> It's common sense that you used and have H1.5 at least.
P.S.
However, I do disagree with you here.
Some one with such strong feelings, harboring reasons right or wrongly about that product ,vehemently advocating never having use that product- In point of fact common sense would dictate that you could never determine the contrary.
![[hr]](/mwf/v12/flags/hr.png "Croatia")
> Lies! You say untruthful hurtful things. It's common sense that you used and have H1.5 at least.
>
What is common sense for you, can be rubbish for me, and vice versa. I don't use also Komodo, Critter, Fire-, Robbo-, Ippo- and many others. I just don't. I have no need for that.
As for the rest, you believe what you want, I believe what I want. You can believe that the Earth is flat, for all I care. You believe a person can come out of nowhere and produce the best chess playing program so far? Oh, yes, Vasik did so, too. The only difference is Vasik made another couple of hundred ELO later, demonstrating he is able to produce mirracles ON HIS OWN. Houdart didn't do nothing, and is trying to sell this nothing around. In my book, it sucks.
Well, for me the ultimate determining factor of this argument will be when Vas programs Rybka 5 to rise out of the ashes of its old image and regains the number one spot. But, the point of fact is Houdini has been here now for a quite some time and no one has been able to over take it. I don't see ANY chess programmer to date - using ippolit source code or black magic over coming this bloody engine. I hate like hell to say this but telling someone to shut up is not the answer. Perhaps, listening, is a better solution.
> telling someone to shut up is not the answer
Agreed, but that was just my way of saying "Ok, I was able to understand what you are doing and am not particularly angry as obviously today many people do this, to the point when it's almost normal, and I could have coped with your existence while you we're quiet, but when I read your posts, and see how difficult and egoistic you are, I go berserk."
![[cr]](/mwf/v12/flags/cr.png "Costa Rica")
Yes, it will be very nice if Vas could get the first place again, but he needs to give a very good jump in strength. In both time controls Houdini 2.0c is 70 Elo points stronger than Rybka 4.1 according to CCRL! Though in the 40/40 list Houdini 2.0c has played very few games still.
http://www.computerchess.org.uk/ccrl/404.live/rating_list_all.html
http://www.computerchess.org.uk/ccrl/4040.live/rating_list_all.html
We will see very soon if Vas is able to surpass Houdini and Rybka get the throne again. I think he can do it, but in the last interview he did not look very confident in seeking only that goal. If his main goal is still the Cluster project, then I doubt Rybka 5 will do it. Probably the gap with Houdini will be reduced a lot but no so much to surpass it. This will be very interesting. We will see.
Regards,
Gaмßito.
http://www.computerchess.org.uk/ccrl/404.live/rating_list_all.html
http://www.computerchess.org.uk/ccrl/4040.live/rating_list_all.html
We will see very soon if Vas is able to surpass Houdini and Rybka get the throne again. I think he can do it, but in the last interview he did not look very confident in seeking only that goal. If his main goal is still the Cluster project, then I doubt Rybka 5 will do it. Probably the gap with Houdini will be reduced a lot but no so much to surpass it. This will be very interesting. We will see.
Regards,
Gaмßito.
If this is any indicator-
Per Lukas (and I hope he doesn't mind my quoting him here)
Yes
Atm Rybka 5 is top priority. After this cluster development will continue
I am hopeful that we will see a state of the art Rybka 5 that will meet all expectations. I think that this time around there is more on the line than just making a version update. I believe he see this and it has been factored into making this his priority.
Per Lukas (and I hope he doesn't mind my quoting him here)
Yes
Atm Rybka 5 is top priority. After this cluster development will continue
I am hopeful that we will see a state of the art Rybka 5 that will meet all expectations. I think that this time around there is more on the line than just making a version update. I believe he see this and it has been factored into making this his priority.
I must object to Komodo being grouped with various Rybka offshoots in your statement "I don't use also Komodo, Critter, Fire-, Robbo-, Ippo- and many others." You don't say so, but you seem to imply that Komodo is an Ippo-derivative in the sense that all the others named (including Houdini) are. This is completely wrong. Komodo has far more in common with Stockfish than with any of the other named programs. We don't even use the basic Rybka/Ippo idea of "expected gain" of moves. The evaluation has some slight similarity to the Rybka family only because I worked on both. The main ideas in the search are similar to Stockfish, not to the very different Rybka family, though of course we have many original ideas that make us stronger than Stockfish. Of course we use a few Ippo ideas that are widely known, but in general we have not found Ippo ideas helpful in Komodo. So I would say that only if you refuse to use Stockfish for some reason would it also be logical to stay away from Komodo.
Hello Larry,
Two questions:
1. Can you please give us some news about the next Komodo 4 and its strength improvement over version 3?
2. Rybka 4.0 was made in May 2010 and 4.1 in March 2011. Certainly Vas has had time to improve the engine. But according to CCRL Houdini 2.0c is now +67 Elo stronger than Rybka 4.1 (both on 6 cores) on 40/4 blitz TC, and +54 Elo stronger on 40/40 TC (both using 4 cores). Do you expect Rybka 5 to be stronger than Houdini 2.0c? Or do you expect Houdini will continue leading for much longer?
Best wishes to you and Don in your work with future Komodo versions, Larry.
Regards,
Gaмßito.
Two questions:
1. Can you please give us some news about the next Komodo 4 and its strength improvement over version 3?
2. Rybka 4.0 was made in May 2010 and 4.1 in March 2011. Certainly Vas has had time to improve the engine. But according to CCRL Houdini 2.0c is now +67 Elo stronger than Rybka 4.1 (both on 6 cores) on 40/4 blitz TC, and +54 Elo stronger on 40/40 TC (both using 4 cores). Do you expect Rybka 5 to be stronger than Houdini 2.0c? Or do you expect Houdini will continue leading for much longer?
Best wishes to you and Don in your work with future Komodo versions, Larry.
Regards,
Gaмßito.
The goal of Rybka 4.1 was NOT to improve playing strength. In fact, its goal was to fix bugs in Rybka 4 WITHOUT improving playing strength (aside from bug related reductions). Vas was very clear about this with the general public, and even clearer with the beta testers.
So the real question is: Has Vas developed enough improvements in the 19 months since May 2010 that he is willing to expose in a released UCI version to overcome the ~70 Elo deficit to Houdini? Please note that this question has two parts; 1) How strong is the strongest Rybka for SMP?, and 2) How much of this strength will be released in R5?
So the real question is: Has Vas developed enough improvements in the 19 months since May 2010 that he is willing to expose in a released UCI version to overcome the ~70 Elo deficit to Houdini? Please note that this question has two parts; 1) How strong is the strongest Rybka for SMP?, and 2) How much of this strength will be released in R5?
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
> The goal of Rybka 4.1 was NOT to improve playing strength. In fact, its goal was to fix bugs in Rybka 4 WITHOUT improving playing strength (aside from bug related reductions). Vas was very clear about this with the general public, and even clearer with the beta testers.
>
> So the real question is: Has Vas developed enough improvements in the 19 months since May 2010 that he is willing to expose in a released UCI version to overcome the ~70 Elo deficit to Houdini? Please note that this question has two parts; 1) How strong is the strongest Rybka for SMP?, and 2) How much of this strength will be released in R5?
Yes this is the most accurate way of seeing the issue. I'm pretty sure (based on my rental of the Rybka Cluster earlier this year) that the answer to Q 1) is > Houdini 2. But the answer to Q 2) may be depressing.
Komodo 4 was supposed to be released today (SP version), apparently there was some delay with getting it compiled suitably. Maybe it will be a Christmas surprise for everyone! It's no longer free, but pretty inexpensive ($19.95).
The MP version should follow in January. Price $39.95, but if you buy the SP now, the MP will be half price so the SP now is still effectively free for anyone planning to buy the MP.
The eval has been changed quite a bit, plus some search improvements. Elo gain probably somewhere in the ballpark of 20, though I'm hoping it will turn out to be more.
