Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Playing strength of simple eval programs?

Author: blass uri

Date: 09:18:46 03/01/00

Go up one level in this thread


On March 01, 2000 at 11:31:50, blass uri wrote:

>On March 01, 2000 at 11:19:56, Kim Roper Jensen wrote:
>
>>I just wondered about what is the strength of a simple computer chess program,
>>by simple i mean that you can have all the 'works' regarding seaching techniques
>>, null move, razoring, hash tables etc. But the only evaluation routine is count
>>of material .
>>
>>My own assumption is that on 'low' searches ( between 1 to 10 plys, or short
>>time controls ) the programs will have little playing strength, but what about
>>longer time controls ?? I have the idea that the speed benefit ( no time
>>consuming eval. ) will give the program, deeper plys so on pure tactical reasons
>>it will not play 'bad' chess ( meaning some really bad positional move )
>
>I think that practically it is going to play bad chess.
>I do not think that you can reject lines like 1.e4 e5 2.Qh5 or 1.c3 e5 2.a3
>based on material only evaluation even if you search 20 plies.
>
>I do not think that there is a speed benefit about plies because ny intuition
>tells me that the program is going to have a bad order of moves and the fact
>that it search more nodes per second is not going to help it to search deeper
>because the disatvantage from bad order of moves is more important.
>
>Uri

I think that it can search deeper in short time control because almost nothing
is losing material but when you try long time control you may have a problem:

example:

suppose 1.f3 is the move you want to play and it is losing material against
1...e5 at ply=20.

If you have a bad order of moves you start searching 1.f3 a6 and only after many
searches you find that 1.f3 is losing material against e5

if you start with 1.f3 e5 and discovers that e5 is winning a pawn for black
the alpha beta let you search less time other moves because you only have to
prove that lines like 1.f3 a6 do not win more than a pawn for black.

Of course you cannot practically prove that 1.f3 is losing a pawn and it was
only an example to explain why I believe that only material evalluation function
is not productive to search deeper at long time control.

You may search sligthly deeper at the beginning position because there is
nothing that lose by force there but when you get to positions when searching
slightly deeper is productive I believe that you are going to be unable to
search deeper.

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.