Topic Rybka Support & Discussion / Rybka Discussion / What is the timeline for the new version of rybka?
1
2
I did not see this in any other thread. VAs, or others in the know, when do you hope to have rybka 3 (and the new gui) ready?
best
Joseph
best
Joseph
Spring. Search is your friend.
Not 100% sure yet, but yes, spring-ish ..
Vas
Vas
Those 3 little letters strike fear into the hearts of all Rybka fans:
"ish"
"ish"
You don´t believe in winter-ish .. :-)
For sure, it will be AFTER the next Freestyle tournament.
Nelson, don´t worry. If you can make it once, you can make it twice :-). Newest version is only for the newbies :-).
Hehe :)
The most glorious is releasing a version for everybody right before the tournament, and then kicking everybody's butt with said version.
The least glorious is losing. In that case, conditions don't really matter.
So, we will think :)
Vas
The most glorious is releasing a version for everybody right before the tournament, and then kicking everybody's butt with said version.
The least glorious is losing. In that case, conditions don't really matter.
So, we will think :)
Vas
Is the next freestyle scheduled yet?
I've heard January or February.
Vas
Vas
Damn, then I have to go shopping for new CPU's real fast ;)
I've heard February-March from Arno. No specifics.
Thx you and Vasik for the reply. I do not play freestyle, since I do not own fast hardware and I have very little experience of centaur playing. BTW, I really like to analyze freestyle games. They are probably the games I like best, together with games with strong/weird-but-interesting opening novelties.
Sorry for the question but, you will share for the next freestyle tournament
What engine you will use , the 3.0 beta ???
What engine you will use , the 3.0 beta ???
Technically, I think it would be an alpha version :-)
since there is no clear definition of alpha and beta version, it depends on how Vas will name it. I guess he will neither call it alpha nor beta :-D
When the first buds show or after the blooming of the flowers?
Which flowers? We had some ivy blooming in early January last year...
I guess ivy was what sid had in mind :)
Not so fast Felix, we have unfinished business in that secret part of this forum!!! :-) :-) actually you guys do, I'm limited in my capacity! :-P
Turbo, your part of the country is not normal to the rest of the worlds flower sprouting dominions. Any buds past the Mason Dixon line will be subject to customs regulations of importing of flowered information. (ahahah) Oh! That's a good one!
sidserious-lays a low plow the southern attempt to rise again! Oh! " Gone With The Wind " You blow me away!
Any questions and we'll have to meet a Gettysburg!
P.S That's in back of Walmart
sidserious-lays a low plow the southern attempt to rise again! Oh! " Gone With The Wind " You blow me away!
Any questions and we'll have to meet a Gettysburg!
P.S That's in back of Walmart
That the same time at which this was going on last year, there were similar reports in Pennsylvania and New York! (They paid for it, though, as that February was one of the coldest months on Earth in 100 years).
That might be so!
But, I want to know what it is like on BETA 8 ?????
The only people that have that information are our Russian Friends in the Russian Federation- because, they've been to BETA 7 and come back alive.
So, what to you have to say to that Turbo!
sidserious- got to go to K-mart for my med's (I think we need to open the X Files again)
But, I want to know what it is like on BETA 8 ?????
The only people that have that information are our Russian Friends in the Russian Federation- because, they've been to BETA 7 and come back alive.
So, what to you have to say to that Turbo!
sidserious- got to go to K-mart for my med's (I think we need to open the X Files again)
So, what to you have to say to that Turbo!
I say that it's time for NATIONAL12 to beam you back to the mothership :-).
I say that it's time for NATIONAL12 to beam you back to the mothership :-).
I think he has the flu! I wish him a speedy recovery!!!
I hear that they are making another X File movie! Do you think it will come out before Rybka 3? Vas, didn't state WHICH SPRING!!!! 2009-2010 :-P
I hear that they are making another X File movie! Do you think it will come out before Rybka 3? Vas, didn't state WHICH SPRING!!!! 2009-2010 :-P
Spring-ish, winter-ish or whatever, we need "The Ultimat-ish" version that stands the test of time at the end. :-)
Vas, and the testers, how much you think the latest version scores compared to 2.3.2a?
Vas, and the testers, how much you think the latest version scores compared to 2.3.2a?
at least about 50 Elo I would guess.
Translation based on previous elo estimates compared with result in practice:
15 elo
I am not, of course, trying to belittle the great work that Vas, Larry and team have performed. I think this reality is more a result of how difficult it is to improve playing level (a) at this incredibly high level and (b) when there isn't anyone who is definitely better.
15 elo
I am not, of course, trying to belittle the great work that Vas, Larry and team have performed. I think this reality is more a result of how difficult it is to improve playing level (a) at this incredibly high level and (b) when there isn't anyone who is definitely better.
Is it that private Rybka is 50 ELO points better when playing against older Rybka, but only 15 ELO better when playing against the rest of the engines?
That's what history has seemed to suggest. The specific number of 15 I take from CEGT comparison of 2.3.2a and 2.3.1, for which I was hearing estimates of about 50 elo improvement.
I'm talking about 2.3.2a compared with 2.3.1 on single processors, not compared with 2.3, which was especially a flop for single processors; 2.3.1 had been the best for singles to that point, though. I recall predictions of +50 elo for 2.3.2a compared with 2.3.1.
It's okay, though--if Rybka 3 comes out and it's only 25 elo of improvement, I'll still be happy :-).
It's okay, though--if Rybka 3 comes out and it's only 25 elo of improvement, I'll still be happy :-).
> It's okay, though--if Rybka 3 comes out and it's only 25 elo of improvement, I'll still be happy :-).
Even if Rybka gets no improvement, I'll be happy, if it gets Persistent Hash :)
I think that Persistent Hash alone will be able to strengthen the engine tremendously, after an enough number of games :)
Yes, I agree--I would say that the stronger the engine, the more effective will be persistent hash when used well.
The gap to 2.3.1 was 59 Elo here:
http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20All%20Versions/rangliste.html
We took our time with Rybka 2.3.2a and it did turn out as we expected. We did have some trouble with Rybka 2.3 and Rybka 2.3.1, those were rushed to fit the schedule and the lesson was learned.
Vas
http://www.husvankempen.de/nunn/40_40%20Rating%20List/40_40%20All%20Versions/rangliste.html
We took our time with Rybka 2.3.2a and it did turn out as we expected. We did have some trouble with Rybka 2.3 and Rybka 2.3.1, those were rushed to fit the schedule and the lesson was learned.
Vas
But here you're talking about multiple CPU issues. I'm talking about improvement in the engine itself, which in the same link you send gives 13 elo for single CPU. That is also far more accurate because here we're dealing in areas where Rybka has competition with other engines, albeit on higher numbers of CPU. When Rybka is on the rating lists >50 elo higher than anything else, those ratings aren't going to be very accurate, and they don't really so much address the raw improvement in the engine. In addition, you've said before that most of your customer base buys the single-CPU version.
The difference between 2 cpus and 1 in terms of scaling is negligible, and even the difference in opponents' relative rating is not a big deal. To be fair, you should average the 13 and 59 values to estimate the real gain, which gives 36.
Also, on CCRL (who didn't test 2.3.11), we have 75 Elo on 1 CPU going from 2.2 to 2.3.2a.
http://computerchess.org.uk/ccrl/4040/rating_list_all.html
Vas
http://computerchess.org.uk/ccrl/4040/rating_list_all.html
Vas
But only 40 elo on 32-bit...
Okay, fair enough :-)...but most of the customers will be experiencing the case of 32-bit, single CPU software, at least for the time being.
If Rybka 3 comes out and is only +25 Elo, I will consider retirement.
Hahaha, don't consider it too seriously! It would definitely disappoint us more than Rybka getting "only" +25 elo of improvement. For Rybka 3 compared with Rybka 2.3.2a on single processor, my prediction is that it will be 30-40 elo--just a guess; enter me in the lottery :-). Perhaps I should create a poll question on this...
Let's look at history, though: improvement from 1.0 beta to 1.01 beta 13d was 50 points over a long string of in-between betas. From 1.01 beta 13d to 1.1, there was an improvement of 30 points. From 1.1 to 1.2f, there was an improvement of 30 points. From 1.2f to 2.2, there was 10 points. From 2.2 to 2.3.1, there was 10 points. From 2.3.1 to 2.3.2a, there was an improvement of about 15 points. My way of separating these might seem arbitrary, but they were all pretty major stages in the realm of Rybka development. This prediction is kind of like going from version 1.1 to something between 1.2f and 2.2, and from the sound of it, this would seem fairly consistent with the amount of time and effort spent here so far. However, I think that improvements compared with other engines, which is how the rating lists must work, will be more and more difficult as the rating gets higher--history seems to confirm this so far, and I think this is partly due to the drawish tendency of the game of chess...
Let's look at history, though: improvement from 1.0 beta to 1.01 beta 13d was 50 points over a long string of in-between betas. From 1.01 beta 13d to 1.1, there was an improvement of 30 points. From 1.1 to 1.2f, there was an improvement of 30 points. From 1.2f to 2.2, there was 10 points. From 2.2 to 2.3.1, there was 10 points. From 2.3.1 to 2.3.2a, there was an improvement of about 15 points. My way of separating these might seem arbitrary, but they were all pretty major stages in the realm of Rybka development. This prediction is kind of like going from version 1.1 to something between 1.2f and 2.2, and from the sound of it, this would seem fairly consistent with the amount of time and effort spent here so far. However, I think that improvements compared with other engines, which is how the rating lists must work, will be more and more difficult as the rating gets higher--history seems to confirm this so far, and I think this is partly due to the drawish tendency of the game of chess...
In general, you are correct that it should be harder to gain points as we go higher. But on the other hand two of us make much faster progress than one, plus we learn more tricks all the time. So at least this jump should be a big one.
Yes, I think that progress is roughly linear per hour worked.
Vas
Vas
Humour alert... stop posting we are wasting elo here
Unfortunately, this is the same for reverse engineering...
Regards,
Alan
Regards,
Alan
Not for Rybka 3. That party is over, just wait and see. Well, technically, 0 Elo/per hour is linear. :)
Vas
Vas
You can be sure that it's more than 15 Elo points :)
Why can I be so sure? :-)
Man, you are stubborn! Because Felix says so of course! No really, I trust him. :)
Topic Rybka Support & Discussion / Rybka Discussion / What is the timeline for the new version of rybka?
1
2
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill