ChessBase: A Gross Miscarriage of Justice in Computer Chess

[Rebel]

[...] [Would you say] I am 100% certain that Vas copied Fruit ?
[...] Ask Zach and Mark, they will not say such things.

I would say that it is (much) beyond a reasonable doubt that VR infringed Fruit's copyright with various versions of Rybka. Whether or not this is what you mean by "copied" is a different question.

Beyond reasonable doubt is not the same thing as 100%, which was the topic at hand.

1. Every action causes a reaction of the same magnitude.

I think this an understatement As the pendulum sways back and forth, over time more and more people get behind each end of its course, shoving it harder and harder.

Working principle in many parts in life. Take politics, the magnitude of the first action defines the magnitude of the reaction. The magnitude of the Vas' ban is huge and so is its counter reaction.

[BB+]

Beyond reasonable doubt is not the same thing as 100%, which was the topic at hand.

As Bob says, this "100%" convention is mere sophistry. Only God can judge in such a manner, and does so justly. Should we thus dispense with earthly justice? Your argument is rapidly becoming one against declaring anyone guilty of anything, ever.

The magnitude of the Vas' ban is huge and so is its counter reaction.

When speaking of hugeness, I think you need to back up about 6-7 years: Yes, the publication of Fruit 2.1 was huge.

[hyatt]

This has just reverted to yet another set of distortions and diversions.

100% certain. I am not even 100% certain I will die. It is POSSIBLE for one to live forever. Even though it hasn't happened. We have sea turtles that survive for > 500 years. So talking about 100% certainty is nonsense. I am not 100% certain that when I flip a coin, that I will get heads or tails. The coin CAN land on its edge. Time to move on. "Beyond a resonable doubt" is used for criminal trials. "A preponderance of the evidence" (which is essentially anything over 50%) is the standard for a civil trial. We were beyond BOTH of those standards.

[BB+]

I am not 100% certain that when I flip a coin, that I will get heads or tails.

Usually I don't spam threads, but this is one is sufficiently divergent that it just don't matter much. So onto the funnies.
http://dilbert.com/strips/comic/1998-01-13/

http://dilbert.com/strips/comic/1998-01-14/

[Rebel]

Beyond reasonable doubt is not the same thing as 100%, which was the topic at hand.

As Bob says, this "100%" convention is mere sophistry. Only God can judge in such a manner, and does so justly. Should we thus dispense with earthly justice? Your argument is rapidly becoming one against declaring anyone guilty of anything, ever.

The magnitude of the Vas' ban is huge and so is its counter reaction.

When speaking of hugeness, I think you need to back up about 6-7 years: Yes, the publication of Fruit 2.1 was huge.

I have not said anything of that kind, nor suggested. There was a question about 100% and I said that is nonsense. I did not say mankind should give up principles like beyond reasonable doubt.

[Judoka]

I have let this issue bounce around in my head and I think perhaps the problem here is that we are facing a paradigm shift in how programming is done on the professional landscape (due to the information explosion on the internet.)

It seems to me that the debate here revolves around the idea of what is acceptable on a continuum we have 100% original to 100% copied with a grey area in between.
the core problem is, I believe both sides agree to this, that neither situation in this case is true. Rybka is not 100% copied or 100% original. I think we can also agree that this grey area, at which point from center this becomes plagiarism is a vague area that is not clearly defined. It is more of "I know it when I see it" definition.
So the question is where does Rybka fall in to this continuum. We tend to as humans fall off the edge to one side or the other once we have made a decision that evidence has swayed us toward a particular side (e.g. John looks guilty there for he IS 100% guilty since there is no "kinda" guilty verdict.) this is why we have a court system.
The definition of cheating is widely varied based on history, culture and area of involvement.
One confound is that of us have been weened in an academic and social environments which we we unconsciously use as our reference point for situations like this. Cheating in the angelo-saxon academic environment is based on academic rules and guidelines that have been figured out over 100's of years. ( It would be an interesting subject for an ethics class. a paper/book on the history of cheating this would be very interesting.) The definition of plagiarism used in the academic environment is probably the most stringent since it extends into the grey area a great deal more than perhaps a legal definition of artistic expression. Things are always under a constant evolution and the internet has accelerated this process.