I would expect Rybka 5 will surpass Houdini. Perhaps all that's needed to do this is to duplicate the changes Houdart made to create Houdini 1.5 from Ivanhoe, which have already been published by Critter author Richard Vida. If Ivanhoe is really pretty much a "clone" of Rybka, that ought to work. Anyway I'm sure Vas has plenty of his own improvements.
We need one really good idea for Komodo to catch Houdini, I think. If Rybka is first, Komodo second, and Houdini third in a few months, that will be fine with me!
The MP version should follow in January. Price $39.95, but if you buy the SP now, the MP will be half price so the SP now is still effectively free for anyone planning to buy the MP.
The eval has been changed quite a bit, plus some search improvements. Elo gain probably somewhere in the ballpark of 20, though I'm hoping it will turn out to be more.
I would expect Rybka 5 will surpass Houdini. Perhaps all that's needed to do this is to duplicate the changes Houdart made to create Houdini 1.5 from Ivanhoe, which have already been published by Critter author Richard Vida. If Ivanhoe is really pretty much a "clone" of Rybka, that ought to work. Anyway I'm sure Vas has plenty of his own improvements.
We need one really good idea for Komodo to catch Houdini, I think. If Rybka is first, Komodo second, and Houdini third in a few months, that will be fine with me!
![[mx]](/mwf/v12/flags/mx.png "Mexico")
Ouch! The Komodo SP version goes commercial
I don't like it, but I think it's a smart move considering many people use Komodo under IDEA and they only need the SP version.
I thought having a SP version free would be a nice thing other engine developers could adopt, that way I could get a shot at using Hiarcs 14 SP or Shredder 14 SP. But now that Komodo abandoned it, it's something that just won't happen...
I don't like it, but I think it's a smart move considering many people use Komodo under IDEA and they only need the SP version.
I thought having a SP version free would be a nice thing other engine developers could adopt, that way I could get a shot at using Hiarcs 14 SP or Shredder 14 SP. But now that Komodo abandoned it, it's something that just won't happen...
We intend to continue to offer the previous major version SP free, at least unless we get so strong that this no longer seems reasonable.
Fair enough. That would be a nice thing other engine developers could adopt (release the SP version of the previous release version for free).
That's a weird statement, considering the SP version has no use whatsoever if you own the MP version, as the MP version can be set to 1CPU.
>the MP will be half price so the SP now is still effectively free for anyone planning to buy the MP.
That's a weird statement, considering the SP version has no use whatsoever if you own the MP version, as the MP version can be set to 1CPU.
Well, in the present instance it means that you can get the SP for free now (in effect) without having to wait for the MP release, as long as you are planning to buy the MP. If they are released simultaneously you would be right.
Larry, it's a good idea to charge a modest fee for a fine product. I am actually glad that you are doing this just so that you have an incentive to keep working on the product.
However, I was unclear as to your estimate of K4's strength ("20 Elo improvement over K3"). Where would that place K4 vs. R4.1?
However, I was unclear as to your estimate of K4's strength ("20 Elo improvement over K3"). Where would that place K4 vs. R4.1?
K3 SSE was slightly ahead of R4 on some (most, I think) lists, so K4 should be comfortably ahead, especially for those with newer computers which can use the SSE version. Of course I'm comparing sp to sp.
LK:
Don's delivery of the download was super efficient and rapid. The Help file explaining Komodo 4 options was exemplary. Hopefully, other engine makers will take note.
I did not see any provision for TB files. Will that be in the Komodo 4 MP?
Don's delivery of the download was super efficient and rapid. The Help file explaining Komodo 4 options was exemplary. Hopefully, other engine makers will take note.
I did not see any provision for TB files. Will that be in the Komodo 4 MP?
We do plan to add TBs, but that is low priority compared to MP. I would expect TBs to be in the next version after the initial MP release.
Is there any plan on a "Preserve Hash" or "Position Learning" mechanism? I think that with the latter (that saves info to Hard Disk) the former is not needed, and recall Don commenting he was interested in this at some point.
I expect we will get to this at some point next year.
Richard Vida has successfully implemented "Position Learning" in his engine, and told Don that he could give Critter's sources if he was interested, so it may be possible for you guys to make contact and implement Critter's Session File system into Komodo.
