Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Symbolic: code example

Author: Uri Blass

Date: 03:19:46 01/24/04

Go up one level in this thread


On January 24, 2004 at 05:11:17, Tord Romstad wrote:

>On January 23, 2004 at 16:07:04, Uri Blass wrote:
>
>>I think that your program is not less complicated than Crafty and you have a lot
>>of things in your evaluation and your search that Crafty does not have.
>>
>>Gothmog is weaker than Crafty but it proves nothing because you also did not
>>spend a lot of time on it.
>
>The above might be true, but it is not very interesting in the present
>discussion, because
>chess programming using conventional techniques is not a very difficult
>programming
>task.  What we all do is basically implementing and fine-tuning relatively
>simple and
>well-known algorithms.  All the hard work of inventing the techniques has been
>done
>in the past.  Like Newton, we're standing on the shoulders of giants.
>
>Tord

It is interesting because Crafty does not implement ideas that are posted in
Ed's page.

basically I decided to go for chess programming because  I believe that chess
programs are stupid and search too big trees(it is the same for movei but I hope
to change it and the problem is not having the right ideas but implementing
them).

For example chessmaster6000(I did not buy later version) searched for a long
time a line when in some move near the beginning of the line white can force a
simple mate(I know because chessmaster6000 also has an option to shows the line
that it search in every second).

Note that my program today is not better and can have similiar behaviour but I
plan to change it after I rewrite my alphabeta.

I think that the problem is that programs do not calculate exact scores for
moves at reduced depth.

Calculating exact score can help the program to save time by not searching these
lines and if the depth is reduced enough the average save is going to be bigger
than the average cost.

The problem is that I found in the past that doing an additional search without
making a move cause my program to need more nodes to finish the same depth even
if I do not use the data of the additional search and reduce the number of nodes
that I search to get the exact score.

The problem is probably that I used the search to get exact scores to update
more information that influence the search so
I decided that I need to rewrite my alphabeta but I prefered to test in the last
months other ideas instead of doing it because even without searching to get an
exact score there are other ideas that can help my program.

Not knowing an exact score at reduced depth is only one example for my opinion
that programs of today are stupid(I guess that not knowing an exact score at
reduced depth is not common for all the programs because the idea seems to me
too simple not to think about it but I saw it also for other top programs when
they wasted time on a move that is losing because of a simple mate).

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.