Author: Robert Hyatt
Date: 21:27:55 10/10/98
Go up one level in this thread
On October 10, 1998 at 11:33:14, blass uri wrote: >I think that in some endgames like KRN vs KR you can save space >by defining rules to do the right moves: > >If there is conversion in a small number of plies you can find it by search and >not use the tablebases to tell you the right move > >In other positions you can use tablebases only if the rules are not right >I believe it is possible to define rules that you are not wrong in 99% of the >cases >and for 1% of the cases you can save a smaller file with the positions and the >right moves. > >you can have rules for win,draw,loss too and again you can store only the >exceptions. > >I think it is more difficult to define good rules in other positions like KRB vs >KR >but at least for part of the endgames you can save a lot of space. > >Uri First, you can "suppose" most anything you want, I "suppose"... but that doesn't make it 'reality' by any stretch. The following "holes" appear in your "theory": 1. KRN vs KR is not something to worry about since it is a draw in almost all cases. 2. Even in QK vs KR, specific "rules" are *very* difficult to define, so difficult that I don't know of a program that can win this ending without databases. 3. In many databases, the methodology to win is not understood anyway, as some of the papers in the JICCA (where a GM tried to find winning plans by probing a database). IE until the database for KQ vs KR was constructed, the "great flying rook" defense was not understood. And there are many more of these sorts of things that are simply not understood. Excepting KQ vs KR, I would hope that Crafty can force mate with any other combination that is winnable, such as KBN vs K, or KBB vs K. But once the 5-piece files are hit, trouble arises... And don't forget how the databases are accessed... there is *no* searching of any kind... just a direct probe to the byte that says mate in N, mated in N or draw... In KRP vs KR, this is quite difficult because there are lots of tricks, lots of exceptions, and lots of problems... The point of databases is to "keep it simple". At present they are...
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.