Rybka Chess Community Forum
When will Rybka 3 be released? (Closed)
|April is not over yet!
This poll was done a few months ago but time has marched on and opinions change!
The answer is simple - Rybka 3 will be released when it is ready. Comprende?
Agreed. May be that Werewolf should change the poll's title: when will Rybka be ready? :)
I'm not so sure--I imagine it will be released sometime after that because even the "ready" product will need to go through some testing, and it will be released when the test results are positive enough to warrant this.
Testing will not take long - but there is still some work to be done although the current version already is super strong ;)
Testing will not take long
Please don´t say that! We want one final version only from Rybka 3 engine. I think, Vas is also interested in it. If I remember correct, Vas has spoken from a period of about 3 weeks for testing. With good and high motivated testers you can make it in this time, but it is a hard job. Testing don´t mean, playing a little bit here and a little bit there (and make some engine - engine matches). You need clear guidelines, what to test.
I remember the release from Rybka 2.3.2 (not a). The very first I tested was the announced new ability of zugzwang detection. And what I see? Nothing! Noone tested it!
Maybe for the team a single version is easier work, but the product will not be perfect anyway. A software can be tested for months, but when people start to use it, they ALWAYS will find bugs that the team will not be able to find. The make "strange" use that the developper could not guess and in a single month of normal use (let us say - 1000 users) you get 80 years of accumulated testing. I think that the 2.3 version model was correct, even if less convinient for the team.
Software is not perfect anyway. Here I agree. Rybka 2.3.2a isn´t perfect and has bugs, but is good enough to live without updates for a year. If Rybka 3 has only similiar small bugs like this, all is okay. I have inhaled the bugs; if they are gone, I have to change myself. I can make it :-).
All here are so eager to have updates from Rybka 3. No way! Nothing! Vas has to show, that he is professional. One release and cash flows! That´s it!
PS: I´m sure, that we will see many updates for the new gui. But this is an other story!
> I´m sure, that we will see many updates for the new gui
Q: Will you only have major releases or do you expect some minor
Victor: There will be frequent updates for the program, databases and
(Interview with the Developers of Rybka Aquarium
I guess they were talking about the engine here... :)
Many thanks for the link. I find, this part is very interesting:Q: When can we expect the first release of Rybka Aquarium?
Pete: I try to keep my word, so I’m not saying anything here (smiles)
Victor: Vas and the ChessOk team pays special attention to the quality of future products, so the exact date has not been determined. But Rybka 3 and Aquarium will be released at the same time. The product will be released when all features have been carefully tested and polished. I recommend checking the news pages at the end of May.
So end of May is an very interesting datum (as I suggested). If Victor (and other people?) says, all is okay, then we have the release date of Rybka 3 in june (16-20). I´m sure, this is the internal schedule, Vas has said here
. All speculations! But I´m convinced :-).
Aslong as they don't tell the year... ;-)
I re-read one of the first Rybka interviews, and Vas promised a GUI "already in the summer of 2006" or something like that :-)
/* Steinar */
Yeah, I have a tendency to be absurdly overoptimistic :)
In reality, we're putting the pieces in place, just at around 20% of the expected speed.
Am I right in thinking that Iweta and others are rigorously testing the engine all along, so that the "final" testing need not take too much time?
You presume that there is something called Rybka 3 beta that is being tested. There are probably many betas and new ones being created all the time.
And presumably these versions get tested for worthiness as they go along . . .
All kinds of little things are put off until later. This is quite normal. The "release cycle" will need about a month.
As a rule of thumb, development testing should take as long as it took to code & debug - then there is system testing! :o(
"ready", IMHO, meant: when all things'll be fixed, and final tests will not evidence any further need of bug cleaning.
"All" is a strong word... And may be quite impossible to achieve... Unless you're ok with getting the program in 2019:)
Sufficient would be all known non-minor issues.
We programmers are slaves to the rule of cybernetic entomology: "There is always one more bug" :)
p.s. Even in industrial code, that has gone through many months of QA tests and deugging cycles, there are bugs. Just... they're minor bugs
It does not matter as long as it improved strong enough over the months. I expect at least 90 elo improvement for a long wait, else i just would be disappointed.
If you're expecting 90 elo improvement then you should stop waiting because I highly doubt Rybka 3 will be anywhere near 90 elo stronger than 2.3.2a. An improvement of 40 elo would be considered excellent.
40 Elo would be a huge disappointment considering the claims Vas and Larry have been making.
I can't remember Vas making any claim about Elo improvement for R3?!
