too.

I suppose that my program should be wrong in analyzing symmetry or checks,
eliminating positions that should have been considered.

Now I would like to find the error and therefore I ask if somebody can kindly
help me.

I found the simplest example of inconsistency in the file kqkn.tbs where I
notice that  White moves and wins in 21 moves in 4 positions, while my program
indicates that such positions, keeping into account symmetry, are only 3, and
exactly:

1.  White: Ka1, Qb1     Black: Kc3, Nc2             White moves
2.  White: Kb1, Qb2     Black: Kd3, Nc3             White moves
3.  White: Kd1, Qe2     Black: Kc3, Nb2             White moves

I checked  these positions successfully with the command  tbgen -q, but I don't
know how to find the missing one, without trying more than one million
positions.

Does a program exist to extract positions with the same result from Nalimov's
tablebases?

Knowing the fourth position I hope to have a good hint to detect the error.

Thanks in advance
Guido






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.