Topic Rybka Support & Discussion / Rybka Support / Really About Fruit &Crafty code? Or vehicle for a Ban!
This isn’t about code! It “was” about getting rid of Rybka as a front runner. There is no mystery here.
For anyone to say- “Rybka” is really “Fruit” in disguise would be just pure madness. There was enough contention for years about the possible Fruit-Rybka connection. Vas indicated his interest in Fruit and spoke of it openly.
The ICGA knew of this for years and Rybka showed a clear and distinct advancement above and beyond any given chess engine -enough so that others were dying to know her secrets. (Here is where it is about code!). For five years she dominated computer chess tournaments . But all this has been said before.
The ICGA – and its list of inquisitors- had to find a loop hole to finally remove Rybka out of contention.
They found it with Bob Hyatt as their leading advocator-pouncing on vestiges of anything left from Rybka’s nascent history of development. Leaving the ICGA then to expel Rybka as unoriginal.
With Bob Hyatt, proclaiming Vas liken unto nothing short of a thieving, murdering rapist.
Red flags should have been raised all over the place!
Forget about the totality of the body of work-that doesn’t count for nothing anymore. Only what Bob Hyatt delineates as essential to his “truth” and the “truth” that will deliver the verdict they wanted from the ICGA.
But! Bob Hyatt and company aren’t here to fight against Rybka being reinstated into the ICGA. They know that isn’t going to happen.
Bob Hyatt is here to keep everyone from seeing how this dirty deal went down.
He has repeatedly-time and time again brought the argument back to stealing of lines of code. It continues to keep everyone off the trail of the real issue-a distraction from the real deal.
No precedent was ever established for this action to take place.
The bottom line here is that the ICGA and its list of supporters want to come out of this - all clean and smelling like roses –all pretty and clean with no hint of a stink. The ICGA is trying to salvage it’s integrity with the likes of Bob Hyatt leading the parade. This is extremely important to them.
However, this may be , and no matter what Hyatt says or what the ICGA proclaims – Rybka remains, intact, an original body of work.
Bob Hyatt has a history of seeing himself as the last word in computer chess programming- discarding out right others opinions for his own- if anyone shows him where his argument is weak he will refute it with semantics that will confirm his own biases. He could be staring the truth in the face and his intentions remain-to make Vas look like a villainous rapacious scoundrel of the lowest order. Yet I cannot hear him for his actions are speaking much too loud for me.
For anyone to say- “Rybka” is really “Fruit” in disguise would be just pure madness. There was enough contention for years about the possible Fruit-Rybka connection. Vas indicated his interest in Fruit and spoke of it openly.
The ICGA knew of this for years and Rybka showed a clear and distinct advancement above and beyond any given chess engine -enough so that others were dying to know her secrets. (Here is where it is about code!). For five years she dominated computer chess tournaments . But all this has been said before.
The ICGA – and its list of inquisitors- had to find a loop hole to finally remove Rybka out of contention.
They found it with Bob Hyatt as their leading advocator-pouncing on vestiges of anything left from Rybka’s nascent history of development. Leaving the ICGA then to expel Rybka as unoriginal.
With Bob Hyatt, proclaiming Vas liken unto nothing short of a thieving, murdering rapist.
Red flags should have been raised all over the place!
Forget about the totality of the body of work-that doesn’t count for nothing anymore. Only what Bob Hyatt delineates as essential to his “truth” and the “truth” that will deliver the verdict they wanted from the ICGA.
But! Bob Hyatt and company aren’t here to fight against Rybka being reinstated into the ICGA. They know that isn’t going to happen.
Bob Hyatt is here to keep everyone from seeing how this dirty deal went down.
He has repeatedly-time and time again brought the argument back to stealing of lines of code. It continues to keep everyone off the trail of the real issue-a distraction from the real deal.
No precedent was ever established for this action to take place.
The bottom line here is that the ICGA and its list of supporters want to come out of this - all clean and smelling like roses –all pretty and clean with no hint of a stink. The ICGA is trying to salvage it’s integrity with the likes of Bob Hyatt leading the parade. This is extremely important to them.
However, this may be , and no matter what Hyatt says or what the ICGA proclaims – Rybka remains, intact, an original body of work.
Bob Hyatt has a history of seeing himself as the last word in computer chess programming- discarding out right others opinions for his own- if anyone shows him where his argument is weak he will refute it with semantics that will confirm his own biases. He could be staring the truth in the face and his intentions remain-to make Vas look like a villainous rapacious scoundrel of the lowest order. Yet I cannot hear him for his actions are speaking much too loud for me.
![[mx]](/mwf/v12/flags/mx.png "Mexico")
> It “was” about getting rid of Rybka as a front runner
You mean about getting rid of Vas as a front runner.
Fact is, if Vas created a new engine from scratch and proved it was clean, they still wouldn't allow him to participate because he still has to prove that Rybka 1.6.1 is clean...
Wasn't 1.6.1 an experimental version? Learning model if you will! It wasn't even supposed to see the light of day. He never entered it into any competition!?
> He never entered it into any competition!?
He entered it into CCT6, a tournament that had the same ICGA rules of the WBCCC.
My point is, though, that if Vas gave them the source code of Rybka 4 and they checked it was clean, they wouldn't lift the ban since they don't care about R4 or any newer version, they only checked 1.6.1, 1.0 Beta and 2.3.2a, they'd want evidence of his innocence for these "relevant" engines.
They didn't ban the engine, the banned the person.
You did a good job with your description of how someone is painted to look alike a cheat. But this is all wordplay, and to be exact, it's undemocratic and inhuman diffamation. Alone the singling out of a singular individual is against the basics of our legal system. So, yes, this here is more a sort of lynch system. Even the place is the same as decades agon. NB Alabama in the South of the USA. At the time and as we see it's still practiced that a single propagandist or a small group is deciding without any legal authority that a scapegoat should be chased around. There is no idea of appealing to a legal court. No, that would no longer be a public Lynching. We all should show our solidarity to Vas and his wife. I would pay 100 or even 200 € for a Rybka 5 with the cover and photo Multiple World Champion Vas Rajlich and his wife Iweta. That program must not be stronger overall than 4 or 4.1. It should just beat Houdini 1.5a, just for fun.
I had to edit it because of the typo Typka. It's Rybka.
I had to edit it because of the typo Typka. It's Rybka.
When it all comes down to it -its all word play-
Finding discrepancies in Watkins report is a first step if you can get someone to listen. But the ICGA's intention from the outset was to secure a ban on Vas. You will not change that no matter what you find. However, what you will do is succeed in showing how dirty they really are- and that is what it is all about. Nothing less and nothing more. Their integrity is the only thing you can go after.
Finding discrepancies in Watkins report is a first step if you can get someone to listen. But the ICGA's intention from the outset was to secure a ban on Vas. You will not change that no matter what you find. However, what you will do is succeed in showing how dirty they really are- and that is what it is all about. Nothing less and nothing more. Their integrity is the only thing you can go after.
![[us]](/mwf/v12/flags/us.png "United States")
Last time I looked, _every_ trial held "singles out a single person." "group trials" are _very_ rare...
Well that reply kinda circumvented the meat of the issue presented by the OP eh wot?
![[nl]](/mwf/v12/flags/nl.png "Netherlands")
It “was” about getting rid of Rybka as a front runner. There is no mystery here.
Exactly, spot on. Unfortunately for them, there is one forum which refuses to accept their dirty tricks, a forum that starts to detonate the so called 'evidence'. Which makes them very nervous. The case gets weaker and weaker and weaker, which forces them to react, to silence the opposition. While verbal bullshit might work on other forums, it is not working here.
Bob Hyatt is here to keep everyone from seeing how this dirty deal went down.
If the case was water tight, they wouldn't discuss opposing views, using verbal diarrhea, over here. The more they discuss it, the weaker it gets. It is very telling that they do not understand this. If your case is 100% sure, you celebrate and leave this forum alone. But apparently it is not.....
This is just a very, very weak case put forward mostly by Rybka haters and people with commercial interests. And the guys with commercial interests profited big time from Fruit and Rybka. Last thing I am going to say about this: they are a bunch of hypocrites. Period.
Now I am going back to my nice hobby: legally 'stealing' from euro, pound sterling and dollar holders, thanks to our central bankers And there is no ICGA, judge or government who will stop me from doing this, ha!
Exactly, spot on. Unfortunately for them, there is one forum which refuses to accept their dirty tricks, a forum that starts to detonate the so called 'evidence'. Which makes them very nervous. The case gets weaker and weaker and weaker, which forces them to react, to silence the opposition. While verbal bullshit might work on other forums, it is not working here.
Bob Hyatt is here to keep everyone from seeing how this dirty deal went down.
If the case was water tight, they wouldn't discuss opposing views, using verbal diarrhea, over here. The more they discuss it, the weaker it gets. It is very telling that they do not understand this. If your case is 100% sure, you celebrate and leave this forum alone. But apparently it is not.....
This is just a very, very weak case put forward mostly by Rybka haters and people with commercial interests. And the guys with commercial interests profited big time from Fruit and Rybka. Last thing I am going to say about this: they are a bunch of hypocrites. Period.
Now I am going back to my nice hobby: legally 'stealing' from euro, pound sterling and dollar holders, thanks to our central bankers
And there is no ICGA, judge or government who will stop me from doing this, ha!
Could you research how much it is actually if I want to buy Alabama. I want to build a memorial for the lynched black people there. You see, I am not following the hypocrisy that I use PC words for the lynched and still being a racist. That's how they are cheating over there.
> Unfortunately for them, there is one forum which refuses to accept their dirty tricks, a forum that starts to detonate the so called 'evidence'. Which makes them very nervous. The case gets weaker and weaker and weaker, which forces them to react, to silence the opposition. While verbal bullshit might work on other forums, it is not working here.
Yes the wrecking crew is doing a pretty effective job here :-)
And Bob keeps biting, repeating same old _read_the_evidence and _you're_too_stupid_to_understand arguments.
Meanwhile he keeps losing ground fast and David probably deeply regrets putting him and his cronies in charge of the panel.
It was a simple plot: get rid of the guy who is way, way in front, try to convince the fans, get them leave the ship and voila, job done. Everybody with a few IQ can understand this. Simply the fact that there is an organised 'read all stuff on forums and verbally refute any counter argument' just shows what a bunch of hypocrites they are.
David probably deeply regrets putting him and his cronies in charge of the panel.
I don't think Levy will regret it, he has no clue about many things. What is quite obvious, though, is that during the next WCCC's there will be lots of Fruit/Rybka/Robbo stuff in participating programs. Their quest is just as hopeless as the UCI and WADA trying to ban doping. The cycling fans do not care about UCI and WADA, nor do the chess players care about ICGA.
David probably deeply regrets putting him and his cronies in charge of the panel.
I don't think Levy will regret it, he has no clue about many things. What is quite obvious, though, is that during the next WCCC's there will be lots of Fruit/Rybka/Robbo stuff in participating programs. Their quest is just as hopeless as the UCI and WADA trying to ban doping. The cycling fans do not care about UCI and WADA, nor do the chess players care about ICGA.
![[de]](/mwf/v12/flags/de.png "Germany")
> Their quest is just as hopeless as the UCI and WADA trying to ban doping. The cycling fans do not care about UCI and WADA, nor do the chess players care about ICGA.
Well at least you correctly identify the equivalent infraction. Apart from a few die-hard deadenders on this site, the world has moved on with the understanding that Rybka began as a hacked-up derivative of another program. And the world also understands that Rybka had its own original ideas and development worthy of appreciation. And yeah, somehow we'll all get through this. Most of us, anyway.
Jeremy
> And the world also understands that Rybka had its own original ideas and development worthy of appreciation.
No, certain parties do not concede this at all without adding all sorts of insidious disclaimers and accusations, which confirms the fundamental dishonesty of their enterprise.
You have this backward. The "dishonesty" came from Vas. NOT the ICGA investigative process.
So you say.
So the _evidence_ says...
So the evidence you helped cook up says.
So the evidence that a _bunch_ of us "cooked up" says...
> So the evidence that a _bunch_ of us "cooked up" says...
Complied with would be more accurately put, Bob.
What's with the word game? This was a group of people looking into the complaint. Not a single person.
How many people collecting data? Does Dr. Watkins and Zack now contain a "bunch" multitudes?
They were not the only two, but don't let small details derail your objections...
No, it confirms their legitimate contempt toward a developer who stole his first 2800 ELO points from someone else without attribution. Their enterprise seems perfectly honest to me. You simply disagree with the outcome of the process.
+1
I must say -IMHO-whatever Vas did or did not do- I dare say it certainly appears to me as well that Vas had to be 'neutralized' by his competitors. I also smell a certain degree of envy along with whatever righteous indignation there was on the part of some (many?) of the detractors. That is just IMHO considering the tonality of some of the posters and past practice treatment of other offenders of such a situation. Seems disproportionate to me IMHO.
What for? His end-user product is already not #1 anymore, and no one knows how the cluster actually performs. At least 3 other "clean" engines are within biting range. Competition for the top is stronger than ever -- it's a great time for competitive CC. There's absolutely no basis for this notion.
The simpler explanation is: the ICGA wanted to enforce their rules, and pursued a violation.
Jeremy
The simpler explanation is: the ICGA wanted to enforce their rules, and pursued a violation.
Jeremy
Competition for the top is stronger than ever -- it's a great time for competitive CC.
Fully agreed. But the ICGA doesn't recognize this, on the contrary: they are trying to prevent this. Just like the programmers who voted for the Rybka ban. They want to put the clock 5 year backwards. The chess players won't care, they will take what is best.
Fully agreed. But the ICGA doesn't recognize this, on the contrary: they are trying to prevent this. Just like the programmers who voted for the Rybka ban. They want to put the clock 5 year backwards. The chess players won't care, they will take what is best.
Er, bullshit?
It is not so much turning the clock back as much as having absolute power to vet members to their private club, and blackball those who display an unsporting propensity to win again and again by crushing margins, making it less and less feasible for competitors to travel to places like Japan on their own nickel only to face certain slaughter.
deep thought won for longer than Rybka. Why wasn't Hsu "black-balled"? Oh yes, he did not copy someone else's code and claim it to be his original work. Or maybe Slate/Atkin from 1970 to 1980. Wonder why they weren't black-balled, nobody could beat them. Oh yes, they did not copy someone else's code and claim it to be their original work.
Notice the trend there?
This discussion appears to be one carried on by a bunch of hillbillies, no internet connections, no newspapers, no email, no contact with the outside world. In short, no facts are uncovered, because you don't know how/where to look. And then you use that lack of information as justification for making comments that are so far out of touch with reality...
Winning does NOT get you banned. Already proven above. (and there are other good examples such as shredder which was untouchable for a few years as well). Cheating DOES get you banned. Copying code DOES get you banned. Lying during the investigative process DOES get you banned. Failure to follow published rules DOES get you banned.
There is a trend there, but I doubt you will see/accept it.
Notice the trend there?
This discussion appears to be one carried on by a bunch of hillbillies, no internet connections, no newspapers, no email, no contact with the outside world. In short, no facts are uncovered, because you don't know how/where to look. And then you use that lack of information as justification for making comments that are so far out of touch with reality...
Winning does NOT get you banned. Already proven above. (and there are other good examples such as shredder which was untouchable for a few years as well). Cheating DOES get you banned. Copying code DOES get you banned. Lying during the investigative process DOES get you banned. Failure to follow published rules DOES get you banned.
There is a trend there, but I doubt you will see/accept it.
Tell you what, I'll accept the 'hillbilly' jibe conditional on your accepting the 'charlatan', 'obsessive-compulsive', 'jackass', 'loser', 'pompous', 'blowhard', 'self-righteous', 'blackguard' and 'mountebank' jibes.
I was not intending to use such words to describe you, so I'll pass. BTW, you might notice I didn't call you a "hillbilly". I said you are carrying on as if you live in a total information blackout, much like hillbilly-types up in the mountains...
Guess it was too subtle...
Guess it was too subtle...
Well yee-haw to you too, sir. Actually, I am a San Francisco native. Grew up there. Not too many hillbillies there. Lots of very peculiar types but not the redneck variety.
![[fr]](/mwf/v12/flags/fr.png "France")
"Lying during the investigative process DOES get you banned. "
why then were you not banned from the investigative process?
you were lying about the constant multiplier that you claimed related fruit and rybka PSTs, this was a serious lie which must have misled many who trusted you, it's a repeated lie. It's a lie that entered the groupthink.
why then were you not banned from the investigative process?
you were lying about the constant multiplier that you claimed related fruit and rybka PSTs, this was a serious lie which must have misled many who trusted you, it's a repeated lie. It's a lie that entered the groupthink.
OK, to bring this back to planet earth. Taking Zach's report. And let's pick on one piece at a time. Are you saying that the knight PST values do not match between rybka 1.0 beta and fruit? no innuendos. No ifs, ands or buts. Just a simple direct question. Do they match or do they not? Then we will advance to bishops and beyond.
your serve...
your serve...
It appears that Vas simply took a few values from those PSTs, but not the PSTs themselves. Since you like to use this analogy, this is like reusing several of Fischer's main lines in the King's Indian, but not like taking all of his Greatest Games and reproducing them.
Going through the PST's there are a total of ten PST tables that Zach argues are parallel in Fruit/Rybka ....
E and EO ...
as explained elsewhere, the build of the tables depends on the patterns in the ramping arrays, and the evaluation weights applied. The evaluation weights are all different. The ramping arrays, if we accept Zach at face value, are the same.
Whather or not a particular PST pair show multiplicative similarity (ie you can multiply one by a constant to get the other) depends on the build method and whether the PST table is generated by using ONE or TWO DIFFERENT ramping arrays. The first produces the multiplicative similarity. The second does not.
pawns
there is one parallel pair of PSTs. It would show multiplicative matching if the values had not also been hit by a kludge factor. But they were. score 0-1
knights
two parallel pairs of PSTs. One shows multiplicative matching, the other does not. score 1-2
bishops
same situation as knights. score 2-3
rooks
matches. score 3-3
queen
Two parallel tables. One matches, one not. Score 4-4
king
two parallel tables
Neither match. Score 4-6
So a better formulation of the expression "all the PST tables match" would be "four out of the ten PST tables match".
It might have been advisable for you to have said that in the first place rather than use your exaggerated, dissembling untruthful factoids to try and influence people. And the Lord only knows how you tried in other ways to influence people on the hidden form wiki, because you have managed to REDACT the entire record of the wiki forum, so that we can't read it.
It would be even more advisable, and a much fairer methodology, to have concentrated on the ramping arrays which are the true cause of the similarity patterns. The PST tables are reflections of the ramping arrays plus noise from the different weights plus huge amounts of redundancy. The ramping arrays just represent simple formulas and the question then becomes: "are they worthy of copyright status"? A much simpler question. And one in which the complexity can't be used by you to dissemble, quite so easily.
E and EO ...
as explained elsewhere, the build of the tables depends on the patterns in the ramping arrays, and the evaluation weights applied. The evaluation weights are all different. The ramping arrays, if we accept Zach at face value, are the same.
Whather or not a particular PST pair show multiplicative similarity (ie you can multiply one by a constant to get the other) depends on the build method and whether the PST table is generated by using ONE or TWO DIFFERENT ramping arrays. The first produces the multiplicative similarity. The second does not.
pawns
there is one parallel pair of PSTs. It would show multiplicative matching if the values had not also been hit by a kludge factor. But they were. score 0-1
knights
two parallel pairs of PSTs. One shows multiplicative matching, the other does not. score 1-2
bishops
same situation as knights. score 2-3
rooks
matches. score 3-3
queen
Two parallel tables. One matches, one not. Score 4-4
king
two parallel tables
Neither match. Score 4-6
So a better formulation of the expression "all the PST tables match" would be "four out of the ten PST tables match".
It might have been advisable for you to have said that in the first place rather than use your exaggerated, dissembling untruthful factoids to try and influence people. And the Lord only knows how you tried in other ways to influence people on the hidden form wiki, because you have managed to REDACT the entire record of the wiki forum, so that we can't read it.
It would be even more advisable, and a much fairer methodology, to have concentrated on the ramping arrays which are the true cause of the similarity patterns. The PST tables are reflections of the ramping arrays plus noise from the different weights plus huge amounts of redundancy. The ramping arrays just represent simple formulas and the question then becomes: "are they worthy of copyright status"? A much simpler question. And one in which the complexity can't be used by you to dissemble, quite so easily.
Maybe the hidden wiki forum could be released with just the names redacted??
First, thanks for actually reading some of the information we posted. I do not think the panel thought the PST information itself was especially damning, although it does show strong evidence of copying (with some tuning). If you look at the stuff Zack supplied here:
http://icga.wikispaces.com/file/view/ZW_Rybka_Fruit.pdf
You will see all 768 PST values (64 squares times 6 piece types times 2 game phases) were generated using the exact same line arrays in both Fruit and Rybka, and these were added and multiplied by about 18 constants (some of the constants were zero). Other than a few pawn values, it is an exact match assuming only these 18 values were modified. This is not a coincidence. The line arrays and basic structure had to have been copied.
You do some strange comparison to see if arrays between the programs are direct multiples. You are right in that some are not (since there are different scaling factors in different directions, and several numbers are multiplied together and added). But the underlying pattern is the same, and unlike any other programs. The conclusion has to be that the Rybka PSTs were a derivative of the Fruit ones.
Please keep reading to see even more copying in the other evaluation features.
http://icga.wikispaces.com/file/view/ZW_Rybka_Fruit.pdf
You will see all 768 PST values (64 squares times 6 piece types times 2 game phases) were generated using the exact same line arrays in both Fruit and Rybka, and these were added and multiplied by about 18 constants (some of the constants were zero). Other than a few pawn values, it is an exact match assuming only these 18 values were modified. This is not a coincidence. The line arrays and basic structure had to have been copied.
You do some strange comparison to see if arrays between the programs are direct multiples. You are right in that some are not (since there are different scaling factors in different directions, and several numbers are multiplied together and added). But the underlying pattern is the same, and unlike any other programs. The conclusion has to be that the Rybka PSTs were a derivative of the Fruit ones.
Please keep reading to see even more copying in the other evaluation features.
Well actually you are wrong, indicating again that even at this late stage you (also Hyatt) didn't read the Zach report with respect to the PST tables at least with any degree of active reading and trying to understand the content.
First you said there were 6 tables. Actually there are 10 in Fruit and 11 in Rybkq, according to the Zach report. Why didn't you read it properly? I know where your "6" is from, it's an assumption based on piece type count, not on actual reading.
Second you said these 6 tables were a constant multiplier of each other. A cursory examination of the source included in Zach's paper would tell any programmer (you are a programmer aren't you?) that they were not multiplicative. As has been demonstrated by someone who actually did read the paper, properly, in an active way, thinking about what it meant. I know where your "multiplicative" is from, it is wishful and biased thinking, assumptive, based on prejudged guilt (well of course they multiply by a constant, he is guilty innit? and that is wot guilty people do innit?)
I accuse you and Bob Hyatt of not actively reading the documents but of skimming them, picking up on small segments, jumping to false and damning conclusions, repeating these false conclusions to others (specially selected and filtered to remove potential critics) and getting the falsities parroted back until they entered your groupthink version of the "truth".
In short, your process was completely corrupted to the point where the desired guilty verdict was quite inevitable.
First you said there were 6 tables. Actually there are 10 in Fruit and 11 in Rybkq, according to the Zach report. Why didn't you read it properly? I know where your "6" is from, it's an assumption based on piece type count, not on actual reading.
Second you said these 6 tables were a constant multiplier of each other. A cursory examination of the source included in Zach's paper would tell any programmer (you are a programmer aren't you?) that they were not multiplicative. As has been demonstrated by someone who actually did read the paper, properly, in an active way, thinking about what it meant. I know where your "multiplicative" is from, it is wishful and biased thinking, assumptive, based on prejudged guilt (well of course they multiply by a constant, he is guilty innit? and that is wot guilty people do innit?)
I accuse you and Bob Hyatt of not actively reading the documents but of skimming them, picking up on small segments, jumping to false and damning conclusions, repeating these false conclusions to others (specially selected and filtered to remove potential critics) and getting the falsities parroted back until they entered your groupthink version of the "truth".
In short, your process was completely corrupted to the point where the desired guilty verdict was quite inevitable.
I specifically started with knights. Which knight PST does not match? I assume you are talking about the opening vs endgame values???
do I have to hold your hand?
you can see quite easily which knight table has had its multiplicity destroyed, just look at Zachs Fruit source listing
you can see quite easily which knight table has had its multiplicity destroyed, just look at Zachs Fruit source listing
Pretty strange. I took his code and compiled and ran it, and then dumped the two PSTs for knights at the end. Did you do that?
edit: I will have to finish this later. My mouse (bluetooth) has gone ape and I can't copy/paste worth a flip. More later...
edit: I will have to finish this later. My mouse (bluetooth) has gone ape and I can't copy/paste worth a flip. More later...
OK. Mouse fixed.
I took Zach's code, _exactly_ as written. For Fruit, using fruit weights, it produces these numbers:
Opening:
-50 -40 -30 -25 -25 -30 -40 -50
-35 -25 -15 -10 -10 -15 -25 -35
-20 -10 0 5 5 0 -10 -20
-10 0 10 15 15 10 0 -10
-5 5 15 20 20 15 5 -5
-5 5 15 20 20 15 5 -5
-20 -10 0 5 5 0 -10 -20
-135 -25 -15 -10 -10 -15 -25 -135
Endgame:
-40 -30 -20 -15 -15 -20 -30 -40
-30 -20 -10 -5 -5 -10 -20 -30
-20 -10 0 5 5 0 -10 -20
-15 -5 5 10 10 5 -5 -15
-15 -5 5 10 10 5 -5 -15
-20 -10 0 5 5 0 -10 -20
-30 -20 -10 -5 -5 -10 -20 -30
-40 -30 -20 -15 -15 -20 -30 -40
I then took the fruit 2.1 source, and modified pst.cpp to dump just the knight values for opening and middlegame, after they are initialized. I got this:
Opening:
-50 -40 -30 -25 -25 -30 -40 -50
-35 -25 -15 -10 -10 -15 -25 -35
-20 -10 0 5 5 0 -10 -20
-10 0 10 15 15 10 0 -10
-5 5 15 20 20 15 5 -5
-5 5 15 20 20 15 5 -5
-20 -10 0 5 5 0 -10 -20
-135 -25 -15 -10 -10 -15 -25 -135
Endgame:
-40 -30 -20 -15 -15 -20 -30 -40
-30 -20 -10 -5 -5 -10 -20 -30
-20 -10 0 5 5 0 -10 -20
-15 -5 5 10 10 5 -5 -15
-15 -5 5 10 10 5 -5 -15
-20 -10 0 5 5 0 -10 -20
-30 -20 -10 -5 -5 -10 -20 -30
-40 -30 -20 -15 -15 -20 -30 -40
Step one. Can we agree they match exactly?
I then took Zach's "rybka pst generator" and ran it. And got this:
Opening:
-3492 -2798 -2104 -1757 -1757 -2104 -2798 -3492
-2440 -1746 -1052 -705 -705 -1052 -1746 -2440
-1388 -694 0 347 347 0 -694 -1388
-683 11 705 1052 1052 705 11 -683
-325 369 1063 1410 1410 1063 369 -325
-314 380 1074 1421 1421 1074 380 -314
-1366 -672 22 369 369 22 -672 -1366
-5618 -1724 -1030 -683 -683 -1030 -1724 -5618
Endgame:
-448 -336 -224 -168 -168 -224 -336 -448
-336 -224 -112 -56 -56 -112 -224 -336
-224 -112 0 56 56 0 -112 -224
-168 -56 56 112 112 56 -56 -168
-168 -56 56 112 112 56 -56 -168
-224 -112 0 56 56 0 -112 -224
-336 -224 -112 -56 -56 -112 -224 -336
-448 -336 -224 -168 -168 -224 -336 -448
Now, from Mark's disassembly of Rybka, which I can't seem to copy correctly, but which are in Mark's report, you find _exactly_ the same numbers. So what is "supposedly broken" that I miss???
The chances of those being identical except for the constant multipliers is vanishingly small...
I took Zach's code, _exactly_ as written. For Fruit, using fruit weights, it produces these numbers:
Opening:
-50 -40 -30 -25 -25 -30 -40 -50
-35 -25 -15 -10 -10 -15 -25 -35
-20 -10 0 5 5 0 -10 -20
-10 0 10 15 15 10 0 -10
-5 5 15 20 20 15 5 -5
-5 5 15 20 20 15 5 -5
-20 -10 0 5 5 0 -10 -20
-135 -25 -15 -10 -10 -15 -25 -135
Endgame:
-40 -30 -20 -15 -15 -20 -30 -40
-30 -20 -10 -5 -5 -10 -20 -30
-20 -10 0 5 5 0 -10 -20
-15 -5 5 10 10 5 -5 -15
-15 -5 5 10 10 5 -5 -15
-20 -10 0 5 5 0 -10 -20
-30 -20 -10 -5 -5 -10 -20 -30
-40 -30 -20 -15 -15 -20 -30 -40
I then took the fruit 2.1 source, and modified pst.cpp to dump just the knight values for opening and middlegame, after they are initialized. I got this:
Opening:
-50 -40 -30 -25 -25 -30 -40 -50
-35 -25 -15 -10 -10 -15 -25 -35
-20 -10 0 5 5 0 -10 -20
-10 0 10 15 15 10 0 -10
-5 5 15 20 20 15 5 -5
-5 5 15 20 20 15 5 -5
-20 -10 0 5 5 0 -10 -20
-135 -25 -15 -10 -10 -15 -25 -135
Endgame:
-40 -30 -20 -15 -15 -20 -30 -40
-30 -20 -10 -5 -5 -10 -20 -30
-20 -10 0 5 5 0 -10 -20
-15 -5 5 10 10 5 -5 -15
-15 -5 5 10 10 5 -5 -15
-20 -10 0 5 5 0 -10 -20
-30 -20 -10 -5 -5 -10 -20 -30
-40 -30 -20 -15 -15 -20 -30 -40
Step one. Can we agree they match exactly?
I then took Zach's "rybka pst generator" and ran it. And got this:
Opening:
-3492 -2798 -2104 -1757 -1757 -2104 -2798 -3492
-2440 -1746 -1052 -705 -705 -1052 -1746 -2440
-1388 -694 0 347 347 0 -694 -1388
-683 11 705 1052 1052 705 11 -683
-325 369 1063 1410 1410 1063 369 -325
-314 380 1074 1421 1421 1074 380 -314
-1366 -672 22 369 369 22 -672 -1366
-5618 -1724 -1030 -683 -683 -1030 -1724 -5618
Endgame:
-448 -336 -224 -168 -168 -224 -336 -448
-336 -224 -112 -56 -56 -112 -224 -336
-224 -112 0 56 56 0 -112 -224
-168 -56 56 112 112 56 -56 -168
-168 -56 56 112 112 56 -56 -168
-224 -112 0 56 56 0 -112 -224
-336 -224 -112 -56 -56 -112 -224 -336
-448 -336 -224 -168 -168 -224 -336 -448
Now, from Mark's disassembly of Rybka, which I can't seem to copy correctly, but which are in Mark's report, you find _exactly_ the same numbers. So what is "supposedly broken" that I miss???
The chances of those being identical except for the constant multipliers is vanishingly small...
I'm assuming a1 is on the top left of each matrix
so, the ratio of the matrix values for square a1, opening phase = 3492 / 50 = 69.84
tracking down to square a6, opening phase, the ratio = 314 / 5 = 62.8
since the a1 ratio (69.84) is not the same as the a6 ratio (62.8) there is no CONSTANT multiplier across the two opening matrices.
The cell multipliers are relative close but not identical.
The relative closeness is a function of the relative closeness of the ratio of the weights used, Fruit 5 equals Fruit 5 and Rybka 347 is close to Rybka 358
Fruit:
static const int KnightCentreOpening = 5;
static const int KnightRankOpening = 5;
Rybka:
static const int KnightCentreOpening = 347;
static const int KnightRankOpening = 358;
I can see we are going to have an argument about the meaning of the word identical .....
In fact the degree of loss of multiplicity is heavily related to the degree of relative difference of the weights. Bishop opening tables have their multiplicity broken by these weights ...
Fruit:
static const int BishopCentreOpening = 2;
static const int BishopBackRankOpening = 10;
static const int BishopDiagonalOpening = 4;
Rybka:
static const int BishopCentreOpening = 147;
static const int BishopBackRankOpening = 251;
static const int BishopDiagonalOpening = 378;
...
which will lead, I assume, to a wilder range of "constants"
so, the ratio of the matrix values for square a1, opening phase = 3492 / 50 = 69.84
tracking down to square a6, opening phase, the ratio = 314 / 5 = 62.8
since the a1 ratio (69.84) is not the same as the a6 ratio (62.8) there is no CONSTANT multiplier across the two opening matrices.
The cell multipliers are relative close but not identical.
The relative closeness is a function of the relative closeness of the ratio of the weights used, Fruit 5 equals Fruit 5 and Rybka 347 is close to Rybka 358
Fruit:
static const int KnightCentreOpening = 5;
static const int KnightRankOpening = 5;
Rybka:
static const int KnightCentreOpening = 347;
static const int KnightRankOpening = 358;
I can see we are going to have an argument about the meaning of the word identical .....
In fact the degree of loss of multiplicity is heavily related to the degree of relative difference of the weights. Bishop opening tables have their multiplicity broken by these weights ...
Fruit:
static const int BishopCentreOpening = 2;
static const int BishopBackRankOpening = 10;
static const int BishopDiagonalOpening = 4;
Rybka:
static const int BishopCentreOpening = 147;
static const int BishopBackRankOpening = 251;
static const int BishopDiagonalOpening = 378;
...
which will lead, I assume, to a wilder range of "constants"
:)
Nice try.
The problem is this: you have code to produce both tables. The _only_ difference in producing fruit or rybka is the 4 "seed numbers" that reflect the difference in material values each uses. The rest of the constants, namely rank, file, etc multipliers are the same. Only those 4 values at the top change. If two tables are connected by just 4 values that need to be changed to translate from one to another, they are "quite close" to be conservative. If you believe that two different programmers will come up with PSTs that are that connected, you probably believe that the lottery is a "sure thing."
The same idea holds for every table. Let's be generous and say "OK, maybe if just the knight PSTs were that closely connected, we could conclude "bad luck." But then bishops? Rooks? before long bad luck runs out. The coin is not fair. The dice are not fair. The laws of probability are being _badly_ violated, which means there is something wrong with the observed data. What is wrong is that the two are "joined at the hip." 12 tables if you count opening and endgame values, 12 * 64 numbers. 768 in total. The laws of probability won't say that is "just a random chance..."
That was the point of the PST analysis. Feel free to take Fruit's code and modify the constants to produce my PSTs. I don't even use those big negative numbers on the corner squares on the enemy's side of the board (ostensibly to deal with the case where you sac a bishop to draw the king to d8, then check on f7 and take the rook on h8 and assume you are a rook and pawn up, and a piece (bishop down). But you are also a knight down if the knight can't get out _right now_.
Point?
PSTs are as unique as anything else in a chess engine. And the PST data is not the _only_ data that was presented. Zach went into all sorts of comparisons for scoring for each piece type... showing code from fruit, and de-compiled code from Rybka...
Too much to wave this off as a serendipitous event.
Nice try.
The problem is this: you have code to produce both tables. The _only_ difference in producing fruit or rybka is the 4 "seed numbers" that reflect the difference in material values each uses. The rest of the constants, namely rank, file, etc multipliers are the same. Only those 4 values at the top change. If two tables are connected by just 4 values that need to be changed to translate from one to another, they are "quite close" to be conservative. If you believe that two different programmers will come up with PSTs that are that connected, you probably believe that the lottery is a "sure thing."
The same idea holds for every table. Let's be generous and say "OK, maybe if just the knight PSTs were that closely connected, we could conclude "bad luck." But then bishops? Rooks? before long bad luck runs out. The coin is not fair. The dice are not fair. The laws of probability are being _badly_ violated, which means there is something wrong with the observed data. What is wrong is that the two are "joined at the hip." 12 tables if you count opening and endgame values, 12 * 64 numbers. 768 in total. The laws of probability won't say that is "just a random chance..."
That was the point of the PST analysis. Feel free to take Fruit's code and modify the constants to produce my PSTs. I don't even use those big negative numbers on the corner squares on the enemy's side of the board (ostensibly to deal with the case where you sac a bishop to draw the king to d8, then check on f7 and take the rook on h8 and assume you are a rook and pawn up, and a piece (bishop down). But you are also a knight down if the knight can't get out _right now_.
Point?
PSTs are as unique as anything else in a chess engine. And the PST data is not the _only_ data that was presented. Zach went into all sorts of comparisons for scoring for each piece type... showing code from fruit, and de-compiled code from Rybka...
Too much to wave this off as a serendipitous event.
Bob, is your need to be right so great that your pride stands in the way of allowing you to even doubt for a second that you might be wrong? I wonder if there isn't something here that you do see that lets you know that you could most certainly have misunderstood Vasik Rajlich intentions and screwed the guy by misinterrupting his actions as some wanton act of unconscionable code copying- when you know it really wasn't that at all.
He copied code. There is no way to "misunderstand" that. There is no possible justification for doing so, then entering tournaments that explicitly forbid that, and then repeatedly denying that it happened after it was exposed...
Give me some idea of what you would consider a "justification" for his copying actions. Then we have something useful to discuss. We might not agree, but I will do my best to explain how the "programmers" feel about this...
Give me some idea of what you would consider a "justification" for his copying actions. Then we have something useful to discuss. We might not agree, but I will do my best to explain how the "programmers" feel about this...
> He copied code. There is no way to "misunderstand" that.
You encouraged him and others to use your code BOB! You're a control freak. The Fabien "stuff" is beginning to look more and more like a matter of interpretation. Rather than hard evidence.
Topic Rybka Support & Discussion / Rybka Support / Really About Fruit &Crafty code? Or vehicle for a Ban!
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill