I know something of Dr.Donninger

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
DarkSide
Member
Posts: 19
Joined: Thu Sep 27, 2007 3:19 pm

Post by DarkSide »

Hi Dr.
Sorry ,About Chess-Advice-Language Che++) it has been developed by you? i found almost nothing in the web, and I have interest in knowing more of, I do not know why it has not been used by others programmers given its virtues, like you can add knowledge without reducing the speed of search, and so on.. I hope you can talk a little more about this topic Dr.
Thanks for you time
Regards :!:
chrilly
Member
Posts: 5
Joined: Wed Oct 10, 2007 1:01 pm

Post by chrilly »

DarkSide wrote:Hi Dr.
Sorry ,About Chess-Advice-Language Che++) it has been developed by you? i found almost nothing in the web, and I have interest in knowing more of, I do not know why it has not been used by others programmers given its virtues, like you can add knowledge without reducing the speed of search, and so on.. I hope you can talk a little more about this topic Dr.
Thanks for you time
Regards :!:
Yes. LIke every "real programmer" I like to develop languages. E.g. I have also developed a scripting language for the multimedia chess programm "Der Schweinehund". And I have made a 16-Bit FPGA processor...
Che was intended only for internal use. Andreas Mader has written a Che++ manual. One could encode chess knowledge in a Basic-like language with some additional chess functions and terms. The code was translated to an internal representation, similar to the Java virtual machine (I designed Che around the same time Java was developed. The idea of making a virtual machine was well known).
This code was primarily processed in the so called Oracle. The Oracle sets parameters of the evaluation and of the search. This is done only once per move at the root-position. As it is done only once the overhead of using an interpreted code does not matter. There is practically no time penalty.
The success was mixed. Some Che enthusiast claimed that they have improved the programm with the Che rules by about 50 percent. I stoped the development after the "Seirawan-desaster" at the Aegon tournament in ????
Nimzo was so far the best ranked programm in the tournament and played on the center-court against GM Seirawan. The Che code had the reasonable rule: If you have not yet castled and the pawns on one side are already pushed forward, keep the pawns on the other side back. If the opponent played e.g. Bg5, Nimzo had the tendency to play h4 and after Bh4, g5.
Seirawan played Bg5, Nimzo wanted to play its favorite moves (which were correct in this position), but the Che-rules said: You have already played b5, h6 followed by g5 is bad. Unfortunately the rule was only specified for the root-position. So the search invented some "Null-Move" and calculated in the variation the favorite line. Nimzo played the "Null-Move", Seirawan played another move, the bishop was still there, the same reasoning repeated. Actually it repeated for a dozen of times, Nimzo played one nonsense move after another, the crowd gathered around the centre-court and cheered each move. Nobody - besides myself - could explain why a programm which has beaten before GMs plays moves no 1200 Elo player would have played. The absolute nightmare for every programmer.
The problem was, that with Che one created 2 knowledge bases. The internal hard coded knowledge and the Che rules. The search is usually very tricky to find the inconsistency between this 2 rule-sets.

Chrilly
DarkSide
Member
Posts: 19
Joined: Thu Sep 27, 2007 3:19 pm

Post by DarkSide »

Hi Dr.
Doctor as you surely know, although since you has said previously that you don´t like chess and don´t pay attention to it, from end of year 2005 to today many people thinks that Rybka, that this in its version 2.3.à and that it has elo of 3100 in 4 processors can beat Hydra.
Could you give us an opinion about Rybka?
Do you have run some test of Hydra vs Rybka(or other engines)?
If so, what was the result?
Again I want to be thankful to you to answer amiably my questions.
Regards
User avatar
Terry McCracken
Senior Member
Posts: 1300
Joined: Tue Jul 31, 2007 5:49 pm

Post by Terry McCracken »

DarkSide wrote:Hi Dr.
Doctor as you surely know, although since you has said previously that you don´t like chess and don´t pay attention to it, from end of year 2005 to today many people thinks that Rybka, that this in its version 2.3.à and that it has elo of 3100 in 4 processors can beat Hydra.
Could you give us an opinion about Rybka?
Do you have run some test of Hydra vs Rybka(or other engines)?
If so, what was the result?
Again I want to be thankful to you to answer amiably my questions.
Regards
Donninger examined Rybkas code after it made a big sensation in CCC.

If he found anything worthwhile he may have looked into this further, I simply don't know.

In any case, could we please forget this guy? Hydra is dormant with no signs of waking up anytime soon.

Terry
User avatar
Harvey Williamson
Site Admin
Posts: 6079
Joined: Sun Jul 29, 2007 6:57 am
Location: Media City, UK
Contact:

Post by Harvey Williamson »

DarkSide wrote:Hi Dr.
Doctor as you surely know, although since you has said previously that you don´t like chess and don´t pay attention to it, from end of year 2005 to today many people thinks that Rybka, that this in its version 2.3.à and that it has elo of 3100 in 4 processors can beat Hydra.
Could you give us an opinion about Rybka?
Do you have run some test of Hydra vs Rybka(or other engines)?
If so, what was the result?
Again I want to be thankful to you to answer amiably my questions.
Regards
The post below is quoted with the kind permission of Chrilly and Enir(Enrique) and first published on the Rebel forum.

Best wishes,

Harvey
Enir wrote:I asked Chrilly Donninger a couple of questions about Hydra and Rybka. These are his answers, that I post here with his permission. Chrilly asks me to "please stress, that its just a hypothesis and no theory." I ask everyone not to reproduce Dr. Donninger's statements below in any form anywhere else.

To my question "What do you think it would be the result Hydra-Rybka in a long match, with Rybka running on a fast quad, for example?", Chrilly answered:

"No idea. I visited last Saturday my old companion Heli Weigel in Vienna. He has build an own parallel system from off the shelf programms called Gecko.
Gecko used Rybka, Zappa and Naum as its basic engines. Rybka was clear, but why on earth Zappa and Naum instead of Fritz and Shredder. Heli informed me, that Zappa and Naum are better. I have heared at least from Zappa, but Naum was completly new to me. I am out of business. The only relation I have to computer chess in the moment are the monthly Hydra-$.
Hydra runs currently on some outdated FPGAs. They are 2 generations back. If Hydra makes a comeback into the computer-chess world we would have to make a new system. I think this system would be considerable stronger than the old one. I have learned some lessons in FPGA design and the new FPGA-chips are much more powerfull. Progress is very fast. Both in hardware, and in computer-chess."

My next question was "what the hell is the secret of Rybka? I find it hard to believe that its superiority is due to a better positional play. Better search once again?". Chrilly answered:

"I think its simple. Rybka has a big table with exchange relations of pieces.
E.g. in the simplest case: A bonus for the bishop-pair. But there are numerous other relations e.g. in the endgame.
Otherwise its basically Fruit. The programm is due to its simplicity very fast. The node count is a fake. They divide the number before sending it to the screen. Same holds for the search depth. The internal search depth is much deeper than the show on the screen. They cut off the principial varation. I have read many arguments, Rybka must be - due to its low node count - very intelligent. This intelligence is in the eye of the observer.
GM Lutz was never impressed by the chess knowledge of Rybka.
They have - in addition to Fruit - a threat extension in the search. If a move threats the opponent and the opponent answers the threat, the search is extended. Threat is defined very general as "the evaluation goes up by a certain amount". To answer the threat means: The evaluation goes down again to the original level.
Its a good mixture of simple ideas.

How to I know? There was a post on a Russion forum by a hacker who dissassembled Rybka. Everybody sayed, he just invents this and its not true.
But I have also disassembled some parts of the first Rybka Version. My findings are in aggreement with the statements of these guy. Maybe they have changed a lot in the meantime and I am speaking of a long outdated Rybka (see Zappa, Naum above). But usually one just adds a few things once one has a successfull version. I think there is no reason to write it completly new."

Unfortunately, I don't think that Chrilly Donninger will post here any time soon. He finds it fun to post in forums once every 2 years or so, but no more. So don't expect he will answer or argue.

Enrique
DarkSide
Member
Posts: 19
Joined: Thu Sep 27, 2007 3:19 pm

Post by DarkSide »

Hi
" :shock: Heli informed me, that Zappa and Naum are better" It will a Heart break for Rybka fan´s :wink: I wonder how many time will happen before the other chess programmers (Mark, Amir, Frans etc) catch Rybka?
I think the Rybka secret is now partially solved. If it is true ( I mean the famous table of exchanges) what dificult is for the rest do a similar table? Hiarcs has a better chess knowledge, Fritz Zappa or Naum has better chess knowledge too, so I´m sure that the dominion of rybka will be end very soon, .... I hope :)
Thanks Harvey
Regards
Enir
Member
Posts: 67
Joined: Fri Aug 31, 2007 6:04 pm
Location: Cadaqués, Mediterranean

Post by Enir »

DarkSide wrote:Hi
" :shock: Heli informed me, that Zappa and Naum are better" It will a Heart break for Rybka fan´s :wink:
Why? Heli said better than Fritz or Shredder, not better than Rybka, no? so Rybka fans can sleep peacefully. :)

Enrique
DarkSide wrote:I wonder how many time will happen before the other chess programmers (Mark, Amir, Frans etc) catch Rybka?
I think the Rybka secret is now partially solved. If it is true ( I mean the famous table of exchanges) what dificult is for the rest do a similar table? Hiarcs has a better chess knowledge, Fritz Zappa or Naum has better chess knowledge too, so I´m sure that the dominion of rybka will be end very soon, .... I hope :)
Thanks Harvey
Regards
Post Reply