>according to our tests Rybka v2.3.2z3 should be 15-20 Elo stronger than Rybka 2.3.2a in normal chess.
CCRL started testing that version last August.
>Actually, Rybka 3 eval is more than a ply better than Rybka 2.3.2 eval
>I spent about a month to six weeks in late 2007 improving the MP algorithm. This gained around 30 Elo on a quad and led to the very interesting search sampling which Rybka 3 will have.
And I think he said the search stuff was looking 'promising' at some point. :-p
I think there's been even more revealing hints:
Larry said 100 elo under certain circumstances (octal, blitz) down to 50 elo in others (humans 40 in 2).
And THAT was with a crippled search!
> 40 Elo would be a huge disappointment considering the claims Vas and Larry have been making.
I like miracles.
Why should I wait for another a-lil-bit-better thing this long?
As time goes by, our expectation will be higher and higher.
I would guess that the actual elo in terms of how much better it would play theoretically in human tournaments would be in the range of 50-60 elo (higher than my previous guess of 30-40 elo). This is based on the fact that some Rybka insiders already have nearly +100 elo improvement on Rybka 3 Beta compared with Rybka 2.3.2a--but one must keep in mind that these are long Rybka vs. Rybka tournaments, and such numbers have historically tended to be inflated compared with actual elo gain.
I would guess that the actual elo in terms of how much better it would play theoretically in human tournaments would be in the range of 50-60 elo
Assume that Rybka 2.3.2a makes today more than 80% in human (GM) tournaments on best hardware (skulltrail) and best opening book. To make 60 Elo more means, Rybka 3 has to make 88% (with black and white). I don´t believe in it. Humans aren´t opponents for measuring Elo increase of an engine. I think, Zappa and new DFritz would perform quite similar against humans, although they are more than 60 Elo worse.
I think that you're right in terms of the number of games that we'd be able to get, but I think that I'm right if you assume that these engines each play 100 games each against each of the top 20 humans in the world. :-)
April is not over yet! (Though this may not be related to the question.) :)
Jokes aside, I'm not the only optimist around!
I think that Rybka "people" is waiting for the releasing of Fritz 11 multiprocesor, so we have to wait when Fritz 11 multiprocesor is finished.
I see no reason for that. What difference would it make?
It is possible that many computer chess fans are planning to buy a new top chess program again, this year - but only ONE. So, if they are not sure yet which one to buy, it is possible that many will simply buy the one which will be released first.
It makes no sense to "wait" for something external. As soon as it's ready for release, it should be done quickly. But I guess that is how it will be done anyway...
Doing that would be a bad strategy. In the remote case that Deep Fritz 11 is stronger than current Rybka, there will be a lapse of time in where DF11 will top all the rating lists until Rybka 3 is released.
Then the competition will be able to brag about being the top 1, strongest chess software pointing to that lapse of time, and that may be bad for Rybka's sales.
I don't see why would Rybka's developers risk such a scenario, so clearly releasing Rybka 3 before Fritz 11 is the best strategy.
I personally think the delay is because of the cloning threat. I think Vas is taking his time trying to work out how to stop people from decompiling it so easily.
You very overestimate the danger of decompilation and cloning.
Hardly, Vas himself has mentioned this as a big problem
And Vas also overestimate.
well, it depends what you mean by overestimate. If I'd written the world's strongest chess engine and there was just one person who decompiled, cloned and distributed it then I'd get very paranoid about it. I think that's a fair reaction.
If "one person" is me?
Decompilation of Rybka is very-very hard for me.
I can make this work only for time more than my life.
I didn't realise you were the person doing this.
It is totally immoral to steal someone else's work by cloning or even attempting to clone it. No one should even attempt this.
About "moral" or "immoral" let's talk 1000... years later.
you must be the only person on the forum that did not know this.
actually it comes back to me now - I forgot his name. A lot of bad software is coming out of russia, virus' etc
Can you estimate the number of hours spent decompiling Rybka beta? One can imagine that this was a very long and tedious process which you are obviously very good at. The interesting thing for me is that the ability to reverse engineer a program with rather sophisticated algorithms is a very valuable skill. I've always wondered why you spend your time reverse engineering chess engines rather than other programs with the potential for more significant financial rewards.
I am not interested for "more significant financial rewards".
I spend my time for more valuable things.
For decompiling Rybka beta quite enough one or two months. For Rybka 2.3.2a - more than 10 years.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill