Author: Aaron Tay
Date: 02:35:44 10/06/01
Go up one level in this thread
On October 06, 2001 at 03:02:16, Uri Blass wrote: >>This is very difficult to say. Most of the important once are those with K?P >>vs KP. The promotion of 'P' would then require that you also get the >>additional five-fold tablebases. >No >programs can use the KPP vs KP without having the other 5 piece tablebases. >They can do stupid mistakes if they get the KPP vs KP without the KQP vs KP if >they are not designed to use correctly part of the tablebases Yes. yace, Crafty etc can handle this, but can most others? >playing without KQP vs KP almost always does not cause programs trouble. yes I think you are right. It will only be a problem if the stronger side needs to queen the pawn to win but does not due to lack of a KQQKP tablebase. But I can't seem to think of any examples. In any case, More often than not the most common conversation case is KPK after exchanges of queens which most programs are more than happy to do. From the weaker side's point of view it will always queen the pawn if possible, since the score found from search after reaching KQPKQ can;t be worse than the certain defeat in KQPKP.. >Having the KPP vs KP without the KQP vs KP may cause programs trouble if they >are not designed correctly but in this case the reason for the trouble is not >only not having the KQP vs KP tablebases and the proof is that the program can >usually avoid the mistake of not going to KQP vs KP endgame without tablebases. I'm being testing out the KRPKP positions [The KQPKP positions are probably similar and I just did a quick test with KPPKP] without any promotion cases and it seems that even in certain limited cases with enough time the program can work out that the best move ie promoting the pawn will lead to a better score [actually faster mate] than the ones in the tablebase. To see what i mean with diagrams see the URL before http://www.chesskit.com/aarontay/Winboard/egtb.html#[A.9]
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.