Moderators: hgm, Rebel, chrisw
You have just shown that it's actually you who has no clue about programming.Zerbinati wrote:but I do not think you have shown that you understand anything about programming.
The options let to limit strength of the engine with tactic and strategy.
https://github.com/Zerbinati/SugaR/comm ... 17a43b7033
I specified I'm not the author of the patch.mar wrote:You have just shown that it's actually you who has no clue about programming.Zerbinati wrote:but I do not think you have shown that you understand anything about programming.
The options let to limit strength of the engine with tactic and strategy.
https://github.com/Zerbinati/SugaR/comm ... 17a43b7033
Probing a red-black tree several times during eval?
Well congratulations because you just made your cheap clone slower
Interesting project, Andrea. I do have the Shashin book, and I once even made a suggestion in these forums that its Tal-Petrosian-Capablanca 'algorithms' could be very well used in a chess engine's eval.amchess wrote:ShashChess isn't at all an April Fool.
I worked on Shashin's theory actively since 2011, based on various internet chess forums.
In particular, based on it, I realized the Vitruvius chess engine opening book (http://www.vitruviuschess.com/article/T ... uvius.html)
When the related book came out (https://www.amazon.com/Best-Play-Method ... 1936277468)
I began working hard also on implementing this on a real chess engine.
To verify this work, I choosed an engine derived from various open source code.
My part is original and meaningful as proved by
- a similar test (well less than the crucial 60%)
- a match against the latest best Stockfish, on 80 games based on a very good test suite based on all center types. The match was won by 7-0 in blitz (5 min) mode
- a test suite of very hard positions: ShashChess solved 282/438 of them. Based on my information, no other known engine can do that.
This last result can be improved using the algorithm options and position types.
"Tal, Capablanca and Petrosian" are in fact the very kernel of Shashin theory. If not, this is a simple non-sense.
The engine dynamically changes his search and evaluation based on the position type.
With the options, you can force them.
To determine the position type and to have an analysis in natural language of a position according to Shashin's theory, I also developed an AI NLG application (a virtual trainer).
I can't render all open source because of the commercial contract with my old and yet current publisher.
So, the engine is private.
If someone is interested, he can contact me privately: the Shashin code can improve all engines and I will be very happy to collaborate with someone.
Andrea Manzo
I'm not so expert of all that. I need to talk to my publisher.carldaman wrote:Interesting project, Andrea. I do have the Shashin book, and I once even made a suggestion in these forums that its Tal-Petrosian-Capablanca 'algorithms' could be very well used in a chess engine's eval.amchess wrote:ShashChess isn't at all an April Fool.
I worked on Shashin's theory actively since 2011, based on various internet chess forums.
In particular, based on it, I realized the Vitruvius chess engine opening book (http://www.vitruviuschess.com/article/T ... uvius.html)
When the related book came out (https://www.amazon.com/Best-Play-Method ... 1936277468)
I began working hard also on implementing this on a real chess engine.
To verify this work, I choosed an engine derived from various open source code.
My part is original and meaningful as proved by
- a similar test (well less than the crucial 60%)
- a match against the latest best Stockfish, on 80 games based on a very good test suite based on all center types. The match was won by 7-0 in blitz (5 min) mode
- a test suite of very hard positions: ShashChess solved 282/438 of them. Based on my information, no other known engine can do that.
This last result can be improved using the algorithm options and position types.
"Tal, Capablanca and Petrosian" are in fact the very kernel of Shashin theory. If not, this is a simple non-sense.
The engine dynamically changes his search and evaluation based on the position type.
With the options, you can force them.
To determine the position type and to have an analysis in natural language of a position according to Shashin's theory, I also developed an AI NLG application (a virtual trainer).
I can't render all open source because of the commercial contract with my old and yet current publisher.
So, the engine is private.
If someone is interested, he can contact me privately: the Shashin code can improve all engines and I will be very happy to collaborate with someone.
Andrea Manzo
You mentioned that the program can't be released as open source, but what about closed source? Do you intend to release such a version at some point, whether free or commercial?
Thanks,
CL