![[cn]](/mwf/v12/flags/cn.png "China")
Wow, that's very generous. I don't know much about Vida, but he's sounding like a decent guy to me at this moment in time!
He is a very decent guy as well as a great programmer. He has already helped us improve Komodo in more ways than one. Critter is a really good program, but it is very similar to Houdini and the other Ippos in many ways. I would like to see Critter be the best of the Ippo-like programs.
> He is a very decent guy as well as a great programmer. He has already helped us improve Komodo in more ways than one. Critter is a really good program, but it is very similar to Houdini and the other Ippos in many ways. I would like to see Critter be the best of the Ippo-like programs.
If that statement isn't a great example of an "asteism" than I would be hard pressed to find one that is.
Thanks for teaching me a new word! But how is pointing out the similarities of Houdini and Critter a criticism? The similarity and the use of nearly all Ippo ideas is not a secret; Houdart and Vida are simply in a race to see who can better implement and improve those ideas. Houdart is currently leading because he got a big head start by actually using the Ippo code, while Vida wrote his own program using Ippo as a "template", more or less as Vas is accused of doing with Fruit. But since Vida doesn't compete in tournaments and has been fairly forthright about the Ippo connections of Critter, he has done nothing wrong and I wish him success in besting Houdart, who has not been similarly forthright.
Larry- you admittedly stated "He has already helped us improve Komodo in more ways than one" Where does that place you (Komodo)? Look! By what you are stating -it appears you are trying to have it both ways.
Painting his engine as tainted and Komodo as , what?
Painting his engine as tainted and Komodo as , what?
There are three very different approaches to writing a new computer chess program:
1. Copying an existing program and then making changes to improve it. Example: Houdini. This is okay if the author of the copied program doesn't object and if you acknowledge that this is what you are doing. By this standard, "Fire" is okay while Houdini is not.
2. Writing a new program "in the image" of an existing one, meaning that you use almost all of the ideas and formulas of the existing one, perhaps with modifications and improvements, but you write your own code. This is what Ippo did with Rybka, what Rybka is accused of doing with Fruit, and what Critter did with Ippo. It is legally defensible, and if the author of the older program does not object it is morally ok. The newer program cannot compete in tournaments though without the express permission of the older one's author, which would be hard to get in the case of Ippo since no one knows who the author is. So I am not criticizing Critter here, or Rybka (the objection from Fruit coming five years too late); Ippo can be criticized on the grounds that Rybka did object.
3. Writing a new program using selected ideas from various other programs, with no strong resemblance to any one. This is what Komodo does, and is the standard for tournament participation without anyone else's permission.
There is a world of difference between each of these 3 methods. All are okay if they follow the above guidelines, which Critter did and which Houdini did not.
1. Copying an existing program and then making changes to improve it. Example: Houdini. This is okay if the author of the copied program doesn't object and if you acknowledge that this is what you are doing. By this standard, "Fire" is okay while Houdini is not.
2. Writing a new program "in the image" of an existing one, meaning that you use almost all of the ideas and formulas of the existing one, perhaps with modifications and improvements, but you write your own code. This is what Ippo did with Rybka, what Rybka is accused of doing with Fruit, and what Critter did with Ippo. It is legally defensible, and if the author of the older program does not object it is morally ok. The newer program cannot compete in tournaments though without the express permission of the older one's author, which would be hard to get in the case of Ippo since no one knows who the author is. So I am not criticizing Critter here, or Rybka (the objection from Fruit coming five years too late); Ippo can be criticized on the grounds that Rybka did object.
3. Writing a new program using selected ideas from various other programs, with no strong resemblance to any one. This is what Komodo does, and is the standard for tournament participation without anyone else's permission.
There is a world of difference between each of these 3 methods. All are okay if they follow the above guidelines, which Critter did and which Houdini did not.
I don't know if I agree with your idea of what you consider original. I'm not sure that each individual programmer should be given leave to run around making their own decisions as to what is pure(r), than the guy before them -where perhaps they were helped " in more ways than one".
Maybe you are right. Do you favor leaving it up to the ICGA to decide? If not, then who?
> Maybe you are right. Do you favor leaving it up to the ICGA to decide? If not, then who?
Ask Komodo's programmer that pip of a question! eh!
> Do you favor leaving it up to the ICGA to decide? If not, then who?
Who?
The origin of Komodo investigated by the same ICGA measuring rod as Rybka was, pushing rule #2 to its limits, would very well not have a typical Hollywood ending.
From the Doch 09.980 README file we read:
Also, much credit goes to the authors of open source chess programs. Many of the ideas and techniques for doch have been borrowed from these wonderful works of art.
Can you not see the similarities between Vas and Don ?
1. Studied and learned from open sources.
2. Coming out of the blue with an incredible elo.
Can this be done without copying? I believe the answer is yes. The post-internet generation learns 10 times faster than the pre-internet generation. I think Richard just proofed that, Don too, Vas and Fabien also. It's just the ICGA who are the slow learners here.
> From the Doch 09.980 README file we read:
> Also, much credit goes to the authors of open source chess programs. Many of the ideas and techniques for doch have been borrowed from these wonderful works of art.
What constitutes a "technique"? Seems to be something different from an idea, as it is split out in the above citation.
You could say that this is just a wild guess - and you'd be right. But I think if the ICGA investigated all the engines that participate, and apply the same approach and interpretation standards that they did with Rybka, 9 out of 10 engines would be in breach of their rule #2. That leaves them without a competition at all.
![[us]](/mwf/v12/flags/us.png "United States")
> That leaves them without a competition at all.
They have already been left without a competition at all, even without doing what you suggest.
Let's make it official and put it on record!
![[fr]](/mwf/v12/flags/fr.png "France")
It will continue. You guys are not the customer, the customer who needs to be kept happy is the foolish sponsor, they can't fool all of the people all of the time, but that doesn't matter, they only have to fool a sponsor for enough of the time. That can be done with any old program's with any old stupid names. Only an alternative tournament with the strongest and basically banned programs will destroy the old order.
"2. Coming out of the blue with an incredible elo."
This is clearly not true. The first released version of Doch was a reasonably good program, but not as good as any of the "name" programs like Shredder, Fritz, Hiarcs etc., never mind Rybka. Each subsequent release was moderately stronger until Komodo eventually passed everyone but Houdini (so far). This is not at all consistent with the idea that we copied from anyone. We have used various search ideas from both Stockfish and Ippo (more from Stockfish than from Ippo), but the details are always different as is of course the code, and we have rejected a huge number of their ideas as poor ones. The eval is completely original, with only the fact that I worked on both Rybka 3 and Komodo giving them any commonality. It is my view that if a new program shows excessive similarity in either search or eval with an existing one, regardless of how it is coded, then it should not be allowed in competition without the express permission of the earlier author. This is of course a judgment call, and it is the right of the sponsoring organization of the competition to make that call. I'm not taking any position on the Rybka-Fruit issue as I have not studied it myself. I have stated that even if one accepts the ICGA judgment on Rybka 1, later versions (let's say at least 2.3 and beyond) were surely sufficiently different to be eligible to play on their own merits, though of course this would have to be demonstrated. The only program that I feel is clearly NOT original enough to be allowed to compete without express permission from another is Houdini, and since we don't know the author of Ippo he can't get that permission. As for Ippo itself, it is a borderline call. At least they went to some trouble not to be so similar to Rybka as to make the case a slam-dunk.
This is clearly not true. The first released version of Doch was a reasonably good program, but not as good as any of the "name" programs like Shredder, Fritz, Hiarcs etc., never mind Rybka. Each subsequent release was moderately stronger until Komodo eventually passed everyone but Houdini (so far). This is not at all consistent with the idea that we copied from anyone. We have used various search ideas from both Stockfish and Ippo (more from Stockfish than from Ippo), but the details are always different as is of course the code, and we have rejected a huge number of their ideas as poor ones. The eval is completely original, with only the fact that I worked on both Rybka 3 and Komodo giving them any commonality. It is my view that if a new program shows excessive similarity in either search or eval with an existing one, regardless of how it is coded, then it should not be allowed in competition without the express permission of the earlier author. This is of course a judgment call, and it is the right of the sponsoring organization of the competition to make that call. I'm not taking any position on the Rybka-Fruit issue as I have not studied it myself. I have stated that even if one accepts the ICGA judgment on Rybka 1, later versions (let's say at least 2.3 and beyond) were surely sufficiently different to be eligible to play on their own merits, though of course this would have to be demonstrated. The only program that I feel is clearly NOT original enough to be allowed to compete without express permission from another is Houdini, and since we don't know the author of Ippo he can't get that permission. As for Ippo itself, it is a borderline call. At least they went to some trouble not to be so similar to Rybka as to make the case a slam-dunk.
Ed - "2. Coming out of the blue with an incredible elo."
Larry - This is clearly not true. The first released version of Doch was a reasonably good program, but not as good as any of the "name" programs like Shredder, Fritz, Hiarcs etc., never mind Rybka.
Doch 09.980 64-bit 2969
Rybka 1.0 64-bit 2919
So I am a bit surprised you saying this.
Did you feel I said or implied this ?
As you perhaps know my beef is with the ICGA and their stupid rule #2 which they pushed to its limits to get the wished effect, my beef is not with Komodo. The similarities between Vas and Don are striking and if the origin of Komodo were hold to the same measuring rod Don could be the next victim if you decide to participate. Something to think about.
It's also quite funny that Don's similarity utility pleads Rybka free and the difference between Rybka 1.0 and the first Komodo is negligible, only 3%, see the second chapter of this page: http://www.top-5000.nl/quick_guide.htm
I have talked with Don about that page, assured him it has nothing to do with Komodo and he was okay with that.
You also might have a look at the real clones, a study of Adam Hair, see: http://www.top-5000.nl/clone.htm
Larry - This is clearly not true. The first released version of Doch was a reasonably good program, but not as good as any of the "name" programs like Shredder, Fritz, Hiarcs etc., never mind Rybka.
Doch 09.980 64-bit 2969
Rybka 1.0 64-bit 2919
So I am a bit surprised you saying this.
> This is not at all consistent with the idea that we copied from anyone.
Did you feel I said or implied this ?
As you perhaps know my beef is with the ICGA and their stupid rule #2 which they pushed to its limits to get the wished effect, my beef is not with Komodo. The similarities between Vas and Don are striking and if the origin of Komodo were hold to the same measuring rod Don could be the next victim if you decide to participate. Something to think about.
It's also quite funny that Don's similarity utility pleads Rybka free and the difference between Rybka 1.0 and the first Komodo is negligible, only 3%, see the second chapter of this page: http://www.top-5000.nl/quick_guide.htm
I have talked with Don about that page, assured him it has nothing to do with Komodo and he was okay with that.
You also might have a look at the real clones, a study of Adam Hair, see: http://www.top-5000.nl/clone.htm
Why compare Doch with Rybka 1, when Rybka 3 had already been out for some time? Doch was nowhere near the level of Rybka 3. Rybka 1 was not a very interesting program in terms of ideas, it was only starting with Rybka 2 that Vas introduced so many original ideas. There was hardly any significant idea in it that had not appeared in earlier programs. It was the strongest when it came out because it was 64 bit, and because it combined the best ideas of known programs. This of course does not imply that anything was improper about Rybka 1, in fact that is pretty much the same course of action we took with Doch/Komodo -- starting with a fairly standard program, then gradually adding in our own ideas (as well as others that came to our attention). It was (and is) very easy to write a program as good as Rybka 1 without copying anything; I think Doch was already stronger because we already did have one or more fairly significant original ideas in it.
I know you didn't accuse us of copying, but many readers might take it that way. I just don't agree with the characterization of early Doch as having an "incredible elo". Rybka 1 was already hopelessly out of date by then.
Without taking any position on the Rybka 1 vs. Fruit issue, I can say that I am confident that any comparison of Komodo with Rybka, Ippo, or any other program will not show any suspicious degree of similarity, except that most of the eval terms (but not the weights) are similar in all the programs that I worked on (Komodo and Rybka 2.3.2.a thru 3.0) and therefore also in Ippo, Houdini etc. So if any eval similarity is demonstrated between Komodo and any Ippo-derived program, it would only be evidence for disqualifying the Ippo program. I suppose that if both Komodo and Rybka wanted to enter an ICGA tournament (assuming no ban on Rybka), they could refuse to allow both on the grounds of one common contributor to both (myself), but I think that would be a stretch.
I know you didn't accuse us of copying, but many readers might take it that way. I just don't agree with the characterization of early Doch as having an "incredible elo". Rybka 1 was already hopelessly out of date by then.
Without taking any position on the Rybka 1 vs. Fruit issue, I can say that I am confident that any comparison of Komodo with Rybka, Ippo, or any other program will not show any suspicious degree of similarity, except that most of the eval terms (but not the weights) are similar in all the programs that I worked on (Komodo and Rybka 2.3.2.a thru 3.0) and therefore also in Ippo, Houdini etc. So if any eval similarity is demonstrated between Komodo and any Ippo-derived program, it would only be evidence for disqualifying the Ippo program. I suppose that if both Komodo and Rybka wanted to enter an ICGA tournament (assuming no ban on Rybka), they could refuse to allow both on the grounds of one common contributor to both (myself), but I think that would be a stretch.
There was hardly any significant idea in it that had not appeared in earlier programs.
Anthony found the extreme futility pruning in early Rybka to be new and significant, and even tried (unsuccessfully) to implement the idea in ZIIM.
Anthony found the extreme futility pruning in early Rybka to be new and significant, and even tried (unsuccessfully) to implement the idea in ZIIM.
Do you know whether Rybka 1 did futility differently than older programs, or did it just do more plies of it and/or use smaller margins? If it was only plies and margins, it wasn't a new idea, just good tuning.
Search for the earliest postings of the user "Anthony C" (relating to his examination of Strelka). R1 futility pruning values varied significantly from Heinz' paper, but by your definition (and Anthony's) it wasn't really a new idea...
I haven't researched this yet but according to Watkins:
"chapter B.1. : The principal new idea in Rybka 1.0 Beta seems to me to be the loosening of the stringent cutoff values for futility previously used in the AEL pruning of Heinz. These add up to about a 75-100 Elo improvement on a 32-bit machine, comparable to the amount that Fruit itself gained over the second half of 2005."
Among other things Larry missed I can mention Rybka's unique idea to use a pawn-value=3200 in EVAL and a pawn-value=100 in search which in the meantime is adopted by many. I think that also was brand new. It's simply brilliant.
"chapter B.1. : The principal new idea in Rybka 1.0 Beta seems to me to be the loosening of the stringent cutoff values for futility previously used in the AEL pruning of Heinz. These add up to about a 75-100 Elo improvement on a 32-bit machine, comparable to the amount that Fruit itself gained over the second half of 2005."
Among other things Larry missed I can mention Rybka's unique idea to use a pawn-value=3200 in EVAL and a pawn-value=100 in search which in the meantime is adopted by many. I think that also was brand new. It's simply brilliant.
![[at]](/mwf/v12/flags/at.png "Austria")
> Among other things Larry missed I can mention Rybka's unique idea to use a pawn-value=3200 in EVAL and a pawn-value=100 in search which in the meantime is adopted by many. I think that also was brand new. It's simply brilliant.
Can you please explain to the layman, what is the purpose of this brilliant idea, and how does it work?
The goal of 3200/32 -> more equal scores but better rounded -> more beta-cutoffs -> (somewhat) faster search.
Are you saying that Rybka was the first program to round scores from the eval to get more cutoffs? I think that is highly unlikely, though I don't know for sure. Even Rybka later switched from 3200 to 1000 per pawn for eval, so there is nothing special about using 3200.
The idea to use different scores in eval and search certainly is new for me.
Thanks.
![[de]](/mwf/v12/flags/de.png "Germany")
Was there any chess program before Rybka 1 that used large tables of material imbalance? I don't think so.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill