To campaign or not

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: To campaign or not

Post by bob »

Albert Silver wrote:
rebel777 wrote:Enrique made me aware of this thread and it made me subscribe again to CCC as I have been the victim of a similar accusation back in 1994 with (the commercial) REBEL-6 when the program was accused of having manipulated a then popular rating test. I tell you it was ugly. But at least the accusers got it right, very good detective work, they came with 100% proof except that they judged my intentions wrong. I immediately pleaded guilty. You see, that's what proof does.
I remember that one. Wasn't it Rebel 8? You had been testing some variant of knight maneuvering code, which solved one of the test positions (Nd1-e3-f5 or something) in 1-2 seconds. You had forgotten to leave this out in the final version and along came the onslaught.

I think most accepted your explanation. I know I did. It was a bit "in the face" to be some dastardly plan after all.... :lol:

Albert
Nope. What happened was that most programs back then put on the front of the boxes "solves 21 of the BK positions, or solves 6 of the XYZ positions, etc." Someone noticed that Rebel would solve the position correctly, but that if you reversed the pieces and made it black to move it would not. Or if you change the position of an unrelated pawn, or you mirror the position right-to-left. It turned out to be a built-in positional learning data that had results for those positions.

Some thought it was an attempt to do better on certain test suites. I didn't follow it very closely as I didn't care. I can tell you I removed positional learning in Crafty fairly recently because I got tired of all the testing mistakes it caused. Play two long matches and forget to erase positional learning and you get different results. Try to fix a bug and forget to delete positional learning and the node count changes.
rebel777

Re: To campaign or not

Post by rebel777 »

El Gringo wrote: Hi Ed,
Hi Johan, nice to meet you here :wink:
El Gringo wrote: Now we have a new 'god' in chess programming. Vas really did a great job to make such a strong program !!!
But i still have some doubts :oops:
Some testers had the latest private version of rybka and it was only around 2100 Elo strong, maybe weaker than rebel 6 or 7.
And then (a while after the release off the fruit source code) his new version Rybka 1.0 beta was 650 Elo's stronger. i NEVER saw such a big elo-jump in the 20 years i have been in computer chess. Pure coincidence ???
Vas admitted that he has studied the source code off Fruit.....but what do you understand with study ?
Of course Vas got loads of elo points from Fruit, I bet most of the chess programmers did, including the commercial ones. And that's what the intend of public source code is about. Nothing wrong so far.

The point this whole debate is if Vas broke the GPL agreement [ http://www.gnu.org/licenses/gpl-3.0.html ] which forbids any copying of source code and claim it as your own. And that's the accusation that is made here. Not very successful so far BTW.

To be complete: studying GPL source code and rewrite it in your own way is not against the GPL agreement. And believe me, I would like to hear one active chess programmer who hasn't taken something from Fruit.

Regards,

Ed
rebel777

Re: To campaign or not

Post by rebel777 »

Hi Bob, nice to disagree with you after so many years :wink:
bob wrote: [You need to reread your quote above. "I posted and then got 95% behind me." Zach's intent was to start a discussion where others could help, in a forum known for having lots of technical expertise around (not to mention lots of idiots, vitriol and such of course).

One has to start somewhere. Some consider it unfair when the police come to their door questioning them about a crime they are suspected of committing. They consider it unfair when the police go to their friends to verify statements or alibis. etc. But you do have to investigate, and this is a far more private place than putting it in the old r.g.c.c for example.
I prefer to do my homework first and then attack considering the inevitable damage to the target in question. An email discussion group comes to mind.

