Rybka Chess Community Forum
Count me! I'm interested.
Yes i would love to test out the cluster software schach. im sure that would be interesting..
i will open a new thread the next days to inform you about the not-yet-named-one,
show you what i have a have already implemented into the project and the algorithms
behind it, and also to collect your additional ideas&contributions. Its currently optimized
for analyzing and playing longer time controls, blitz will be possible (needs some kind of
optimizations there though) but i doubt bullet would make to much sense :)...
As you know such a stub-approach has advantages and disadvantages over a native
version. Advantages for example are that you can use a tactical solver as a kind of
'preprocessor' (as known to compilers) and therefore overcome known "features" in
some engines like the bishop-underpromotions and others pretty easily.
btw., have you had a look into the intranet/remote possibilites of Aquarium (2010)?
I've used the remote feature of Aquarium, it's useful! I"m still having trouble getting used to it as my primary analysis tool.
I'd very interested in seeing what you have, and helping in any way! Thanks Alex!
that sounds like a very interesting approach. The idea is similar to GridChess (from Kai Himstedt). I'd love to test this program with my "mini-cluster" (4x4). Maybe you could send me a version to computerschach(at)gmx.de
Please carry on with this interesting project - computer chess urgently needs fresh ideas!
Thanks Timo and Brandon for your kind interest.
I read about Kai's GridChess-Ph.D. and it was one of the ideas behind my not-yet-named-one.
Is Kai member of the forum here?
I will open an own thread soon to show you what is already there and to collect your ideas&opinions.
I don't know if Kai is member here, but I guess he is reading the forum from time to time. I see him from time to time when some computer chess enthusiasts in the region of Hamburg join together to our "Stammtisch im Schachcafé".
Your "not-yet-named-one" reminds me also to the Gecko project by Heli Weigl. I haven't heard from it for a while now, so maybe he quit the work on it.
Did you fix the evaluation issues, when Rybka shows a winning score in a completely drawn position, where in a Rook and Pawn Endgame, it is up a pawn , but one of it 's rooks is in front of the pawn:
I mentioned this in the bug thread here:http://rybkaforum.net/cgi-bin/rybkaforum/topic_show.pl?pid=122204
You had replied:
Yes, this is something for later.
Is now, later?!
No! Now is now! Later is later! If there were going to be flowers in the windows, Vas would have said so!!!
There are flowers in the windows Rybka 4 is coming in May!
Flowers are overrated. But that bugfix is incredibly important for reliable analysis results.
Absolutely correct. Endgame evaluations were the most materially screwed-up part of R3. I can understand why Vas might make this a lower priority than relative strength but poor endgame evals are still a major irritant for any serious analyst. I think Vas (or was it Lukas?) showed us that some improvements had already been made there a year ago or more, so I am optimistic that we won't be totally disappointed.
It would be interesting to learn the sequential improvements Vas has made over the last 21 months. He may have changed hundreds of things, how fascinating to learn what he changed and what each incremental improvement was worth. Also what things failed to pan out and were set aside.
In part of the December interview that I edited out Vas described his programming technique a little (he was pretty vague on this, mostly). He said that chess programming is a matter of picking a particular area you want to improve, and in the process of fixing that you wind up working on a tangential problem, which leads to another and another. You might find yourself five tangents away from the original issue, but dealing with the tangents is how you solve the original issue. How I wish he could give a detailed example.
I don't see how this answer would worry you. What do you expect? "Everything about Rybka 3 is fixed and Rybka 4 is perfect in every way."? His answer was the obvious truth and he's not making any false promises.
> Kinda reminds me of people that buy Hiarcs and other of these engines.
The main reason could be for the playing style. Well, others are just anti-Rajlich :)
The release is good news for the Global Economic Situation, as the purchases of the Long Awaited release will stimulate the world economy. Vas should release the cluster version as part of the package to stimulate the world economy even more! We can all go out and buy more pc's to put together our own clusters to make our Rybka's stronger!!
I know this question might be a little out of place here, but are you planning an iPad version of Rybka/Aquarium? Another way to make things work on the iPad/iPhone would be to offer Rybka as a server based solution accessible via a browser. I have read about plans for this. Any comments?
Question on the new Rybka 4 Opening book by Jiri Dufik:
If the Chessbase version of the book can also be used with Aquarium, why bother making a Convekta version?
1) Superior format
2) Aquarium can't write to ctg books
Or the other way around.
I guess the "original" format is Aquarium and it was converted to ctg. but I don't know any details.
In what language Rybka 4 is written?
Vas has stated that it's really closer to C.
Closer to C?
But not C itself?
Is it a different language, which is less known?
The language is C++ but the code is closer to C than typical C++ code.
Vas has also said that the code is in hundreds of files and that other programmers would think him crazy if they ever saw it.
then it probably is crazy.
hopefully Vas used the past two years to do some housecleaning in his code, clean code = better performance and rules!
>hopefully Vas used the past two years to do some housecleaning in his code
Did you already began to test Rybka 4 beta?
It is too late.
Now is release time ! ;-)
I think Vasik said he developed his own language within c/c++ that helps him think better - and other developers would think he's crazy if they saw that. I don't recall it was the many-files that triggered that comment from him?
Possibly some pre-processor sub-language that helped him code chess conditions and masks bitboard operations or something?
Maybe some day he'll share some more as this sounds quite interesting to understand in concept, if not in detail.
> I think Vasik said he developed his own language within c/c++ that helps him think better - and other developers would think he's crazy if they saw that. I don't recall it was the many-files that triggered that comment from him?
He said both things in the same post, I think.
> Maybe some day he'll share some more as this sounds quite interesting to understand in concept, if not in detail.
Well, he said he would even reveal commented Rybka sources once he retires and moves to a different game, if the time comes.
That would be a great day for computer chess, considering that other already successful programmers (e.g. of Stockfish) could complement their (hopefully) different ideas with Vas' ideas and techniques.
Different ideas? Is that why I was running an engine tournament from a starting position yesterday, and Stockfish 1.7.1 was the ONLY "non-Rybka" engine that picked exactly the same first 12 moves as the Rybka engines against a certain opponent?
It actually really pissed me off--I had previously been critical of the idea that Stockfish is a Rybka derivative to any extent, but I am now starting to wonder if more than just an idea or two were taken from the Forsaken codes.
If that is your idea of a scientific analysis...Why would one even post such a null statement. If this were twelve tactical moves, they were possibly forced or typical computerlike moves. The only thing that shows is roughly equal tactical depth, and maybe the effect of playing a lot of games against Rybka. If these were all positional moves it was either coincidence or Rybka is a clone of Glaurung because the evaluation function has hardly changed at all since Glaurung 2.0. I think you can conclude absolutely nothing from just move choices from one starting position. The evaluation has not changed at all from 1.6.3 to 1.7.1, only quick_evaluate() is not used anymore. The source is open for everyone so everyone can see what changes have been made ever since Stockfish started.
The reason that it really pissed me off was because of the fact that I was using quite a number of different engines here, and I started it from a position that was no longer tactical, i.e. a transition from tactical to positional, where I wanted the opinions of different engines. Zappa, Fritz, Hiarcs, Glaurung, and Sjeng succeeded in this naturally because it was the type of situation where very slightly different evaluations yield quite different results. Rybka, Fire, and Stockfish were the ones who had the same moves. I also noticed it (though for not as many moves) in a sideline, too. Anyway, if I have a chance over the summer, if I notice that this seems to be a regular pattern, I'll check with a more scientific analysis to see about the possibility that Stockfish could actually be considered a clone of Rybka in the same "chess sense" as the known derivatives. Obviously, there are many types of positions where Stockfish gives very different results than Rybka, though these were ones where the great thing about using the different engines was that it should give a very wide tree of possibilities (and indeed it did with the other engines).
Stockfish a clone of Rybka. This observation is premature, but the method of data sure points that way.
When you think about it out of the blue sky all these new engines pop up and are stronger than programs
taken years and years to reach their current status, suggest that there is a rat in the wood pile. Sorry
about that Doc H.
> Stockfish a clone of Rybka.
I believe this statement is very unfair to the remarkable team behind Stockfish - and also very much inaccurate.
What is more fair to say, imho, is that at most they've learned some ideas from the Ippo* sources, and used them in the two recent Stockfish releases.
This does not make them derivatives nor clones, and Stockfish in an incredible purely original work by a selfless team that contributes their hobby time to the community in a remarkable way.
I would not consider it a clone at all. That is why I used the term "chess sense" because people are often misusing the term "clone". Obviously, it would be less of a derivative work than the Forsaken engines. The question comes to how much Forsaken information is it "okay" for a public engine to use "in good faith". Different people would certainly have different standards for this. I mention this because the real dope comes to the point of the term "derivative work" and whether such works are okay to use in tournaments, on the Chessbase server, etc. If the term is an absolute "no", then obviously Stockfish is bye-bye. Heck, for that matter, even Rybka could be considered bye-bye with the little code it has from Fruit that is known in such areas as the UCI implementation, though I think that most people consider that to be a non-issue. Thus, there must be a clear line drawn somewhere in the sand on this issue.
I highly doubt that Stockfish is a clone simply because it has different strengths and weaknesses than rybka. The fact that there is a native mac os version makes it pure genius too.
Why does the lack of Mac support make Stockfish pure genius?
No, that is not true. Stockfish is no Rybka clone, although it may contain ideas from Rybka, which are taken from the clones. But this is what any programmer will do, that's just the way it is now. Taking ideas is ok, I don't see why this should be wrong.
I see no reason to accuse the stockfish team for cloning, their engine is original to my knowledge.
I think looking at the sources would be a better method.
Sources (especially Rybka) are obfuscated, so that makes it more difficult except for the original programmers. It seems, though, that Vas doesn't have any complaints about Stockfish at this point.
> It seems, though, that Vas doesn't have any complaints about Stockfish at this point.
Because he's probably sitting at his desk cackling watching Rybka 4 easily defeat all the engines he's testing it against. We did wait two years...
Yes, but he had problems with the Forsaken engines, and stated such. Depending on who is doing the testing, Stockfish 1.7.1 might be just as strong overall.
I mean comparison of the derivatives' sources with Stockfish's sources, I bet several people jumped to compare them when the big improvement news came in and I've seen 0 accusations.
The only thing I've seen is your 12 moves claim, can you share these moves so we take a look? It could have been pure luck.
I can share them at a later point, but not now--they are from a sideline in my game against Alex. Also, don't consider this an accusation--as you said, it could have simply been luck, and that looking at many positions, one will eventually get many engines doing something like that. As I said, if I have time later this summer (which I doubt), I can check this out in more detail if things are looking more suspicious.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill