Author: martin fierz
Date: 05:24:31 02/18/04
Go up one level in this thread
On February 18, 2004 at 06:38:33, Uri Blass wrote: >On February 18, 2004 at 05:37:41, Tord Romstad wrote: > >>On February 17, 2004 at 20:55:13, Dann Corbit wrote: >> >>>Here's a dumb idea: >>> >>>Write a program to scan a Nalimov database, but throw away everything except >>>won/lost/drawn/broken (needs 2 bits per reflected board position to store the >>>outcome state). >>> >>>Then write a table. >> >>There are two problems with following this approach: >> >>1. If you just use a table, you risk to miss the opportunity to discover >> principles which can be useful even in more complicated endgames. >> >>2. The memory requirements are big. A few MBs of RAM may not seem like a >> lot on modern computers, but it is not very aesthetically pleasing to >> use so much memory in order to do something as simple as evaluating >> KRKP endgames. Besides, some of us (or at least one of us) want to >> port our engines to Palm OS and similar platforms, where memory is >> limited. >> >>Tord > >I think that the first problem is KPK >I still have 1843 positions white to move (pawns in a-d file) that the >evaluation that I have now returns do not know. > >I wonder if your evaluation knows to detect win or draw in the following >positions(I think the second position may be more easy to generalize a rule >based on it): mine doesn't know about this. i have a simple and fool-proof win detection, but in most cases i don't know. however, i agree that it is worthwhile working on this if you don't use TBs. even if you do use TBs you might find good evaluation terms for more complex endgames. cheers martin >[D]8/k7/8/8/8/2P5/K7/8 w - - 0 1 >[D]8/8/1k6/8/8/3P4/K7/8 w - - 0 1 >[D]8/k7/8/8/8/3P4/K7/8 w - - 0 1 > >Uri
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.