![[us]](/mwf/v12/flags/us.png "United States")
No, there was no code copying. however, there was "Data Copying" which is exactly the same thing in the given context. Not allowable.
![[gb]](/mwf/v12/flags/gb.png "United Kingdom")
> Not allowable.
I'm interested in this. Can I ask if you think the following is allowable?
1) I write an opening book, lots of green and red moves in it, created through 1000's hours of analysis/tuning/game playing and sell it (copyrightable etc)
2) A competitor wants to get the lines out of my book and put them in his book to sell
3) The competitor does not look in my book to steal my lines, instead he plays two engines against eachother, both using my book, truncates all the games at book exit and imports them into his book
So effectively he gets my lines that way.
Is this allowable?
I do not believe so, but that is getting into a technical issue. Since both programs use your book, the book lines played are clearly yours. The grey area would be if he claims that he obtained the games from other methods/sources.
I think the argument would be simply that he did use your book to produce his book, and I don't think there would be any possible way the copyright infringement claim could be denied, if he explains the process exactly as you explained it.
I think the argument would be simply that he did use your book to produce his book, and I don't think there would be any possible way the copyright infringement claim could be denied, if he explains the process exactly as you explained it.
My apologies, I added your post above. As we both agree there was no code copying for the PST issue, I don't think that our conclusions differ. What the ICGA consider to be a derivative is up to them.
It _would_ be completely acceptable. But if you copy a couple of hundred lines of code, and change just 4 constants for each piece type, that is _not_ OK. That's the point.
For the Fruit/Crafty question, Nope. The algorithms are different. The implementations are different. In fact, the basic "concepts" are different, because I have two. One that expects to be given a move to test, one that expects to be given a position and a target square, where the move has already been made. If they are "joined at the hip" then you should find _some_ code that is identical. There is none.
For the Fruit/Crafty question, Nope. The algorithms are different. The implementations are different. In fact, the basic "concepts" are different, because I have two. One that expects to be given a move to test, one that expects to be given a position and a target square, where the move has already been made. If they are "joined at the hip" then you should find _some_ code that is identical. There is none.
> But if you copy a couple of hundred lines of code, and change just 4 constants for each piece type, that is _not_ OK. That's the point.
That might work. But I thought we'd got past the PST issue being anything to do with alleged code copying? It's now in the realm of alleged data copying. Writing a Microsoft Excel macro to produce the PST values is also a possibility, then we would not have code copying we would have idea copying, a Fruit idea re-implemented in a spreadsheet.
I don't think you can claim that 12 x 64 numbers represent an "idea". They are far closer to "code" than to "idea" because they are all added, subtracted, multiplied and divided in various ways, and reduced to a single number that represents the goodness or badness of a specific position. Change just one number and you change the personality of that program. Ideas are abstract. A large table of numbers is anything but abstract.
In another post on this forum Ed said he saw some pseudo-code for how hash tables worked in an early 90's ICGA Journal. He looked at it and then implemented the idea in Rebel.
Something similar could have happened here:
"Vas looked at the Fruit code, saw how PSTs were generated and implemented the idea, with some changes, in a spreadsheet".
Neither you or I can prove he did this or if he copied Fruit code. In either case he kept both at home and did not ship this PST generating thing, whatever it was, just the output of it (=PST numbers), in Rybka. Go back to your own statement -> "There was NO CODE COPYING for the PST issue. NONE. NADA. ZILCH. ZIPPO. " <- you were spot on with that.
Something similar could have happened here:
"Vas looked at the Fruit code, saw how PSTs were generated and implemented the idea, with some changes, in a spreadsheet".
Neither you or I can prove he did this or if he copied Fruit code. In either case he kept both at home and did not ship this PST generating thing, whatever it was, just the output of it (=PST numbers), in Rybka. Go back to your own statement -> "There was NO CODE COPYING for the PST issue. NONE. NADA. ZILCH. ZIPPO. " <- you were spot on with that.
You are COMPLETELY missing the point. Two programs can have hash tables that are functionally equivalent. That is, they both provide a "best move" to search first, they both provide for cutoffs when the table value compares favorable to the search bounds and the depth is adequate. But that is the "functional equivalence" part. Nobody cares. But how the tables are implemented? Some use one table with multiple buckets for entries / address collisions. Some use two tables ala' Belle's algorithm. Some use both. The replacement strategies vary all over the place. The table entries are not the same. The contain the same information, but arranged differently. It is not convenient to store negative numbers in an odd number of bits. Some do. Some bias all scores with a constant large enough to guarantee that the result is positive, and then they subtract that same constant after they retrieve the value but before using it. Some store multiple scores. Some store a static evaluation. That is where things differ. Wildly.
I would challenge you to study Fruit's PSTs and then produce a spread sheet that produces those numbers. Without having Fruit's PSTs right beside you so that you can continue to tweak the cells until you get a dead match. That's a completely bogus argument, and anyone using it is just grasping at straws.
As I said, there was no "code copying". Because vas took the _output_ of the fruit PST initialization code and copied _that_.
I would challenge you to study Fruit's PSTs and then produce a spread sheet that produces those numbers. Without having Fruit's PSTs right beside you so that you can continue to tweak the cells until you get a dead match. That's a completely bogus argument, and anyone using it is just grasping at straws.
As I said, there was no "code copying". Because vas took the _output_ of the fruit PST initialization code and copied _that_.
> I would challenge you to study Fruit's PSTs and then produce a spread sheet that produces those numbers.
?? I said you'd look at the Fruit code and re-implement the PST formula, with some changes, in a spreadsheet. This removes any code copying issue and makes it idea copying. What you are saying is that copying this idea makes Rybka a derivative of Fruit, which is something that an impartial third party needs to decide, in my opinion.
You realize that this "formula" is not a formula in any normal sense? That it is a bunch of lines of C code with multiple loops? You won't get this kind of PST matching unless you intentionally use the 64 fruit values as a starting point. And at that point, you are no longer using an idea...
> 64 fruit values as a starting point. And at that point, you are no longer using an idea...
If these 64 values were used to make calculations in the Rybka PSTs (and I don't know if they were or not) .. is that a bad thing? Is there a sign saying "do not use these numbers under pain of death" on them? I guess what I am asking is are they copyrighted, those numbers, so that no-one else can use them? (If they were used at all).
Good question. My only answer is that we have always used ideas from others. Where ideas are abstract expressions. For example, back in the 70's, I had an idea that instead of "praying that my program would find a way out of a big fail low before time ran out, I would just note that the score had dropped and used more time." I explained that idea at a CC paper session we had the next year. The idea "if your score drops, use more time to try to bring it back up. If it drops a lot, use even more time." That can translate to many different implementations. In CB, we detected "positional drops" (score dropped by by less than 0.4), "losing a pawn", "losing a piece" and "losing a bunch". Those "thresholds" are uniq in Cray Blitz. Others just decided "if the score drops by .5 pawns, we extend the time limit." In Crafty, if the score drops _any_ I extend the time, even .01...
As you can see, the "idea" is quite abstract. The implementations are unique and vary all over the place. DB did it a different way. If the "shape" of the tree changed dramatically, indicating that move ordering is suddenly not working as well as it should, they would extend the time even before the score dropped, because that "shape change" was often a precursor to a "score change." At times, the score change was +, but at times -.
The difference between idea and implementation is significant.
As you can see, the "idea" is quite abstract. The implementations are unique and vary all over the place. DB did it a different way. If the "shape" of the tree changed dramatically, indicating that move ordering is suddenly not working as well as it should, they would extend the time even before the score dropped, because that "shape change" was often a precursor to a "score change." At times, the score change was +, but at times -.
The difference between idea and implementation is significant.
> The difference between idea and implementation is significant.
I 100% agree with this Bob. Back during the Rybka 4.1 beta testing I ran the following three cutechess matches for fun:
TC 40/5s+1s repeating, three 100 game matches:
Score of Rybka1.0beta-w32 vs Fruit21-w32: 50 - 14 - 36
ELO difference: 131
Score of Rybka1.0beta-x64 vs Fruit21-w32: 59 - 5 - 36
ELO difference: 210
Score of Rybka1.0beta-x64 vs Fruit21-x64: 68 - 8 - 24
ELO difference: 241Amazing that these allegedly copied 64 numbers work so much better in Rybka. I would have no idea of how to get so much more value out of these numbers than Fruit did, could be in the different implementation?
The difference is _outside_ the numbers. PSTs are just a first-level approximation of the final evaluation. Then you have individual piece evaluations, commonly called first-order evaluations. Then you have second-order piece evaluations where you now evaluate pieces that are coordinating. For example, two pieces closer to your king is far worse than just 1. 3 is far worse than just 2.
Nobody has not said he didn't change anything. You can find two consecutive versions of Crafty that are almost 200 Elo apart. One before we did the evaluation tuning using our cluster, the second after a few hundred million games had been played. Yet the code is the same. We just tweaked scores here and there, maybe deleted something that didn't look so good here, added a few lines over there. And Bingo, after considerable effort, and several months of work, +200.
That's no different than the fruit->rybka case.
The problem here is that programmers object to someone copying someone else's code, and taking a _huge_ short-cut through the software life-cycle. Starting with the best open-source program lets you start, from day 1, working on improvements, where otherwise you would spend day 1 just figuring out where to start with the design. And then the coding. And then all the testing and debugging. Just to get to fruit 2.1's "state of the art".
That's the issue here.
Nobody has not said he didn't change anything. You can find two consecutive versions of Crafty that are almost 200 Elo apart. One before we did the evaluation tuning using our cluster, the second after a few hundred million games had been played. Yet the code is the same. We just tweaked scores here and there, maybe deleted something that didn't look so good here, added a few lines over there. And Bingo, after considerable effort, and several months of work, +200.
That's no different than the fruit->rybka case.
The problem here is that programmers object to someone copying someone else's code, and taking a _huge_ short-cut through the software life-cycle. Starting with the best open-source program lets you start, from day 1, working on improvements, where otherwise you would spend day 1 just figuring out where to start with the design. And then the coding. And then all the testing and debugging. Just to get to fruit 2.1's "state of the art".
That's the issue here.
> You can find two consecutive versions of Crafty that are almost 200 Elo apart. One before we did the evaluation tuning using our cluster, the second after a few hundred million games had been played. Yet the code is the same.
Ok so we have two crafty's that are just about the same code-wise but play very differently due to tuning (well done btw).
> That's no different than the fruit->rybka case.
That's what we haven't seen yet, anything that makes the rybka code the same code as fruit, in the same way that your two versions of crafty were just about the same.
I really don't follow that. We've been on the PST copying for a couple of days, and that seems like a slam dunk after recent analysis... my 69.4 multiplier, for example. That is just not going to happen with two independent programs...
> That is just not going to happen with two independent programs...
Rybka and Fruit cannot be totally independent because Vas said he went through Fruit and took many things. But you were suggesting that Rybka and Fruit were as dependent on eachother as your two crafty's and there is no evidence for that at all.
No evidence at all? PSTs from Fruit are trivially multiplied to produce those in Rybka. Rybka eval terms mimic fruit _perfectly_. I really don't get what you are trying to say.
I'm saying that Rybka and Fruit are NOT independent, because Vas already said he took many things from Fruit. Why are you saying that they are or should be independent?
All of the top modern engines are co-dependent in some form or another. Even crafty and fruit. You took from Fruit (not like Vas "going backwards and forwards taking many things") but in another way.
When you used open source fruit as an opponent (albeit compiled with your libs to allow it to work in your cluster), you were effectively using a brute-force approach to suck the knowledge out of fruit, it's chess strength in the form of games against crafty, and then you subsequently used this knowledge to tune crafty to make it play better. You wouldn't have got such a large elo boost if you had used say gnuchess as an opponent. The testing that you did is a good idea in my opinion, look at the elo boost you got.
Your crafty program has gained a lot from fruit. *If* (I don't know) you've done the same brute-force cluster approach using crafty against open source Stockfish (again with Stockfish compiled using your cluster support libs), then you've taken from Stockfish too. If you didn't have these stronger engines to pit crafty against, you would not have achieved the elo improvements that you've made in crafty through cluster tuning.
Imagine how you would feel if crafty had been at the top for the last five years using this approach and then a bunch of individuals came along and said "Ugh, this Hyatt is at the top for five years because he cluster-tuned his crafty against all the best engines out there, he used their outputs to make his program play better, it makes us sick!"
All of the top modern engines are co-dependent in some form or another. Even crafty and fruit. You took from Fruit (not like Vas "going backwards and forwards taking many things") but in another way.
When you used open source fruit as an opponent (albeit compiled with your libs to allow it to work in your cluster), you were effectively using a brute-force approach to suck the knowledge out of fruit, it's chess strength in the form of games against crafty, and then you subsequently used this knowledge to tune crafty to make it play better. You wouldn't have got such a large elo boost if you had used say gnuchess as an opponent. The testing that you did is a good idea in my opinion, look at the elo boost you got.
Your crafty program has gained a lot from fruit. *If* (I don't know) you've done the same brute-force cluster approach using crafty against open source Stockfish (again with Stockfish compiled using your cluster support libs), then you've taken from Stockfish too. If you didn't have these stronger engines to pit crafty against, you would not have achieved the elo improvements that you've made in crafty through cluster tuning.
Imagine how you would feel if crafty had been at the top for the last five years using this approach and then a bunch of individuals came along and said "Ugh, this Hyatt is at the top for five years because he cluster-tuned his crafty against all the best engines out there, he used their outputs to make his program play better, it makes us sick!"
You want to make that "I took many things" to mean something _other_ than "I copied large pieces of fruit's code." If he only took ideas, the two programs _would_ be independent. They would not use the same type of flag structure.. Rybka 1.6.1 would not have used my unique finite-state-machine approach to ordering moves and choosing the next one to search.
Ideas are abstract. Implementing them takes many lines of code. Code that won't look anything alike unless the code is copied. I give relatively simple programming assignments every week. And not a single program looks even to be "close" to another program. I'll remind you, for a given functional description of a program, there are an _infinite_ number of different source programs that will produce _exactly_ that functional behavior. What's the chances two people pick the _identical_ syntactical construction out of that infinite set of choices? Pretty damned low.
Ideas are abstract. Implementing them takes many lines of code. Code that won't look anything alike unless the code is copied. I give relatively simple programming assignments every week. And not a single program looks even to be "close" to another program. I'll remind you, for a given functional description of a program, there are an _infinite_ number of different source programs that will produce _exactly_ that functional behavior. What's the chances two people pick the _identical_ syntactical construction out of that infinite set of choices? Pretty damned low.
You keep going on about this "Rybka 1.6.1" being a ripoff of crafty and that it participated in CCT6, an event held under ICGA rules so it's relevant to the ICGA inquiry. Then I check the accusers own wiki (Harvey recently posted a link to it) and see:
http://chessprogramming.wikispaces.com/Rybka-Fruit+Controversy
"2004
In January, Rybka competed in CCT6, the largest online author-based tournament (a posting of Rajlich's from time gives a "Rybka 1.3" as a version number). In April, Rybka 1.5 enters the privately-run author-based tournaments Chess War V and Le Systeme du Saison. In October(?), Rybka 1.6 enters Chess War VI and AEGT 3. Rybka 1.6 also competed in Chess War VII."
So it definitely was a different Rybka (1.3) than the one that you disassembled (1.6.1) which participated in CCT6. You completely failed to mention this in our thread where we discussed it, and it explains why you won't check the output of the binary you have against the moves that Rybka 1.3 played in CCT6, because you *know* they won't match.
There is the (remote due to it's performance) possibility you mention that Vas copied crafty and entered it in CCT6, but there are at least two others:
1) Rybka 1.3 was completely different from 1.6.1 and Vas moved to a crafty codebase after Rybka 1.3, erroneously entering these in Chess War (but no crafty code in the ICGA ruled CCT6 though)
2) As Trotsky pointed out in another post, Vas himself did not publish this Rybka 1.6.1 on a link available to the general public, so we can't check it. This Rybka 1.6.1 which you have, cannot reliably ("provenance" I believe he said a court called it) be traced back to Vas, it could have been created by some Rybka-hater (which I know that *you* know there are many of) that hacked a crafty to make it look like a Rybka
You and I don't know, but the standard of evidence needs to be higher.
You won't remember, but you and I first conversed via email over 15 years ago where you kindly helped me get crafty running on an IBM mainframe, there was initially some problem with the inline asm on a big-endian machine, IIRC. You were very helpful, my deep respect for you is turning into deep disagreement over the way you are handling these current issues in trying to cover all dissent, I know that you don't care but I find it unfortunate.
http://chessprogramming.wikispaces.com/Rybka-Fruit+Controversy
"2004
In January, Rybka competed in CCT6, the largest online author-based tournament (a posting of Rajlich's from time gives a "Rybka 1.3" as a version number). In April, Rybka 1.5 enters the privately-run author-based tournaments Chess War V and Le Systeme du Saison. In October(?), Rybka 1.6 enters Chess War VI and AEGT 3. Rybka 1.6 also competed in Chess War VII."
So it definitely was a different Rybka (1.3) than the one that you disassembled (1.6.1) which participated in CCT6. You completely failed to mention this in our thread where we discussed it, and it explains why you won't check the output of the binary you have against the moves that Rybka 1.3 played in CCT6, because you *know* they won't match.
There is the (remote due to it's performance) possibility you mention that Vas copied crafty and entered it in CCT6, but there are at least two others:
1) Rybka 1.3 was completely different from 1.6.1 and Vas moved to a crafty codebase after Rybka 1.3, erroneously entering these in Chess War (but no crafty code in the ICGA ruled CCT6 though)
2) As Trotsky pointed out in another post, Vas himself did not publish this Rybka 1.6.1 on a link available to the general public, so we can't check it. This Rybka 1.6.1 which you have, cannot reliably ("provenance" I believe he said a court called it) be traced back to Vas, it could have been created by some Rybka-hater (which I know that *you* know there are many of) that hacked a crafty to make it look like a Rybka
You and I don't know, but the standard of evidence needs to be higher.
You won't remember, but you and I first conversed via email over 15 years ago where you kindly helped me get crafty running on an IBM mainframe, there was initially some problem with the inline asm on a big-endian machine, IIRC. You were very helpful, my deep respect for you is turning into deep disagreement over the way you are handling these current issues in trying to cover all dissent, I know that you don't care but I find it unfortunate.
One key statement: Using fruit as an opponent "sucked the knowledge out of fruit." But notice I did not "suck any code out of fruit." This is about copying code, which is (1) in direct violation of ICGA rule 2, which is why they investigated and then handed down penalties; (2) in violation of copyright laws, which is why the FSF is looking into the stuff right now.
I do not see why it is so hard to differentiate "idea" from "implementation of an idea". They are _not_ the same thing. Ideas are abstract, and are often quite easy to come up with. They can be hellishly hard to implement, particularly when the code has to be as efficient as possible to avoid dragging the search speed down and hurting more than the idea helps.
For example, "I want to build a model airplane with a gas engine and use a radio transmitter to control the flight surfaces and engine speed." That was a pretty simple idea many years ago. Implementing that idea took years and led to a _huge_ industry today. There are dozens of transmitter/receiver/servo manufacturing companies scattered all over the world, hundreds of model airplane kit and engine manufacturing companies scattered all over the world. And thousands of companies that make accessories for them. The "idea" was easy to come up with. The "implementation" was a many-year exercise. There is an enormous difference between "idea" and "code". Here, we claim that Rybka took _code_. Not _ideas_.
For the provenance of Rybka, we had a tester for a well-known rating list provide the code. We had at least 3 early versions of Rybka. I have these three on my laptop:
Rybka-1.5.32.zip
Rybka-1.4.32.zip
Rybka-1.6.1.32.zip
The tester received them _directly_ from Vas. And I do not believe that anyone doubts that that happened...
I do not see why it is so hard to differentiate "idea" from "implementation of an idea". They are _not_ the same thing. Ideas are abstract, and are often quite easy to come up with. They can be hellishly hard to implement, particularly when the code has to be as efficient as possible to avoid dragging the search speed down and hurting more than the idea helps.
For example, "I want to build a model airplane with a gas engine and use a radio transmitter to control the flight surfaces and engine speed." That was a pretty simple idea many years ago. Implementing that idea took years and led to a _huge_ industry today. There are dozens of transmitter/receiver/servo manufacturing companies scattered all over the world, hundreds of model airplane kit and engine manufacturing companies scattered all over the world. And thousands of companies that make accessories for them. The "idea" was easy to come up with. The "implementation" was a many-year exercise. There is an enormous difference between "idea" and "code". Here, we claim that Rybka took _code_. Not _ideas_.
For the provenance of Rybka, we had a tester for a well-known rating list provide the code. We had at least 3 early versions of Rybka. I have these three on my laptop:
Rybka-1.5.32.zip
Rybka-1.4.32.zip
Rybka-1.6.1.32.zip
The tester received them _directly_ from Vas. And I do not believe that anyone doubts that that happened...
> Rybka-1.5.32.zip
> Rybka-1.4.32.zip
> Rybka-1.6.1.32.zip
Did you look at 1.3, the one that is said to have played in CCT6?
I will remind you, again. We only looked at those to verify the "all versions of Rybka were 100% original, including the early versions." According to CCT6 records (I will try to find a link), rybka 1.6.1 was the program that was entered. Not 1.3... So I am not sure why 1.3 is in the discussion at all.
> So I am not sure why 1.3 is in the discussion at all.
I mentioned it as it is on your wiki.
http://chessprogramming.wikispaces.com/Rybka-Fruit+Controversy
"2004
In January, Rybka competed in CCT6, the largest online author-based tournament (a posting of Rajlich's from time gives a "Rybka 1.3" as a version number). In April, Rybka 1.5 enters the privately-run author-based tournaments Chess War V and Le Systeme du Saison. In October(?), Rybka 1.6 enters Chess War VI and AEGT 3. Rybka 1.6 also competed in Chess War VII."
I don't know who wrote that. Someone posted a crosstable, perhaps Peter Skinner, the TD. I am not sure. And it said 1.6.1 on the thing. If we were investigating for a CCT6 violation, we'd need to make certain that we looked at the version he used. Mark simply tested the veracity of "all versions of Rybka were 100% original including the early versions" and found that to be highly inaccurate.
> But notice I did not "suck any code out of fruit."
Yes I agree with that, but the code is only a means to an end, chess playing. You didn't take the means, you took the end. I guess it's where you draw the line.
We _clearly_ draw the line at copying code. If you don't want anyone to discover any of your secrets, you don't even release your binary so that nobody can do so. Or, you might try to obfuscate your output. Say mangle the node counter to mislead others? Or report a depth that is 3 plies shallower than what you really searched, to make it appear your search is more sophisticated? Wait. Didn't _someone_ do just that? :)
> Wait. Didn't _someone_ do just that? :)
Maybe he did it TWICE to convince some guy that he was splitting at the root
He _was_ splitting at the root on that occasion. 100% sure, 0% doubt...
I even have the output to absolutely prove it if you want to debate it further...
I even have the output to absolutely prove it if you want to debate it further...
Not necessary, Lukas already said he was in those earlier versions in another post on this forum. What you missed was my irony in that if he hadn't been splitting at the root but had been sending messages that he was (it's called obfuscation I believe) -- that would have been funny.
OK, misunderstood you. Yes, that would be yet another level of obfuscation...
point taken.
point taken.
I don't even begin to follow this. I played in hundreds of human chess tournaments over the years. And in each game I played, I learned something about the opponent. Is that somehow wrong? As opposed to breaking into their house, breaking into their computer, and looking to see what they have in the folder "opening preparation" under the player "Hyatt"???
There is a big difference between copying a block of code, and playing games against that same program and learning from them. You _still_ have to write the code for your own engine, playing games doesn't magically create new code out of thin air...
There is a big difference between copying a block of code, and playing games against that same program and learning from them. You _still_ have to write the code for your own engine, playing games doesn't magically create new code out of thin air...
All I am saying is that without Fruit you wouldn't have got the elo gain. Next time try playing against just crafty. But using this brute-force cluster approach you'll never be better than the engine you are taking from, it's like in the matrix when Morpheus said to Neo:
Before you tuned, Fruit was stronger than crafty. You tuned against Fruit (and others), and crafty became stronger than Fruit. To be stronger than the thing you are tuning against you need to invent something.
"their strength, and their speed, are still based in a world that is built on rules. Because of that, they will never be as strong, or as fast, as *you* "Before you tuned, Fruit was stronger than crafty. You tuned against Fruit (and others), and crafty became stronger than Fruit. To be stronger than the thing you are tuning against you need to invent something.
That's one of the more ridiculous statements I have seen. Rybka tests against itself. It doesn't "get any better"? :)
Sorry I don't understand? Are you saying that you only test crafty v crafty in your cluster? I think you test crafty v stronger engines.
I test crafty against weaker _and_ stronger opponents on my cluster. Vas, for the longest, and I assume to the present, played rybka vs rybka games only. Larry talked about his testing methodology a few years ago when he was playing 30-40K "hyper-bullet" games overnight on a 4 or 8 core box, rybka vs rybka. My point was two-fold. (1) testing of any kind is better than nothing; (2) not _everybody_ can find a program stronger than theirs to test. If that is the case, then the best can't test and get better???
I only have one engine in my current testing that is stronger than Crafty, that being stockfish. I have to have source to test because of some cluster library issues. I have several opponents that are weaker. Doesn't hurt a thing.
For many years, programmers had no choice but to test against their own engines. You don't think we got stronger each year, back then? If the newer engine beats the older one, then it is better, obviously.
I only have one engine in my current testing that is stronger than Crafty, that being stockfish. I have to have source to test because of some cluster library issues. I have several opponents that are weaker. Doesn't hurt a thing.
For many years, programmers had no choice but to test against their own engines. You don't think we got stronger each year, back then? If the newer engine beats the older one, then it is better, obviously.
> I only have one engine in my current testing that is stronger than Crafty, that being stockfish.
Perhaps you should only cluster test crafty-v-crafty then you will not be open to the challenge that you are taking the fruits of other engines. It's your call.
Btw, I firmly believe that ideas are more important than code (see last post from "McKoder" on this forum). The folks that wrote the Inception movie said the following which I agree with:
"What is the most resilient parasite? Bacteria? A virus? An intestinal worm? An idea. Resilient... highly contagious. Once an idea has taken hold of the brain it's almost impossible to eradicate."
My point is by testing against other programs you can only go so far, after that you need to invent something new.
I have never heard that remark made before. Otherwise, one could claim that the chess tournaments "exploit the opponent's knowledge" since you can learn things by going over the game later.
I think this is a ridiculous line of reasoning...
It is not even on the same planet with copying stuff...
I think this is a ridiculous line of reasoning...
It is not even on the same planet with copying stuff...
> I think this is a ridiculous line of reasoning...
You are, of course, entitled to your opinion.
Do you _really_ expect any commercial or amateur authors to pay any attention to such a suggestion? Do you really expect everyone to stop playing in tournaments so that they no longer "learn" from their opponents? The idea really is ridiculous. And that is _not_ "just an opinion."
> Do you really expect everyone to stop playing in tournaments so that they no longer "learn" from their opponents?
I am not talking about playing in tournaments I am talking about brute force cluster testing (millions of games, etc).
How, exactly, are they different, other than in quantity of games? 20 - 30 years ago, games were rare. And we went over them, and over them, looking to see what we could learn about our opponent, and about our own programs... Now we can play enough games that we can draw conclusions on results and not have to look at the individual games very often...
Again, this is completely unrelated to copying code...
Again, this is completely unrelated to copying code...
> Again, this is completely unrelated to copying code...
Of course it's unrelated, just as you admit the Rybka PST issue is unrelated to code copying, it's about alleged data copying.
With the massive cluster testing you do, however you want to spin it, you are sucking on the fruits of another programmers work to improve your program. There is no doubt. Incidentally, I think it's a good thing.
I train against them just as surely as they train against me. But I don't copy their code...
> This Rybka 1.6.1 which you have, cannot reliably ("provenance" I believe he said a court called it) be traced back to Vas,
Several early versions were forwarded to us still attached to the original emails from Vas so the provenance can be proved.
You have relentlessly and doggedly covered every post that has come out of this. I might be wrong, but my instincts are telling me that you do know something that you want to keep hidden, and you're covering every post to make sure no clue of it comes to surface or leads in the direction of where the bodies are buried.
I'll let some one else go over the technical details of your explanation.
I'll let some one else go over the technical details of your explanation.
I have not even covered 25% of the posts on this topic, so again, do your homework and get your numbers right...
Sum up all of my posts since I joined, and then sum up all of the posts on this topic, and I will not account for even 1/4 of them. Maybe not 10%.
Sum up all of my posts since I joined, and then sum up all of the posts on this topic, and I will not account for even 1/4 of them. Maybe not 10%.
1200 or more posts -since the ICGA capped the case with a verdict, and you fearfully hang on- to make sure of what?
You are obsessed with making sure nothing gets by. Is something still out there that threatens you and keeps you on, Bob.
Do you know of some data that may have been obfuscated and you want to be here to challenge it when it comes through? Somehow you feel you can't afford to leave the scene can you? It isn't a potential crime scene is it? I hope this is not the case for your sake- what that would do to your reputation and standing in the computer chess committee would be devastating and pale before anything that has happened to Vas.
You are obsessed with making sure nothing gets by. Is something still out there that threatens you and keeps you on, Bob.
Do you know of some data that may have been obfuscated and you want to be here to challenge it when it comes through? Somehow you feel you can't afford to leave the scene can you? It isn't a potential crime scene is it? I hope this is not the case for your sake- what that would do to your reputation and standing in the computer chess committee would be devastating and pale before anything that has happened to Vas.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill