In the Glarean Magazine you can read a new Interview with Vasik Rajlich (Rybka) in german and english:
GM: What is your main work to be done with Rybka right now?
VR: Right now I am in «release mode», there are a ton of little things to do now.
GM: What innovations may we expect from Rybka 4?
VR: The eval and search are revamped, Ichanged it completely three times last year. There will also be a few new analysis features.
GM: When approximately may we expect Rybka 4?
VR: This is still TBD.
GM: Will it be sold and distributed in the same ways as formerly?
VR: Yes, Convekta and ChessBase will be the publishers.
GM: Has there been any change in the team of developers?
VR: The development work is still done only by me, but we have a great team. Lukas Cimiotti has helped tremendously with the clustering and with tournament preparations, his contribution over the past year has been enormous. Our book team has changed a bit, we have added Jiri Dufek, while Jeroen Noomen remains involved. We will give more details later. I also should mention Felix Kling and his brother Christoph for their work on our web site, Hans van der Zijden as the Rybka operator, my wife Iweta for testing, Larry Kaufman for man-vs-machine matches and parameter tuning, and Nick Carlin for book work and tournament preparations.
GM: Will we have a public Rybka cluster-version sooner or later?
VR: This won’t be a part of the Rybka 4 release, but we do have plans for this in addition to competing in tournaments. It will be something a little different, you’ll have to stay tuned.
GM: Maybe you remember me asking you once at Rybka forum about multi-variant-mode of analysis. My special wish of automatic backward analysis in mv- mode was answered by you then as a matter of GUI, which it is, of course. But would you support it as useful, even sometimes in normal game mode, especially as for cluster version?
VR: Can you say what you mean by „backward analysis“?
GM: I just meant the feature of some GUIs to step back automatically in the game analyzed.
VR: The output of our cluster is a tricky issue which needs some thinking. Outputting a single PV is a poor fit to how the cluster searches. In principle, the cluster could provide a sort of multi-pv analysis even in its more efficient „game-play“ mode. So far we have only used the cluster for competitions, so this issue has not been resolved.
GM: Will the „find win“ mode be new too in Rybka 4 or will even a new WinFinder come up again?
VR: Making more interesting versions (including some WinFinders) is on my to-do list, but I haven’t touched it since Rybka 3. Ditto for „find win“ mode – it could be improved, but so far hasn’t been. Rybka herself is much stronger tactically now than Rybka 3.
GM: Or do you think nullmove to be switched off as an option of the engine would also be a feature worth adding?
VR: This feature probably doesn’t have enough value to add to the parameter list.
GM: Any news to be expected as for shared analysis and persistent hash?
VR: Not yet, so far there are only some bug fixes. This is another topic which will eventually get a lot of work.
> Rybka herself is much stronger tactically now than Rybka 3.
>> Rybka herself is much stronger tactically now than Rybka 3.
That's Vas for you :)
> I can see every other chess programmer out there roll their eyes. Like Rybka 3 is not enough of a tactical beast :-)
Are you rubbing salt into the wound? :)
But the tactical issue isn´t the only issue R3 lacks! I know at least one more! :-)
But I can give you an issue, where Rybka has to become better! Bad luck: Rybka avoids this issue by winning before this issue becomes relevant! :-)
So I´m waiting for better opponents for Rybka!
> I can see every other chess programmer out there roll their eyes. Like Rybka 3 is not enough of a tactical beast <IMG alt=":-)" src="/mwf/epx.png" class="sic sic_sml_pos"/>
Yes. I think the other programs will be obsolete in a few months. Rybka 3 seems to win most of its games with tactics, Rybka 4 will barely need to play with any positional discernment: one tactical blow after another and all the flies will be dead.
Not in 2009 and not in 2010!
This was the most remarkable information in this interview!
PS: R4 comes, when it is ready!
You're worried about the availability of another version of Rybka?
I'd worry about whether Vas' brain hasn't turned to mush! You gotta wonder if they even know who they are!
You can buy Fritz 12 NOW!
Let's put it this way: if Rybka 4 were 100 points weaker than Rybka 3, I'd call it a piece of shit too.
> Still could beat 99% of the GMs in the world.
By that metric, all engines are worthless because there are free engines that are strong enough to do that.
> 2900 and 3100-does it really matter? Still could beat 99% of the GMs in the world.
I (and most people on this forum, as well as some other computer chess users) don't really want the engines as opponents to play against.
They are used as Analysis tools. For me, it's for analyzing games of mine. In that area, yes, 2900 and 3100 is a huge difference. Some engines are useful mostly in tactical blunders detection only. Nothing else. Many times when weaker programs were used, they fail to properly judge certain situations, and I have to spend hours to be certain if their judgement is correct. Ok, Rybka sometimes fails too, but a LOT less often.
