1
2
Rybka 3
August 6, 2008
Overview
Rybka 3 is the latest UCI chess engine from the Rybka line. Our previous version, Rybka 2.3.2a, is now nearly 14 months old, so the changes are extensive.
Major Changes
- The evaluation function is more accurate. There is a long list of useful new evaluation terms. The terms also interact in a more balanced and mature way.
- The search is more efficient, especially in the sorts of sharp, murky positions in which chess games are often decided.
- The multi-processor support is improved.
- The code and algorithms have been cleaned up, leading to more solid and predictable behavior.
- The opening book is improved and expanded and now fills the role of a general reference. Note: this book is sold separately, details here.
New Features
- Persistent hash, for storing and loading previous analysis
- More efficient multi-variation mode
- Chess 960 support
- More intelligent contempt, which takes into account the trading of material, symmetrical structures and other drawish features
- A special "human" version, whose evaluation aims at consistency with human grandmaster practice rather than at success in computer vs computer play
- A special "dynamic" version, whose evaluation aims at speculative play
There are also a number of other new features which enhance the analytical process but which require interface support. These can be used in the ChessBase and Convekta interfaces which are sold with Rybka 3. Details can be found in the documentation for those interfaces.
User Area
All information relevant to Rybka 3 usage is collected here. This includes parameter descriptions, FAQ, installation instructions, known engine issues, etc.
Credits
A huge thanks to everyone on the Rybka team, without whom this project would be much less successful and a lot less fun:
- Larry Kaufman for his incredibly accurate work with Rybka's static evaluator (and for not letting me forget about Monte Carlo)
- Jeroen Noomen for again finding a way to push the boundaries in opening book quality and for his never-satisfied approach
- Iweta for Rybka testing, ideas and moral support :)
- Felix Kling and Christoph Kling for an excellent and innovative web site
- Lukas Cimiotti for always having an 8-core machine or three ready for testing, tournament play or analysis (by patzers and non-patzers alike)
- Hans van der Zijden for traveling around the world (or at least North-Western Europe) as our tournament operator
- Nick Carlin for ICC play, opening preparation and testing help
- Dagh Nielsen for contributing to Jeroen's opening book while entertaining the rest of us with his unique working style
- Victor Zakharov and his team from Convekta and Mathias Feist from ChessBase for their enthusiastic interface work
- Rybka testers for not letting the obvious perfection of the code discourage them
- Rybka forum regulars for a steady stream of often-interesting and occasionally-practical ideas
What's Next?
We will consider a number of Rybka ports to other platforms in the next few months. Aside from that, we'll go back into our cave and in the summer of 2009 we'll emerge with Rybka 4.
August 6, 2008
Overview
Rybka 3 is the latest UCI chess engine from the Rybka line. Our previous version, Rybka 2.3.2a, is now nearly 14 months old, so the changes are extensive.
Major Changes
- The evaluation function is more accurate. There is a long list of useful new evaluation terms. The terms also interact in a more balanced and mature way.
- The search is more efficient, especially in the sorts of sharp, murky positions in which chess games are often decided.
- The multi-processor support is improved.
- The code and algorithms have been cleaned up, leading to more solid and predictable behavior.
- The opening book is improved and expanded and now fills the role of a general reference. Note: this book is sold separately, details here.
New Features
- Persistent hash, for storing and loading previous analysis
- More efficient multi-variation mode
- Chess 960 support
- More intelligent contempt, which takes into account the trading of material, symmetrical structures and other drawish features
- A special "human" version, whose evaluation aims at consistency with human grandmaster practice rather than at success in computer vs computer play
- A special "dynamic" version, whose evaluation aims at speculative play
There are also a number of other new features which enhance the analytical process but which require interface support. These can be used in the ChessBase and Convekta interfaces which are sold with Rybka 3. Details can be found in the documentation for those interfaces.
User Area
All information relevant to Rybka 3 usage is collected here. This includes parameter descriptions, FAQ, installation instructions, known engine issues, etc.
