New engine: Rhetoric

[asanjuan]

Hello,
I would like to present the first release of my own engine: Rhetoric. Not a very strong engine, because is in the 2200 - 2300 elo range, but is surprisingly strong playing against humans.

But what makes Rhetoric special is the fact that it has been trained with Genetic Algorithms trying to fit grandmaster moves from a collection of GM Karpov games against other elite GMs.

The web site is in spanish, i hope to update it with new sections in english in the near future.
For the moment you can download the Win32 version of Rhetoric clicking in the "Descarga ya" button placed in the home page.

http://www.chessrhetoric.com

Regards from Seville, Spain.
Alberto Sanjuan

[bpfliegel]

Sounds very interesting - I was wondering how GAs work in a discreet problem space. Is it public how it internally works - how the genotype looks like, mutations, crossovers, etc?

Good luck with it!
Cheers, Balint

[Andres Valverde]

Enhorabuena !

[velmarin]

Muchas gracias , probando ya¡¡.

[asanjuan]

Sounds very interesting - I was wondering how GAs work in a discreet problem space. Is it public how it internally works - how the genotype looks like, mutations, crossovers, etc?

Rhetoric is based on this article. You can download it from my site:
http://www.chessrhetoric.com/index.php/ ... bliografia
I don't know who wrote it, but is very very interesting.
After reading, i implemented a simplified version of this methodology using arrays of integers as adn secuences, and uniform corossover. I don't plan to publish this code because is the only thing that makes Rhetoric so different.

[Graham Banks]

Hello,
I would like to present the first release of my own engine: Rhetoric. Not a very strong engine, because is in the 2200 - 2300 elo range, but is surprisingly strong playing against humans.

But what makes Rhetoric special is the fact that it has been trained with Genetic Algorithms trying to fit grandmaster moves from a collection of GM Karpov games against other elite GMs.

The web site is in spanish, i hope to update it with new sections in english in the near future.
For the moment you can download the Win32 version of Rhetoric clicking in the "Descarga ya" button placed in the home page.

http://www.chessrhetoric.com

Regards from Seville, Spain.
Alberto Sanjuan

Thanks Alberto.

[tpetzke]

Hola,

congratulations to your engine. Glad to see that GA seem to work somehow despite many people telling that GAs are the 2nd worst thing in chess programming.

Buena suerte
Thomas...

[asanjuan]

It still loses against Eveann, but as we say: "ya tenía ganas de publicarlo".

[Ajedrecista]

Hi Alberto:

Sounds very interesting - I was wondering how GAs work in a discreet problem space. Is it public how it internally works - how the genotype looks like, mutations, crossovers, etc?

Rhetoric is based on this article. You can download it from my site:
http://www.chessrhetoric.com/index.php/ ... bliografia
I don't know who wrote it, but is very very interesting.
After reading, i implemented a simplified version of this methodology using arrays of integers as adn secuences, and uniform corossover. I don't plan to publish this code because is the only thing that makes Rhetoric so different.

When I saw this paper, the overall view remembered me a lot another paper of Omid David-Tabibi (Falcon engine) that Ruxy Sylwyka posted in TalkChess three months ago; searching in my posts I found one where I linked to the original post. I give the link of that paper:

Expert-Driven Genetic Algorithms for Simulating Evaluation Functions.

If you read it carefully, I am sure that there are some similar things that appear in the paper you have in your website. So, I think that Simulating Human Grandmasters: Evolution and Coevolution of Evaluation Functions is also by Omid David-Tabibi.

I hope that this paper will be interesting for you and will help you in improving Rethoric. I am amazed of the complex algorithms that people are able to implement. ¡Buena suerte en el futuro!

Regards from Spain.

Ajedrecista.

[Andres Valverde]

It still loses against Eveann, but as we say: "ya tenía ganas de publicarlo".

. Well done!