That's the way me and many others use it.
Some others on this forum use it for correspondence chess, or advanced chess,
some use it for opening analysis and finding novelties.
For many years now no one is using the engines purely as playing partners ot play against. Almost pointless.
But for analysis there's a big difference.
If you want adjustable playing strength and controllable styles to play against, go with Chessmaster... None of the best commercial engines even tried to compete against it with that. The Rybka/Naum/Stockfish/Shredder/Hiarcs/Fritz/Junior etc. userbase usually use their engine/s for other reasons (Mainly analysis. Some for Correspondence/Advanced Chess).
> In that area, yes, 2900 and 3100 is a huge difference.
YES. This is what loads of people don't understand. They're is a HUGE difference between 2900 and 3100 elo that a player 2000 elo can take advantage of in analysis of his games.
If you don't care for +/- 200 Elo of strength, you don't need to buy any engine, and not the newest quad cpu, and no 64-bit Windows, etc. Freeware, 32-bit on an average computer can beat 99% of the (otb.) GMs in the world, too. But that is not what top computer chess is about. If "99% GMs" would have been the ultimate goal, chess engine developement could have been stopped years ago. There is constant strength progress disregarding if even 100% of GMs lose. That is not the benchmark, the previous version and the other strong engines make the comparison.
E.g. a top correspondence player who wants to have good analysis tools, will certainly not be content with an engine 200 Elo weaker if he must assume that his opponents use the strongest. Same for the hardware. That's just one example why that matters.
Another aspect is to save time. With other factors being equal, a rather old engine might require like 5 or 10 times the calculation time for a certain chess strength, than a new top engine needs. I mean on same hardware! The engine software improvements which are found still, after so many years in PC chess (~20), are amazing...
>It's about 100 points weaker than the previous version on equal hardware! (4CPU)
That information is not correct and it is misleading.
Fritz 12 on 4 CPU is weaker than Deep Fritz 11 on 4 CPU by about 100 points in the CEGT list(ignoring the error ranges, etc) but one can also say that Fritz 12 on 1 CPU is weaker than Deep Fritz 11 on 1 CPU by about 8 points in CEGT list.
So the expression on equal hardware is misleading and may make people believe that it is also valid for 1 CPU(which isn't). You should have used the statement "on 4 CPU" and not "on equal hardware".
> You should have used the statement "on 4 CPU"
I did use it.
> and not "on equal hardware".
Bah. Single-CPU performance is only relevant on a netbook. Fritz 12 has nobody but its author(s) to blame for being crippled.
>Because the engine is a piece of shit, even compared to Deep Fritz 11.
What is your definition of "the engine is a piece of shit" ?
> What is your definition of "the engine is a piece of shit" ?
The following conditions must be met:
1. It costs money
2. It's weaker than <some arbitrary limit>. I don't have an infinite amount of money so my personal limit is the strongest free engine, currently Rybka 2.2n2. I picked Deep Fritz 11 to highlight how pathetic it is.
The engine being locked to its native GUI doesn't exactly improve my opinion of it.
>so my personal limit is the strongest free engine, currently Rybka 2.2n2.
As i see on CEGT 40/40 list Stockfish 1.5.1 64bit is stronger than the free Rybka 2.2.n2 64 bit.
Also how do you know that Fritz 12 is less strong than the free Stockfish 1.5.1 64 bit in playing against humans? I say that because some people may have that criterion of whether they should call a program a piece of shit or not a piece of shit. :-)
> As i see on CEGT 40/40 list Stockfish 1.5.1 64bit is stronger than the free Rybka 2.2.n2 64 bit.
*updates limit* Thanks.
> Also how do you know that Fritz 12 is less strong than the free Stockfish 1.5.1 64 bit in playing against humans? I say that because some people may have the criterion of whether they should call a program a piece of shit or not a piece of shit. :-)
My magic-8-ball says it is. :)
> That's unkind.
Charging money for an engine that's weaker than Rybka 2.2n2 is unkind.
But even if they charge money for the engine is not unkind because:
- they don't force people to buy but only giving them the chance to do it
- Fritz 12 has some areas where it plays better some positions than Rybka 2.2n2(and Stockfish 1.5.1). I mean no engine is perfect so each engine has some advantages over the other and strength at eng-eng games is not the only criterion for when an engine is good.
Also a person who uses 1 CPU and has a 32 bit operating system will most probably find Fritz 12 stronger from Rybka 2.2n2 32 bit in his computer at long time controls(look CEGT) so he would say different things from you and would disagree with your opinion.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill