Re: Searching simple opensource chess program

-nospam- <[email protected]>
Newsgroups gmane.comp.gnu.chess.announce
Organization Telenet Internet
Message-ID <[email protected]>
Paul Pichota wrote:

> For a university project I want to compare different algorithms 
> (minimax, alpha-beta and some variants). So I am looking for rather 
> simple and working opensource program where I can easily exchange the 
> algorithms.

I have tried the same stuff but I came to following conclusion: In order 
to evaluate an algorithm you have to cross-check the results with the 
results obtained from a brute-force search.  The brute-force attack must 
come to a position where the score is agreed with all parties involved.

This has 3 major drawbacks.

1) Brute force search in chess is very time consuming.  Move generation 
is complex and takes a lot of time if a large tree is to be searched. 
(e.g. if you try the go game, you will understand where the compexity lies.)
2) A definite chess score is obtained in the same chess program in the 
evaluation function.  When comparing the chess score obtained with other 
chess programs or a human evaluation, scores may vary.  Only mate and 
technical draw positions are definite.
3) Before all rules (ep, 50moves, 3 fold-repetition) are implemented, it 
takes a lot of time and effort.  Effort that does not contribute to the 
algorithm development.

So in order to compare different algorithms, I designed a program with 
different algorithms and different games.  The only thing you have to do 
is to select the algorithm, and the game-ruleset.  Because of the 
complexity of a chess program, it is left out however.  But I did 
include checkers, connect 4, tictactoe, draughts, othello/reversi... 
etc.  I also implemented different algorithms and you can ask for a move 
to be generated with all algorithms in parallel.  You can easily compare 
the output from minimax/brute-force/alpha-beta/negascout.  It is called 
strategicgames.

 > I have already tried GNU Chess but the search routine is rather
 > complex because of various prunings, heuristics etc.

I can send you the source of my chess program named 'Belofte'.  I can 
assure you, it is much simpler than gnuchess.  You can test playing 
against it on freechess.org.  Finger belofte for more details.  Mail me 
for getting the source code.  www.internetcontact.be/~macchess

You can also download the source code of the strategic games.  It can be 
downloaded from users.telenet.be/artificialintelligence.

PS: Strategic Games does have some errors in the negascout and the 
move-generation for some checkers variants.... Feel free to correct.

Good luck,

Yves
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.