Ed
User avatar
Bill Rogers
Posts: 3562
Joined: Thu Mar 09, 2006 3:54 am
Location: San Jose, California

Re: To campaign or not

Post by Bill Rogers »

Norman
I just answered you question in another post. In the thousands of chess programs out there and believe me there are thousands, just how many do you supect use almost the exact same search routine? How many use the same Alpha/beta routine, quiensence search? How about interanal board representation. How many chess program licensed under GPL are trying to prove orginality buy have simply used algorithms that have been made public and have been free for dozens of years.
While it may be true that some search routines may have some unique coding as well as the evaluation routine, the rest of the programs logic is in my opinion copied of free software and thus not protected in any way.
Even if an author come up with his own design it would not be the first time that one discovery was made by more that one person about the same time. This was true with alpha/beta along with dozens of other examples.
Once again I will restate the laws of slander and that is if you accuse someone of doing something and in reality they are not guilty then you can be sued for libel. Your friends little piece of coding does not prove anything at this point in the game. It would take a massive amount of line for line identical code to even begin to imply someone cheated.
Bill
rebel777

Re: To campaign or not

Post by rebel777 »

Albert Silver wrote:
rebel777 wrote:Enrique made me aware of this thread and it made me subscribe again to CCC as I have been the victim of a similar accusation back in 1994 with (the commercial) REBEL-6 when the program was accused of having manipulated a then popular rating test. I tell you it was ugly. But at least the accusers got it right, very good detective work, they came with 100% proof except that they judged my intentions wrong. I immediately pleaded guilty. You see, that's what proof does.
I remember that one. Wasn't it Rebel 8? You had been testing some variant of knight maneuvering code, which solved one of the test positions (Nd1-e3-f5 or something) in 1-2 seconds. You had forgotten to leave this out in the final version and along came the onslaught.

I think most accepted your explanation. I know I did. It was a bit "in the face" to be some dastardly plan after all.... :lol:

Albert
Hi Albert, nice to see the real oldies still post :wink:

It was REBEL-6 for sure. And there was no way around the fact 3 positions were solved too quickly :lol:

Ed
Terry McCracken
Posts: 16465
Joined: Wed Aug 01, 2007 4:16 am
Location: Canada

Re: To campaign or not

Post by Terry McCracken »

rebel777 wrote:
El Gringo wrote: Hi Ed,
Hi Johan, nice to meet you here :wink:
El Gringo wrote: Now we have a new 'god' in chess programming. Vas really did a great job to make such a strong program !!!
But i still have some doubts :oops:
Some testers had the latest private version of rybka and it was only around 2100 Elo strong, maybe weaker than rebel 6 or 7.
And then (a while after the release off the fruit source code) his new version Rybka 1.0 beta was 650 Elo's stronger. i NEVER saw such a big elo-jump in the 20 years i have been in computer chess. Pure coincidence ???
Vas admitted that he has studied the source code off Fruit.....but what do you understand with study ?
Of course Vas got loads of elo points from Fruit, I bet most of the chess programmers did, including the commercial ones. And that's what the intend of public source code is about. Nothing wrong so far.

The point this whole debate is if Vas broke the GPL agreement [ http://www.gnu.org/licenses/gpl-3.0.html ] which forbids any copying of source code and claim it as your own. And that's the accusation that is made here. Not very successful so far BTW.

To be complete: studying GPL source code and rewrite it in your own way is not against the GPL agreement. And believe me, I would like to hear one active chess programmer who hasn't taken something from Fruit.

Regards,

Ed
I guess you didn't read what I wrote or just chose to ignore it.

There is no campaign. There are no accusations.

Just questions, and certain data provided to work with, are they forbidden?

It seems to me the spirit of the CC community isn't there, it's dead.

You also leave Christophe whom you worked closely with and should know him rather well, and Bob, (whom BTW doesn't use fruit code), and Zach to hang and dry in the Sahara Sun!

I think that's very irresponsible and unkind to say the least.
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: To campaign or not

Post by bob »

chrisw wrote:
bob wrote:
rebel777 wrote:
Zach Wegner wrote: Thanks for the support guys.
Well, you don't have mine :wink:
Zach Wegner wrote: It's good to know that there's a bit of common sense left on this forum.

Common sense is that you provide proof. I am sorry but you have not. When you attack do it right, with proof else it will backfire on you, as it does now. The rule innocent until proven applies.