Since his is a chess subject lets look at the dramatic age drop in the age of Grandmasters in the last 10-15 years since the advent of chess databases and data being proliferated on the internet. Fischers record stood for decades then was broken by Judit polgar then a deluge of younger and younger GMsby broke her record, at first by years then months and now we have bottomed out with weeks and days becoming the.


Why is this so important?
The key article for the proof of cheating presents evidence by presenting Rybka code and fruit code side by side (pages and pages of evidence) but this code was later shown to be modified to help prove guilt.
Why not use 3 columns? fruit code - (modified code rybka WITH the changes shown)- original rybka code?

This point is important because the definition of guilt is somewhere where in the grey area.
Hyatt mentioned in a previous post that the changes were made to show that PST tables can be produced with a minimal number of changes. Fine but the term "minimal" puts the situation into a grey area and is critical for determination of guilt or innocence by the people reviewing the evidence. Is it really that hard to present the facts, unmodified and unaltered for review? providing reams of paper that is technical in approach that they 'could' look at also becomes biased.

What is an acceptable amount of change before something differs from the original enough to be considered it falls into the more "original" category?

There are experts on both sides so it comes down to opinions. What is lacking is a 3rd party and now that they are starting to take notice of the situation we are gaining clarity....The 3rd party posts [on chessvibes] by programming experts that have started to indicate that there are problems with the evidence and process in the ICGA Decision.

[hyatt]

Please supply the "original Rybka code" and I will PERSONALLY be quite happy to compare them "side by side" and we won't need that middle column you mentioned.

This process we used is based on a very simple idea. A block of C source expresses some sort of semantic concept. It has precise inputs and precise outputs for those inputs. A block of assembly language does exactly the same. But it looks FAR different, particularly when written by an experienced programmer or when produced by an optimizing compiler. What one has to do is take EVERY assembly language statement and map it back to its corresponding source code line. When finished, no lines of source can be left out, and no lines of assembly language can be left out. If that is done, one has proven "semantic equivalence" between the two blocks of code.

That is what was done in the evidence. "Show that these assembly language statements do EXACTLY what this Fruit C source code does, leaving nothing out nor adding anything in that is not already present." Ed/Chris like to refer to this as "fruitification". A deliberate attempt to distort and discredit the evidence. Yet this is EXACTLY what a compiler person does when he modifies the optimizer, compiles a test program, and then discovers that the compiled executable does not produce the correct results. He has to figure out why, and then fix it. I have done this MANY times. I have debugged GCC several times when it was producing bad code for "long long" operations (Crafty was one of the first heavy users of long long and gcc broke the binary in many ways with faulty optimizations...) So we were not trying to take the asm and produce just ANY equivalent C source, we were trying to answer the direct question "Does the C in Fruit (or Crafty for the crafty/rybka evidence) match the asm in the rybka binary exactly in terms of semantic operations. The answer was yes. Over and over.

For Crafty 19.x vs rybka 1.4 - rybka 1.6.1, this was a simple and direct process, because the code was copied "as is." For fruit and rybka, it is more complicated, because Rybka uses bitboards and Fruit uses a mailbox board representation. Hence the mentioned "abstraction - filtration - comparison" process where you have to "remove" the board representation from the C and asm (abstraction) so that you can filter those parts that are useful (or not useful) in the comparison, and then compare the blocks of code without the data structures causing what appears to be unequal comparisons...

[lmader]

It seems to me that the debate here revolves around the idea of what is acceptable on a continuum we have 100% original to 100% copied with a grey area in between.
the core problem is, I believe both sides agree to this, that neither situation in this case is true. Rybka is not 100% copied or 100% original. I think we can also agree that this grey area, at which point from center this becomes plagiarism is a vague area that is not clearly defined. It is more of "I know it when I see it" definition.
So the question is where does Rybka fall in to this continuum. We tend to as humans fall off the edge to one side or the other once we have made a decision that evidence has swayed us toward a particular side (e.g. John looks guilty there for he IS 100% guilty since there is no "kinda" guilty verdict.) this is why we have a court system.
The definition of cheating is widely varied based on history, culture and area of involvement.
One confound is that of us have been weened in an academic and social environments which we we unconsciously use as our reference point for situations like this. Cheating in the angelo-saxon academic environment is based on academic rules and guidelines that have been figured out over 100's of years. ( It would be an interesting subject for an ethics class. a paper/book on the history of cheating this would be very interesting.) The definition of plagiarism used in the academic environment is probably the most stringent since it extends into the grey area a great deal more than perhaps a legal definition of artistic expression. Things are always under a constant evolution and the internet has accelerated this process.