Credits
A huge thanks to everyone on the Rybka team, without whom this project would be much less successful and a lot less fun:
- Larry Kaufman for his incredibly accurate work with Rybka's static evaluator (and for not letting me forget about Monte Carlo)
- Jeroen Noomen for again finding a way to push the boundaries in opening book quality and for his never-satisfied approach
- Iweta for Rybka testing, ideas and moral support :)
- Felix Kling and Christoph Kling for an excellent and innovative web site
- Lukas Cimiotti for always having an 8-core machine or three ready for testing, tournament play or analysis (by patzers and non-patzers alike)
- Hans van der Zijden for traveling around the world (or at least North-Western Europe) as our tournament operator
- Nick Carlin for ICC play, opening preparation and testing help
- Dagh Nielsen for contributing to Jeroen's opening book while entertaining the rest of us with his unique working style
- Victor Zakharov and his team from Convekta and Mathias Feist from ChessBase for their enthusiastic interface work
- Rybka testers for not letting the obvious perfection of the code discourage them
- Rybka forum regulars for a steady stream of often-interesting and occasionally-practical ideas
What's Next?
We will consider a number of Rybka ports to other platforms in the next few months. Aside from that, we'll go back into our cave and in the summer of 2009 we'll emerge with Rybka 4.
lets see
Excellent! I was going to wait for the OS X version but I couldn't resist buying everything now. ;)
Back into the cave! I thought you guys were knuckle draggers.haha
CONGRATULATIONS on a fine product,one that changes the landscape of computer chess!
CONGRATULATIONS on a fine product,one that changes the landscape of computer chess!
and in the summer of 2009 we'll emerge with Rybka 4.
He Vas, don´t let us speak about summer 2009 but autumn. I think, summer is the worst time to sell engines. I would like to heat my rooms with Rybka 4 in october 2009. :-)
He Vas, don´t let us speak about summer 2009 but autumn. I think, summer is the worst time to sell engines. I would like to heat my rooms with Rybka 4 in october 2009. :-)
Or: Let's not speak about Rybka 4 before we received Rybka 3 ;)
I think this version of Rybka should have been called Rybka 4!
no kidding
Or Fritz 12 ;)
Fritz 12 will be nowhere near as strong as this version. I think that if they are lucky, Fritz 12 might catch Rybka 2.3.2a.
Vas will give the Fritz guys some tips as to make it stronger. So, maybe Fritz 12 will be much stronger than 2.3.2a :)
What I meant is: now that Chessbase distributes Rybka, why not just rename it to Fritz 12?
Chessbase should aim for a one ,two punch by making Fritz the number two engine with Vas's help.But not number one!
Ahh, I see...interesting--though Convekta is still basically the primary distributor.
What primary distributor means?
I see that not all components of Rybka 3 are sold by Convekta(Vas said that the Rybka 3 book is sold separately, but it's only available from Chessbase) while the opposite it's true for Chessbase as it sells all Rybka's components.
If you consider Aquarium as Rybka's part then of course the opposite isn't true in this case also, since Chessbase doesn't sell Aquarium, but i can't find a reason why we should consider Aquarium as part of Rybka 3 and also Vas has not mentioned anything about Aquarium in his lasts reports/FAQs about Rybka 3.
I see that not all components of Rybka 3 are sold by Convekta(Vas said that the Rybka 3 book is sold separately, but it's only available from Chessbase) while the opposite it's true for Chessbase as it sells all Rybka's components.
If you consider Aquarium as Rybka's part then of course the opposite isn't true in this case also, since Chessbase doesn't sell Aquarium, but i can't find a reason why we should consider Aquarium as part of Rybka 3 and also Vas has not mentioned anything about Aquarium in his lasts reports/FAQs about Rybka 3.
> Ahh, I see...interesting--though Convekta is still basically the primary distributor.
Well, they named Pocket Hiarcs "Pocket Fritz 3" and got away with it, it wouldn't surprise me if Rybka 4 powers Fritz 12 and they don't mention Rybka anywhere.
> Well, they named Pocket Hiarcs "Pocket Fritz 3" ...
Exactly. If they rename Hiarcs for marketing reasons, who is going to stop them from renaming Rybka?
I hope that Vas is not going to agree to something like that.
I see no reason to rename everything that chessbase is selling to Fritz.
There are other programs in the world and I see no reason to call every program in the name Fritz.
Uri
I see no reason to rename everything that chessbase is selling to Fritz.
There are other programs in the world and I see no reason to call every program in the name Fritz.
Uri
Either you don't believe branding can be successful, or you don't believe Pocket Fritz is a successful brand.
If market research indicated that Chessbase would sell twice as many copies if they called it Pocket Fritz rather than Pocket Fish, it would be a business decision for CB and Vas, just as it was for CB and Stephan, and CB and Mark for previous versions.
Alan
If market research indicated that Chessbase would sell twice as many copies if they called it Pocket Fritz rather than Pocket Fish, it would be a business decision for CB and Vas, just as it was for CB and Stephan, and CB and Mark for previous versions.
Alan
>Either you don't believe branding can be successful, or you don't believe Pocket Fritz is a successful brand.
Maybe this would make good business sense, but my objection is that it would give the false impression that Vas was not the programmer. In my mind, it comes close to false advertising.
Regards,
Milton
Well, the term Pocket Fritz could be said to refer mainly to the GUI, just as we currently talk about the Fritz GUI. Most users are probably oblivious to what's under the hood. Please note that Pocket Fritz is currently based on Hiarcs and before that it was based on Shredder.
Regards,
Alan
Regards,
Alan
rather than Pocket Fish
Be sure, noone will buy it! :-) But why not Pocket Rybka or Pocket Hiarcs?
Be sure, noone will buy it! :-) But why not Pocket Rybka or Pocket Hiarcs?
> rather than Pocket Fish
>
> Be sure, noone will buy it! :-)
You might be surprised :)
Momentous occasion! Very well done indeed ;)
I'm very pleased to see the improved multi-variation mode -- it's going to be very useful for live analysis of human games, especially for us patzers (answering questions like "uh, why on Earth is he not taking the queen?").
/* Steinar */
/* Steinar */
I see only this "- More efficient multi-variation mode".
How do you know what changes there are and how different it will be than the current way of how multi-variation works and how do you know that with the new way it will be easier to understand the reasons of the moves on the board?
Also does this new multi-variation work on all GUIs(Arena,Shredder Classic) or on Aquarium only or on Chessbase GUI that comes with Rybka 3 only?
How do you know what changes there are and how different it will be than the current way of how multi-variation works and how do you know that with the new way it will be easier to understand the reasons of the moves on the board?
Also does this new multi-variation work on all GUIs(Arena,Shredder Classic) or on Aquarium only or on Chessbase GUI that comes with Rybka 3 only?
The option was mentioned in a previous post; specifically, you can ask Rybka not to calculate non-primary variations further than usual, unless they're within some given number of centipawns (and I'd assume you can set that to 0 or 1).
I don't know if it works with a given GUI or not; I don't use a GUI for Rybka. From my understanding of it all it does is change how much time is spent on each variation, though, so I'd assume it should work in any GUI that can set normal UCI options.
/* Steinar */
I don't know if it works with a given GUI or not; I don't use a GUI for Rybka. From my understanding of it all it does is change how much time is spent on each variation, though, so I'd assume it should work in any GUI that can set normal UCI options.
/* Steinar */
Congratulations on what should be a huge success commercially, in terms of theory and overall prestige. Well deserved. Now, brace yourself for the deluge of criticisms from idlers! How many times will you hear "Rybka is a great product, BUT..." and "[New Engine] released - Rybka in danger?" Finally you will have to get used to, "Rybka is so dominant it is BORING" and "Rybka is the death of chess." In fact all of these imagined quotes are perfect subject headers for you idlers out there!
A better subject header is "Rybka 3 is an idiot, it can't solve this position. Program-X can solve it."
Program-X of course is about 150-200 ELO weaker than R3.
I really hope that Rybka 3 is improved in best move testpositions, since i'm tired of reading all these Rybka-can't-find-this-move threads.....
I will not escape from hearing all the Rybka-is-bad-at-endgames threads though.... :-(
Program-X of course is about 150-200 ELO weaker than R3.
I really hope that Rybka 3 is improved in best move testpositions, since i'm tired of reading all these Rybka-can't-find-this-move threads.....
I will not escape from hearing all the Rybka-is-bad-at-endgames threads though.... :-(
Good points. The problem with endgames is that there are general principles that all engines should have down pat, which I think Rybka mostly will have, but there are always special cases that some engines can immediately solve (counterintuitively) and others, apparently much stronger, cannot. All you can do to neutralize the nit-pickers is present them favorable results from a large test suite, I suppose. I'm sure someone will be coming out with that data soon enough.
IMHO most people's aim, when posting positions like that here, is not to mock rybka, but to show Rybka developers where she can be improved.
I may be wrong, of course ;)
I may be wrong, of course ;)
It would possibly be in Vas' best interests to help other programs improve. One thing which drives sales, is the desire to have the strongest program. If improvements to Rybka just make an already strongest program stronger, surely sales will decline, but, if another program, after a suitable period of time, surpasses Rybka, and then a new, improved, stronger Rybka is released, people will flock to it.
Just a thought, but based on my own experiences over the years. I haven't bought a new release in years, as there doesn't seem to be much point.
Just a thought, but based on my own experiences over the years. I haven't bought a new release in years, as there doesn't seem to be much point.
If the desire to have the best program drives sales. Then, why won't the release of a significantly stronger version of the currently best program drive sales. (It's not necessary to be better than another program - it's OK to be better than it's previous version, as long as the improvement is significant enough.)
I agree that competition may create additional attention. But, I believe Rybka 3 will be very successful - even though Rybka 2 is already the strongest chess program. Likewise, Rybka 4 can be very successful - even if Rybka 3 is still the best chess program.
I agree that competition may create additional attention. But, I believe Rybka 3 will be very successful - even though Rybka 2 is already the strongest chess program. Likewise, Rybka 4 can be very successful - even if Rybka 3 is still the best chess program.
Credits
You forgot to mention the most important person: yourself :-). Thanks for building the strongest engine ever, I am sure August 6th 2008 will be a memorable day in the history of computer chess, as was December 5th 2006!
Kind regards, Jeroen
You forgot to mention the most important person: yourself :-). Thanks for building the strongest engine ever, I am sure August 6th 2008 will be a memorable day in the history of computer chess, as was December 5th 2006!
Kind regards, Jeroen
Hear, hear!
Definitely.
Well said Jeroen!
and he forgot the muse that inspires him ;)
Woops, time flies when you are having fun! Of course the second date has to be December 5th 2005 instead of 2006 :-)
Thanks! I was really wondering what I might´ve missed :)
Perhaps this was the real date of the release of the fish called Rybka.... But I recall very well that the first beta version was offered for a free download at December 5th. It was 'Sinterklaas evening' when she entered my PC :-). And from that day the Rybka boom began.
You left out "improved time management" from the list of major changes. Maybe you didn't consider it "major", but it is very much improved.
This would have been a major blemish had it not been corrected. It's great this got done. Kudos.
Rybka has revolutionized the world of computer chess, thanks that you have created this marvel.
The special thanks are for you and ONLY YOU Vas.
Best Regards,
Gaмßito.
The special thanks are for you and ONLY YOU Vas.
Best Regards,
Gaмßito.
Hey everyone i purchased rybka 3 anyways it has a different file different than the previous versions that used exe at the end example
rybka 3.2.a.exe like that this new file is completely different so my shredder interface wont allow me to add it to be able to use the
program any reasons why what can i do about it or do i need to just buy the cd when it comes out to chessbase instead?
rybka 3.2.a.exe like that this new file is completely different so my shredder interface wont allow me to add it to be able to use the
program any reasons why what can i do about it or do i need to just buy the cd when it comes out to chessbase instead?
After you extract Rybka 3 from your archive, it will end in .exe, just like any other engine.
Vas
Vas
Rybka is another chess engine like Fritz or Hiarcs and Junior. I think Psion Chess beat Rybka. and Psion Chess is very old dos chess program. I love my Rybka 2.3.2a.mp.exe But my master plan is beat Rybka byte by byte. I wanna make my own chess engine to beat Rybka byte by byte.
I think Rybka code is similar to Crafty code. Crafty is not UCI engine but play well chess.
1
2
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill