Author: Uri Blass
Date: 09:45:21 05/24/01
Go up one level in this thread
On May 24, 2001 at 12:20:18, Miguel A. Ballicora wrote: >On May 24, 2001 at 06:49:09, Ferdinand S. Mosca wrote: > >>On May 24, 2001 at 05:14:07, Miguel A. Ballicora wrote: >> >>>On May 24, 2001 at 05:06:48, Miguel A. Ballicora wrote: >>> >>>>epd record is >>>>5rk1/1pp3p1/p1p5/7p/6P1/4b3/PPP2RPP/5RK1 b - - bm h4; >>>> >>>>[D]5rk1/1pp3p1/p1p5/7p/6P1/4b3/PPP2RPP/5RK1 b - - >>>>It is a modified version of a position from real tournament play that >>>>I saw in www.inforchess.com. I modified so a program won't find the solution >>>>for the wrong reasons. I tested crafty and could not find the solution >>>>even after 17 plies. I did not continue further. >>>>Is there any program that find the solution in a reasonable time? >>>>This is very easy for humans, I find the solution myself instantanously >>>>and took me few seconds to verify it. The original position is without the >>>>white pawn in g4. >> >>>> >>>>Regards, >>>>Miguel >>> >>>Of course, the solution is h4 and white is paralyzed. >>> >>>Miguel >> >> >>Very difficult indeed, Yace and Chezzz like 1...hxg4 and could not see 1...h4!! >>if then 2. g3 h3!, could this be a good test for pin and king mobility perhaps. >>Black has to see that a pawn on h3 is paralyzing for white. > >This position also involved a long term zugswang. Once white is paralyzed >it should move the pawns in the q side until it runs up of moves or black >captures them all. >I have an idea about making the test even more difficult for computers, where >there is paralysis but no zugswang. Zunzwang is exactly the difficult point for computers because null movers cannot detect zunzwang. If you give no zunzwang it will become easier for computers. A lot of programs are going to fail even in the following position [D]5rk1/6p1/8/7p/6P1/4b3/5RPP/5RK1 b - - 0 1 But firs, there should be a program >able to solve this "easy" problem... Of course if you give a non null mover 999999999999999999999999 years then it is probably going to solve it but it is faster to wait to better hardware and not start now. I do not believe that a top program can solve it because no programs can calculate all the pawn moves by search. A program that is not going to solve it in few seconds will probably never solve it. If you want to test the search of chess programs then do not give so many pawns in the board. Here is a test position that can be solved by search but it is not so easy [D]5rk1/p5p1/P7/7p/6P1/4b3/P4RPP/5RK1 b - - 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.