This kind of thinking, the idea that there is some vague gray area of interpretation here, is a common misconception. There are two domains to consider, that of Fruit's source code license (GPL), and that of the ICGA's decision. With respect to the source code license, the definition of "cheating" or "plagiarism" or whatever you want to call it here is very clearly defined. The Fruit code is protected under the GPL. If there is _any_ code that was _copied_ (actual cut and paste of source code, not borrowing of conceptual ideas), this is a violation of the license. The investigation's results really aren't very ambiguous - it absolutely looks like code was copied and the GPL was therefore violated. I suppose that one can argue that the ICGA rules regarding the definition of an original work are somewhat looser, but once one accepts that code was copied and that this fact was hidden by the author, then the ICGA is on pretty firm ground to demand that Vas explain (which he chose not to do).

Why is this so important?
The key article for the proof of cheating presents evidence by presenting Rybka code and fruit code side by side (pages and pages of evidence) but this code was later shown to be modified to help prove guilt.
Why not use 3 columns? fruit code - (modified code rybka WITH the changes shown)- original rybka code?

I think there is a misunderstanding here. There is no original Rybka source code to compare to, until Vas releases it. Until then the code comparison is done based on the semantics of each representation (disassembled binary vs Fruit c code). For people that don't understand what it means to compare disassembled code (from Rybka) to actual c source code (from Fruit), this gets confusing, but the bottom line is that his kind of comparison is totally legitimate and can be done with reasonable results. A couple of people have attempted to introduce FUD (fear uncertainty and doubt) into the interpretation of the investigation's code comparison, but none of that FUD is valid. Sadly I guess one has to have some programming experience to be able to evaluate that statement.

This point is important because the definition of guilt is somewhere where in the grey area.
Hyatt mentioned in a previous post that the changes were made to show that PST tables can be produced with a minimal number of changes. Fine but the term "minimal" puts the situation into a grey area and is critical for determination of guilt or innocence by the people reviewing the evidence.

The PST thing has been given way too much importance. Without it even considered, there is an overwhelming amount of damning evidence. It was included because it is meant to add yet-another-example of a smoking gun. By itself the PSTs wouldn't be enough. Taken together with other stuff, it's just that much more evidence. Left out it wouldn't affect the strength of the case.

Is it really that hard to present the facts, unmodified and unaltered for review? providing reams of paper that is technical in approach that they 'could' look at also becomes biased.

Well, it isn't hard to present the code comparisons unaltered, side by side. But this only makes it much more technically demanding to interpret. The code comparisons were filtered to make them more easily understandable. The fact that people still seem confused about just how damning the evidence is makes me think that it does require some technical knowledge to understand.

What is an acceptable amount of change before something differs from the original enough to be considered it falls into the more "original" category?

The Fruit code was protected by a software license that doesn't allow code to be copied without adhering to the terms of that license. The answer is that there cannot be _any_ code copied. Just because some parts of the program were dramatically changed, does not move it along some continuum towards being ok. If there is code left in Rybka that was copied, it's a violation. No gray area.

There are experts on both sides so it comes down to opinions. What is lacking is a 3rd party and now that they are starting to take notice of the situation we are gaining clarity....The 3rd party posts [on chessvibes] by programming experts that have started to indicate that there are problems with the evidence and process in the ICGA Decision.

I realize that it can seem this way to the lay person - that one would think that the people who are supposedly pointing out problems with the evidence are credible experts. The problem is that this just isn't true. Ed Shroeder is the poster child for this and has been the most vocal. But his arguments verge on insanity (or deliberate deceit, hard to know which).

