Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strange error in Hiarcs 7.32 game

Author: leonid

Date: 12:53:33 02/02/02

Go up one level in this thread


On February 02, 2002 at 10:31:12, Uri Blass wrote:

>On February 02, 2002 at 09:24:40, leonid wrote:
>
>>On February 01, 2002 at 02:44:18, Mike Hood wrote:
>>
>>>I'm posting the whole game of an engine-engine match I played on my PC. White
>>>was Hiarcs 7.32, Black was Chess Tiger 14.0, both were using 128 MB Ram. The
>>>time conrol was 40'/40+40'/40+40'
>>>
>>>The strange move is move 121. The position looks to me like a drawn game. Hiarcs
>>>thought for 25 seconds to a search depth 13, evaluated Ka6 as 0.0, made his
>>>move, and oops! -- he'd overlooked a mate in 1.
>>>
>>>I've tried duplicating this error in vain. In analysis mode Hiarcs sees the
>>>Mate. Has anyone else noticed similar things happening with Hiarcs?
>>
>>
>>Probably, every chess program should have one mate solver chained to its search
>>for needed move. First work must be done by mate solver and only when mate not
>>found, second part should be activated. Mate solver must look (my guess) 6 ply
>>deep by brute force and later (if first search failed) 14 ply by selective. All
>>search by mate solver must take 0.05 sec, or even less. So, enough time for
>>second part of chess program to look for needed move, if mate is not found.
>>
>>By seeing efficency of brute force search done by Heiner's mate solver and speed
>>of actual chips, I think that 4 move brute force search (before each move) is
>>possible. Second selective search for mate could easily go as far as 8 moves.
>>This way 1 move mate blunder will be impossible for ever in every program.
>>
>>Usually mate, in average game, is only between 2 and 5 moves deep.
>>
>>Leonid.
>
>Unfortunately it is not going to help in the relevant case because hiarcs did
>not miss a mate for itself but miss a mate for the opponent.

Ha, ha! This is true.

Leonid.


>A mate solver could tell hiarcs that a mate is not found and not prevent hiarcs
>to do the blunder that let the opponent to do a mate in 1.
>
>A mate solver can help to prevent that kind of mistake only if it is used after
>the move is decided by the program but before the move is played to check that
>the move is not a blunder but in this case a change in the program is needed to
>force it to calculate again when the mate solver tells it that the move is a
>blunder.
>
>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.