Zach, you are obviously a bright guy and looking at your age you must have a lot of potential. Realize that (chess) programming is something entirely different than politics, the latter being an issue you apparently not master. No real surprise looking at your age!

I suggest you to work on a document that beyond any doubt provides the evidence Rybka contains Fruit code. Else it is better for you to step down and hope that people will forget real fast.

Enrique made me aware of this thread and it made me subscribe again to CCC as I have been the victim of a similar accusation back in 1994 with (the commercial) REBEL-6 when the program was accused of having manipulated a then popular rating test. I tell you it was ugly. But at least the accusers got it right, very good detective work, they came with 100% proof except that they judged my intentions wrong. I immediately pleaded guilty. You see, that's what proof does.

Another example from the past, old-timers here will surely remember. In 1995/96 MCHESS-5 was leading the SSDF-list for the wrong reasons. I discovered it and provided the 100% proof in RGCC (the precursor of CCC so to say) and got 95% of the experts behind me.

Proof, proof, proof, proof, proof, proof, proof, proof, proof...............

I hope this all makes some sense to you and wish you all the best.

Ed
You need to reread your quote above. "I posted and then got 95% behind me." Zach's intent was to start a discussion where others could help, in a forum known for having lots of technical expertise around (not to mention lots of idiots, vitriol and such of course).

One has to start somewhere. Some consider it unfair when the police come to their door questioning them about a crime they are suspected of committing. They consider it unfair when the police go to their friends to verify statements or alibis. etc. But you do have to investigate, and this is a far more private place than putting it in the old r.g.c.c for example.
Oh, police again?!

The police are subject to vigorous selection, training and discipline. And for good reason. They do not employ, for example, criminals who recently committed the same crime that allegedly requires investigation.
Bullshit, bullshit and bullshit. We had a local police officer last year that was found guilty of murder. He and his wife murdered his ex in-laws who were trying to meddle with his kids. We have had two local police officers found guilty of various crimes and also are now in prison. For asking for "sexual favors" and in return they would tear up a ticket they had written. The list goes on and on. A Detective in Texas was found guilty of manufacturing evidence, and put in prison. And thousands of guilty verdicts in the cases he prosecuted were immediately appealed and many were released on bond. We had a state trooper that beat someone to death and claimed self-defense. Later it turned out that the evidence did not match his report, and witnesses came forth to dispute his explanation. It then turns out he had prior criminal issues that should have disqualified him from serving but they somehow were overlooked.

Police are people just like everyone else. There are good ones and bad ones, just like everyone else

They carry out much of their investigative process quietly and at their own offices. They only start breaking down doors as part of their investigation after assembling good evidence and getting a warrant from a judge. They refrain from publicly accusing suspects unless and until they have assembled evidence considered good wnough to win a prosecution.
False. Friend of ours had a child that got into a small difficulty with the law. He was out camping with his dad, when the local deputies came by to serve the warrant and arrest him. He was not home. They checked the house, and in so doing, noted a couple of phone numbers displayed somewhere. They had the phone company look those up and "went a calling" to see if those people might be hiding the child. One of those was us. My wife had called that home to ask about something advertised in the paper for sale. We got a visit and a detailed explanation of why they had come and who they were looking for. Is that "in the office?" Don't think so. Now we knew about "the problem". And they visited others that "then knew". And this happens all the time. Don't know what utopian world you live in, but it isn't the way things actually work, and never has been. It might be how we would like for them to work, but it isn't how they really work.

In short, they go about their business in a professional manner.

Finally, are you right in stating "it was Zach's intent to start a discussion ....". That rather implies he started this off and is responsible for it. Are you quite sure that is how it happened? 20 year old Zach was the prime mover and got this whole game rolling?
For me, yes. Whether he made the first post or Christophe, I could not say. I believe the first thing I responded to was in Zach's post, but even that could be wrong. So let's say "either Zach or Christophe posted some information to start a discussion." Is that better? Are we going to have to continually argue about every _minor_ point when the bigger issue is still sitting there? My issue in joining in was the rather ridiculous claims that "duplicate code is not surprising at all in a program this big." More bullshit.

Sorry for the expletive, but to say "crap" does "crap" an injustice.
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: To campaign or not

Post by bob »

I'm lost. Ed didn't discover this had happened. Others discovered it and offered 100% proof that somehow, Rebel was "trained" to solve a specific set of positions in a way that did not work if you mirror or reverse the board. He explained how it happend, later. I never gave it much thought. I have reported too many bogus results myself by running the same position twice and forgetting about the learning angle. At least it won't happen in the future for me anyway.
Albert Silver
Posts: 3019
Joined: Wed Mar 08, 2006 9:57 pm
Location: Rio de Janeiro, Brazil

Re: To campaign or not

Post by Albert Silver »

bob wrote:
Albert Silver wrote:
rebel777 wrote:Enrique made me aware of this thread and it made me subscribe again to CCC as I have been the victim of a similar accusation back in 1994 with (the commercial) REBEL-6 when the program was accused of having manipulated a then popular rating test. I tell you it was ugly. But at least the accusers got it right, very good detective work, they came with 100% proof except that they judged my intentions wrong. I immediately pleaded guilty. You see, that's what proof does.
I remember that one. Wasn't it Rebel 8? You had been testing some variant of knight maneuvering code, which solved one of the test positions (Nd1-e3-f5 or something) in 1-2 seconds. You had forgotten to leave this out in the final version and along came the onslaught.

I think most accepted your explanation. I know I did. It was a bit "in the face" to be some dastardly plan after all.... :lol:

Albert
Nope. What happened was that most programs back then put on the front of the boxes "solves 21 of the BK positions, or solves 6 of the XYZ positions, etc." Someone noticed that Rebel would solve the position correctly, but that if you reversed the pieces and made it black to move it would not. Or if you change the position of an unrelated pawn, or you mirror the position right-to-left. It turned out to be a built-in positional learning data that had results for those positions.
Yes, now I remember. I only remembered one of the positions, but not the precise details. Only that it had been an unintentional gaffe.
Some thought it was an attempt to do better on certain test suites. I didn't follow it very closely as I didn't care. I can tell you I removed positional learning in Crafty fairly recently because I got tired of all the testing mistakes it caused. Play two long matches and forget to erase positional learning and you get different results.
Yes, when testing using an openings suite, I am always careful to disable book learning to prevent any issues. I never know if this is necessary, but it certainly can't hurt.

I recall when I first saw learning in a Fidelity stand-alone. Maybe '88 or '89. Its book stopped 2 moves early in a main line of the Italian, and it would get clobbered by not knowing the next two essential moves. I played the line against it some 15 times in blitz, fascinated as it tried a new move every game, until it finally knew the correct two moves. This was new, and I recall being impressed.
"Tactics are the bricks and sticks that make up a game, but positional play is the architectural blueprint."
bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: To campaign or not

Post by bob »

Albert Silver wrote:
bob wrote:
rebel777 wrote:Another example from the past, old-timers here will surely remember. In 1995/96 MCHESS-5 was leading the SSDF-list for the wrong reasons. I discovered it and provided the 100% proof in RGCC (the precursor of CCC so to say) and got 95% of the experts behind me.

Proof, proof, proof, proof, proof, proof, proof, proof, proof...............

I hope this all makes some sense to you and wish you all the best.

Ed
You need to reread your quote above. "I posted and then got 95% behind me."
That isn't exactly what he wrote:

"I discovered it and provided the 100% proof in RGCC and got 95% of the experts behind me"
I simply factored in what I know of the discussion, as I was there. Others found this behavior and offered 100% proof that something was wrong. What Ed means by "I found 100% proof" I am not sure, unless he is talking about something else entirely. The users offered the proof for the test suite that was being discussed at the time.

The _users_ made it public. And then Ed offered an explanation. It actually _was_ that simple. And it was made public before an explanation from Ed was available, just as what has happened here.