[Judoka]

This kind of thinking, the idea that there is some vague gray area of interpretation here, is a common misconception. There are two domains to consider, that of Fruit's source code license (GPL), and that of the ICGA's decision. With respect to the source code license, the definition of "cheating" or "plagiarism" or whatever you want to call it here is very clearly defined. The Fruit code is protected under the GPL. If there is _any_ code that was _copied_ (actual cut and paste of source code, not borrowing of conceptual ideas), this is a violation of the license. The investigation's results really aren't very ambiguous - it absolutely looks like code was copied and the GPL was therefore violated. I suppose that one can argue that the ICGA rules regarding the definition of an original work are somewhat looser, but once one accepts that code was copied and that this fact was hidden by the author, then the ICGA is on pretty firm ground to demand that Vas explain (which he chose not to do).

With respect to the source code license, the definition of "cheating" or "plagiarism" or whatever you want to call it here is very clearly defined.

I have seen NO clear definition or guidelines presented by ICGA in their examination of Rybka.

The Closest I have seen is: from chessvibes.com

Watkins: "Furthermore, [Riis] is similarly lacking in any knowledge of the relevant aspects of copyright law, particularly the Abstraction-Filtration-Comparison Test that formed a basis for one part of the ICGA Panel analysis. He appears to apply a minimalist copy/paste standard to what “copying” might mean, ignoring any other creative aspects."

emphasis is mine.... what are the other parts? WHY add other parts? the test was created to determine guilt (

http://en.wikipedia.org/wiki/Abstractio ... rison_test
"The AFC test is a three-step process for determining substantial similarity of the non-literal elements of a computer program. The process requires the court to first identify the increasing levels of abstraction of the program. Then, at each level of abstraction, material that is not protectable by copyright is identified and filtered out from further examination. The final step is to compare the defendant's program to the plaintiff's, looking only at the copyright-protected material as identified in the previous two steps, and determine whether the plaintiff's work was copied. In addition, the court will assess the relative significance of any copied material with respect to the entire program."

one part? That begs the question of what the other parts were. Why add additional aspects to the process since the test seems pretty comprehensive, fair and unbiased.


Essentially ICGA made a ruling and they can do what they want since it is a private organization. they need no evidence or even proof just a vote. In this case a simple we dont like you and we THINK you did something wrong would be enough evidence for the ICGA to throw Rybka out of their system. Again ICGA can do what they want.
the problem they tried to create an illusion of a legal process to indicate credibility.

WE assume it used the same or close legal standards we have in the court systems around the country.
WE assume that the process was a fair.
It was at the very least seriously questionable when questions are asked they are blown off.

it absolutelylooks likecode was copied and the GPL was therefore violated.

Looks like , similar, vaguely ,.... understand the base code isnt available but this begs doubt

A couple of people have attempted to introduce FUD (fear uncertainty and doubt) into the interpretation of the investigation's code comparison, but none of that FUD is valid. Sadly I guess one has to have some programming experience to be able to evaluate that statement

.
And how do you consider the evaluation of those people that are expert 3rd party programmers that have expressed doubt?

Well, it isn't hard to present the code comparisons unaltered, side by side. But this only makes it much more technically demanding to interpret. The code comparisons were filtered to make them more easily understandable. The fact that people still seem confused about just how damning the evidence is makes me think that it does require some technical knowledge to understand.

The Fruit code was protected by a software license that doesn't allow code to be copied without adhering to the terms of that license. The answer is that there cannot be _any_ code copied. Just because some parts of the program were dramatically changed, does not move it along some continuum towards being ok. If there is code left in Rybka that was copied, it's a violation. No gray area.

Maybe in an academic world yes but read the article http://en.wikipedia.org/wiki/Abstractio ... rison_test
"Eliminating elements dictated by external factors is an application of the scènes à faire doctrine to computer programs. The doctrine holds that elements necessary for, or standard to, expression in some particular theme can not be protected by copyright.[8] Elements dictated by external factors may include hardware specifications, interoperability and compatibility requirements, design standards, demands of the market being served, and standard programming techniques.[9]
Finally, material that exists in the public domain can not be copyrighted and is also removed from the analysis.[2]"
So code can be copied... again the modification of things are covered as well in the article so the area is grey .

I am a layperson in this issue. I have actually zero interest in Vas as a person or anyone else.
My questions are based on a sense of justice. When I look at what happened I see some serious gaps in what I would consider a fair, open and just process.

The [ICGA] board formed the panel by asking Mark, Harvey and myself to "chair" the panel and to enlist the aid of others as we saw fit. -Hyatt

Shouldnt the main requirement been to have programmers on the panel? Not SOME not partially but entirely composed of experts and programmers of equal level not 1 or 2 that could sway the group. The problem with 1 expert and 10 amateurs would be that the 1 experts opinion could sway the others. Hyatt formed and led the panel even with this the panel returned a 14-0 out of 34 members .
. with 34 members a 14-0 result in guilt so 20 members of the panel abstained?! if they chose not to vote or participate or anything else why were they asked to be on the panel? That amount indicates a clear problem in 1) the evidence was not convincing enough to sway enough members to get at 50% to vote guilty. 2) or that the committee was flawed in its formation with members on the panel who shouldn't have been there in the first place.

The ONLY evidence presented was by the opposition. Vas was asked to offer a defense but well providing code to a group that was formed and led by your critics? Who would jump on that wagon.

The evidence was presented in a biased manner.- Was it stated in the paper/evidence that the code was modified in the side by side comparison? No. The excuse that the code that wasn't real rybka code was modified to make it easier to understand is so leading that it would be thrown out of court by judge judy. Shouldn't the panel have been formed by people who didnt need this crutch? Asking laypersons to be a part of the panel just usually results in them falling in line behind the experts.

The program examined was NOT the same version that competed in the ICGA championships.

- Convicting someone for an activity that didn't take place in their tournament seems to be grasping for straws. If this was the case it would have been much easier to determine guilt by using version that was used that had clear cases of copying from Crafty that participated in events?
The panel did not present a clear guideline and criteria to follow to prove guilt or innocence."The Abstraction-Filtration-Comparison test (AFC) is a method of identifying substantial similarity for the purposes of applying copyright law." This IS the method to use to add steps makes the process questionable.The AFC process was only 1 step? What were the other steps used?
While evidence is very critical so is the manner in which it is examined. There have been people who have been tried and convicted by juries that were 100% sure of their decision based on the evidence presented only to be later proved innocent by additional evidence. The DNA in this case is the original code but after 5 years of accusations drama I dont blame Vas for basically giving the group the silent treatment when they asked him for it.

[K I Hyams]

Shouldnt the main requirement been to have programmers on the panel? Not SOME not partially but entirely composed of experts and programmers of equal level not 1 or 2 that could sway the group. The problem with 1 expert and 10 amateurs would be that the 1 experts opinion could sway the others. Hyatt formed and led the panel even with this the panel returned a 14-0 out of 34 members .

The panel members are listed below. Look at their names. look at their qualifications......................................................
Panel members:
Albert Silver (software designer for Chess Assistant (1999-2002); currently editor of
Chessbase News (2010-present))
Amir Ban (author of Junior: World Champion 2002, 2004, 2006, World microcomputer
Champion 1997, 2001)
Charles Roberson (author of NoonianChess)
Christophe Theron (author of Chess Tiger)
Dariusz Czechowski (author of Darmenios)
Don Dailey (author of Cilkchess, Star Socrates, Rex, Komodo)
Eric Hallsworth (part of Hiarcs Team, Publisher of Selective Search magazine)
Fabien Letousky (author of Fruit)
Frederic Friedel (Chessbase.com)
Gerd Isenberg (author of IsiChess)
Gyula Horvath (author of Pandix, Brainstorm)
Ingo Bauer (Shredder team)
Jan Krabbenbos (Tournament Director of Leiden tournaments)
Kai Himstedt (author of Gridchess and Cluster Toga)
Ken Thompson (creator of Belle Chess Machine, World Computer Chess Champion
1980, Turing Award winner 1983, creator of B and C programming languages,
Unix and Plan 9 developer). More Information about Ken can be found here http://
en.wikipedia.org/wiki/Ken_Thompson
Marcel van Kervinck (author of Rookie)
Maciej Szmit (assistant professor at Technical University of Lodz)
Mark Watkins (MAGMA Computer Algebra Group, School of Mathematics and
Statistics, University of Sydney)
Mark Uniacke (Hiarcs, World Microcomputer Champion 1993)
Mincho Georgiev (Pawny)
Olivier Deville (Tournament Director of ChessWars)
Omid David (author of Falcon)
Peter Skinner (Tournament Director of CCT--the major annual online computer chess
tournament)
Ralf Schäfer (author of Spike)
Richard Vida (author of Critter)
Richard Pijl (author of The Baron)
Stefan Meyer-Kahlen (author of Shredder, multiple world champions from 1996-2007)
Thomas Mayer (author of Quark)
Tord Romstad (author of Stockfish, Glaurung)
Tom Pronk (ProChess, Much)
Vladan Vuckovic (Axon, Achilles)
Wylie Garvin (game Programmer at Ubisoft Montreal)
Yngvi Björnsson (The Turk)
Zach Wegner (author of ZCT and Rondo, an upgraded version of Anthony Cozzie’s
Zappa program, which was world champion in 2005)

Asking laypersons to be a part of the panel just usually results in them falling in line behind the experts.

To which “laypersons” do you refer? Your logic is flawed anyway, even if there were “laypersons" and those hypothetical "laypersons" fell in line “behind the experts”, there would have been fewer abstentions, wouldn’t there?

- Convicting someone for an activity that didn't take place in their tournament seems to be grasping for straws. If this was the case it would have been much easier to determine guilt by using version that was used that had clear cases of copying from Crafty that participated in events?
The panel did not present a clear guideline and criteria to follow to prove guilt or innocence."The Abstraction-Filtration-Comparison test (AFC) is a method of identifying substantial similarity for the purposes of applying copyright law." This IS the method to use to add steps makes the process questionable.The AFC process was only 1 step? What were the other steps used?
While evidence is very critical so is the manner in which it is examined. There have been people who have been tried and convicted by juries that were 100% sure of their decision based on the evidence presented only to be later proved innocent by additional evidence. The DNA in this case is the original code but after 5 years of accusations drama I dont blame Vas for basically giving the group the silent treatment when they asked him for it.

More nonsense. This was not a jury trying a legal case, a "beyond reasonable doubt" verdict was not required. This was simply an internal affair; a panel deciding on whether someone had broken the rules of a private club. The composition of the panel and the way in which they reached a decision was therefore a private matter. There is no obligation for a member of a private club to be present at a disciplinary hearing and there is not even an obligation on those who are conducting the hearing to invite him.

The program examined was NOT the same version that competed in the ICGA championships.

The version of Rybka 1 was effectively the same and the version of Rybka 2 was exactly the same. Rybka 2.3.2a featured in the tournament and Rybka 2.3.2a was the version examined.

In the post below, Cimiotti admits that a version of Rybka that was inspected by the ICGA played in the WCCC:
By Lukas Cimiotti‭ (*****) [‬de‭] ‬Date‭ ‬2007-08-19‭ ‬19:36
Rybka‭ ‬-‭ ‬2.3.2a MP x64‭ ‬gets a reasonable boost from going from‭ ‬4‭ ‬to‭ ‬8‭ ‬cores.‭ ‬Rybka is continously beeing improoved,‭ ‬so the version that will‭ (‬hopefully‭) ‬be playing in Mexico isn't ready yet‭ ‬-‭ ‬it won't be publicly available,‭ ‬but i think,‭ ‬multiprocessor efficiency will not be improoved until then.‭ ‬Btw.‭ ‬WCCC in Amsterdam was run the first day using Rybka‭ ‬2.3.2‭ (‬with TBs and hash size hardcoded‭ ‬-‭ ‬no other modification‭)‬,‭ ‬from the second day it was‭ ‬2.3.2a‭ ‬-‭ ‬the version that was released later to the public.
http://rybkaforum.net/cgi-bin/rybkaforu ... ‭;‬hl=wccc


If you did some basic research before shooting off your mouth, you wouldn’t look quite so stupid.