How hard is?

General discussion about computer chess...
veritas
Posts: 111
Joined: Thu Jun 16, 2011 2:35 pm

Re: How hard is?

Post by veritas » Sat Jul 23, 2011 2:42 am

Richard Vida wrote:
BB+ wrote:Another "original->recycled" analogy might be IPPOLIT's PST now being re-used in Critter [when RV found it worked so well].
Yeah, and my mailbox was full of flames from anonymous ippolit fans :(
I am still pretty convinced that the PST's are mostly just adding noise. Nevertheless, for the next Critter I had developed an 'PST-autotune' method custom tailored for my engine. Alas, judging from early test results it seems to be completely elo-agnostic (no gain, no loss). So far the only gain is that my mailbox will be probably more spam-free in the future :)

Edit: And I am very curious how it will affect the similarity % from Don's "sim03" tester...
Then they are not genuine Ippolit supporters as those Angels 77 (comrades of Ippolit ) have openly thanked you on behalf of Ippolit on the Ippolit Wiki

User avatar
Uly
Posts: 838
Joined: Thu Jun 10, 2010 5:33 am

Re: How hard is?

Post by Uly » Sat Jul 23, 2011 7:52 am

BB+ wrote:It's never easy to have a "fair" anything when the accused won't participate. Indeed, this latter fact was most likely almost solely the reason for the quasi-gratuitous life-ban. My own attitude was more of: there's little point in checking/discussing later versions until he responds to the evidence against R232a.
Participate how? By helping you guys decompile his own engine with ideas of how his code worked? Considering he forgot how to do time management properly, I don't think he could have helped much in his case.
BB+ wrote: There were two reasons for this, the first being the previously mentioned desire to avoid work ( :cool: ), in that if Rajlich couldn't care enough to offer a defense of R1/R232a, then I personally felt that this was enough to impose drastic penalties, even w/o R3/R4 evidence being catalogued. The second reason was that it would be much easier to judge the "originality" of later Rybka versions with Rajlich participating (either with the Board or the Panel).
Oh, so Vas is guilty of not caring?

For all we know he could have a perfectly clean [from Fruit code] chess engine and he wouldn't be able to participate because he can't give evidence of his innocence about ancient engines that are not relevant for future ICGA's tournaments (they are only relevant for the tournaments in where they were used). It almost seems personal (specially if what has been said is true that it is Vas himself who is banned for his actions, and not the chess engines, so it's irrelevant if they're clean or not).

@K I Hyams: I apologize for the misquoting and have fixed it. I failed at copy pasting but blame the forum software in that, it shows Topic review at the bottom of the post a reply page, but their "quote" button opens a new window instead of appending the quoted message to the current post, so it's just faster to copy paste the name of the quoted user manually, which is more error-prone.

veritas
Posts: 111
Joined: Thu Jun 16, 2011 2:35 pm

Re: How hard is?

Post by veritas » Sat Jul 23, 2011 8:30 am

Uly wrote:
BB+ wrote:It's never easy to have a "fair" anything when the accused won't participate. Indeed, this latter fact was most likely almost solely the reason for the quasi-gratuitous life-ban. My own attitude was more of: there's little point in checking/discussing later versions until he responds to the evidence against R232a.
Participate how? By helping you guys decompile his own engine with ideas of how his code worked? Considering he forgot how to do time management properly, I don't think he could have helped much in his case.
BB+ wrote: There were two reasons for this, the first being the previously mentioned desire to avoid work ( :cool: ), in that if Rajlich couldn't care enough to offer a defense of R1/R232a, then I personally felt that this was enough to impose drastic penalties, even w/o R3/R4 evidence being catalogued. The second reason was that it would be much easier to judge the "originality" of later Rybka versions with Rajlich participating (either with the Board or the Panel).
Oh, so Vas is guilty of not caring?

For all we know he could have a perfectly clean [from Fruit code] chess engine and he wouldn't be able to participate because he can't give evidence of his innocence about ancient engines that are not relevant for future ICGA's tournaments (they are only relevant for the tournaments in where they were used). It almost seems personal (specially if what has been said is true that it is Vas himself who is banned for his actions, and not the chess engines, so it's irrelevant if they're clean or not).

@K I Hyams: I apologize for the misquoting and have fixed it. I failed at copy pasting but blame the forum software in that, it shows Topic review at the bottom of the post a reply page, but their "quote" button opens a new window instead of appending the quoted message to the current post, so it's just faster to copy paste the name of the quoted user manually, which is more error-prone.
ulay wrote
Oh, so Vas is guilty of not caring?

well he never cared to "care for the interests of those that kept his bank manager happy
he never cared about Fabien or anyone else he had abused

ulay wrote
For all we know he could have a perfectly clean [from Fruit code]

and perfectly clean clean ( of all traces of other plagiarized codes ?)yes anythings possible

and if common sense was indeed common every one would have it , even you

ulay wrote
I apologize for the misquoting and have fixed it. I failed at copy pasting but blame the forum software in that,

poor workmen often do blame there tools

strange that you don't blame Vas for leaving chunks of useless code hed copied and pasted from other code into his to leave CLEAR EVIDENCE of plagiarism / cloning/ lieing and deceit isn't it

your hero got caught red handed , get over it pal, get a life ,

Jeremy Bernstein
Site Admin
Posts: 1226
Joined: Wed Jun 09, 2010 7:49 am
Real Name: Jeremy Bernstein
Location: Berlin, Germany
Contact:

Re: How hard is?

Post by Jeremy Bernstein » Sat Jul 23, 2011 9:52 am

Uly wrote:Oh, so Vas is guilty of not caring?
That's a glib way of expressing this part of the problem: Vas agreed to assist exactly this sort of investigation when he signed up for the tournament. By "not caring", he's (once again) breaking the rules and adding months of work to the investigation. If Vas doesn't care, doesn't care about the WCCC, doesn't care about his titles and trophies, he's more than welcome to give them up. The investigation demonstrates that he violated the 'originality' rule at least twice. He violated the 'provide source code in the event of an investigation wrt originality' rule at least 4 times.

Keith's point about the man being banned, and not the engine, is 100% correct.

jb

K I Hyams
Posts: 19
Joined: Sun Jun 13, 2010 6:51 pm

Re: How hard is?

Post by K I Hyams » Sat Jul 23, 2011 11:59 am

Uly wrote: @K I Hyams: I apologize for the misquoting and have fixed it. I failed at copy pasting but blame the forum software in that, it shows Topic review at the bottom of the post a reply page, but their "quote" button opens a new window instead of appending the quoted message to the current post, so it's just faster to copy paste the name of the quoted user manually, which is more error-prone.
Thank you.

User avatar
JcMaTe
Posts: 80
Joined: Thu Jun 10, 2010 11:09 am
Real Name: Julio Cesar

Re: How hard is?

Post by JcMaTe » Sat Jul 23, 2011 12:35 pm

Uly wrote:
Jeremy Bernstein wrote:As for laziness, how many volunteer man-hours of someone else's time spent performing unpleasant disassembly and analysis would you consider appropriate?
What about 1/3 of the work that was done? Because, they spent all that work and time checking three different engines, 1.6.1, Beta 1.0 and 2.3.2a, while just checking Rybka 4 would have sufficed.
Jeremy Bernstein wrote:reasonable assumption of continued violation
Backwards assumption. I would have had no problem with them finding Rybka 4 "dirty", and having the reasonable assumption that all previous versions were dirty, because it's obvious. It wouldn't make sense for Vas to have a clean engine all the time, and then suddenly insert dirty code into it in the latest public version.

But it would make sense that at some point he rewrote the engine from scratch and that later versions are "clean", that's why it's much more efficient to just check latest versions for cleanliness.
Jeremy Bernstein wrote:he could have been disqualified simply on the basis of refusing to provide source code as part of such an investigation.
Well, they asked him for an impossible in a definitively non timely manner. As most people know Vas didn't keep any version of his old engine sources, he made a successful change, and he deleted old versions. So he doesn't have any Rybka 3 sources, and probably by now he doesn't have the R4 and R4.1 sources (unless he got more careful by now). How is he going to provide something he doesn't have? Something he deleted 6 years ago (for the Rybka 1.0 Beta).

And I know this is true, because by the time I was beta testing Rybka 4, it was clear that he only had the bare bones and had to adapt the cluster software back to UCI, he didn't even have time control code and had to improvise, which probably cost Rybka 40 elo or more due to terrible time management.

It's probably stupid to not have any version control at all, but don't act like Vas didn't want to show his sources, there just weren't any sources to show.
JcMaTe wrote:1) what make you think when a guy as VAS say that he lost rybka's source code?
I knew what he meant, it's not like he had a drive failure, or some accident in where he lost the sources of the engine, it's that he employed a "burn the ships" programming strategy in where he just wasn't interested in keeping the source of old versions of the engine.

What for? For when he breaks time management so he could easily retrieve the time management code of an old version? Nah...

For people implying that Vas is lying about this, he made this statement back in April 2009.
JcMaTe wrote:2) what do you think when he haven't defend his self yet?
From what I've read about the interview that he gave, he doesn't care about it, as it hasn't affected him negatively (on the contrary, with his engine selling 4 times as much and all).
JcMaTe wrote:3) do you really believe that Rybka 3 and 4 are clean?
I don't know, and actually Rybka 4 maybe is not clean. Both versions could have been found guilty and then there wouldn't be any discussion on how e.g. proving that 1.6.1 is guilty is irrelevant.
JcMaTe wrote:what do you think that he spend hours reading his forum,talkches and maybe open-chess and he do nothing to prove us wrong?
That he doesn't? Whenever he reads Rybka Forum he goes on a posting spree replying to anything he finds relevant, and he hasn't done this in quite a while. My expectations would be that once he has free time and is on the mood he'll read Rybka Forum first, and he hasn't done so.

[^Change "reading" to "spends hours reading" up above if necessary]

The only report of such a thing I've read about was when he spent 10 hours straight reading CCC because he was misinformed about his situation with the ICGA.

I maintain: Innocent until proven guilty, not the other way around. To me stopping the investigation at 2.3.2a is like stopping the investigation at 1.6.1. They spent time checking versions that didn't need to be checked, and didn't check versions that would be relevant for a fair trial of "banned for life", since there's no reason to ban the engine if it is clean.

@hyatt: Check BB+'s message, he got it.

thanks uly for you answers
I knew what he meant, it's not like he had a drive failure, or some accident in where he lost the sources of the engine, it's that he employed a "burn the ships" programming strategy in where he just wasn't interested in keeping the source of old versions of the engine.
I dont believe in this there could be others reason that i believe

1) he will never show us de source code because could be more than ( crafty-fruit)
2) may be he is not the real rybka's autor and thats why he can not defend his self yet
From what I've read about the interview that he gave, he doesn't care about it, as it hasn't affected him negatively (on the contrary, with his engine selling 4 times as much and all).
ok fine he doesn't care, he can return all the money and trophies and plus
That he doesn't? Whenever he reads Rybka Forum he goes on a posting spree replying to anything he finds relevant, and he hasn't done this in quite a while. My expectations would be that once he has free time and is on the mood he'll read Rybka Forum first, and he hasn't done so.
how do you know he doesn't?

BB+
Posts: 1484
Joined: Thu Jun 10, 2010 4:26 am

Re: How hard is?

Post by BB+ » Mon Jul 25, 2011 3:49 am

Richard Vida wrote:I am still pretty convinced that the PST's are mostly just adding noise. Nevertheless, for the next Critter I had developed an 'PST-autotune' method custom tailored for my engine. Alas, judging from early test results it seems to be completely elo-agnostic (no gain, no loss). So far the only gain is that my mailbox will be probably more spam-free in the future :)
The same opinion (for eval in general, to some extent) was stated when I met with some programmers a couple of weeks ago. Namely, that you can munge up eval quite a bit, and still see little Elo variation.

BB+
Posts: 1484
Joined: Thu Jun 10, 2010 4:26 am

Re: How hard is?

Post by BB+ » Mon Jul 25, 2011 4:08 am

Uly wrote:For all we know he could have a perfectly clean [from Fruit code] chess engine and he wouldn't be able to participate because he can't give evidence of his innocence about ancient engines that are not relevant for future ICGA's tournaments (they are only relevant for the tournaments in where they were used). It almost seems personal (specially if what has been said is true that it is Vas himself who is banned for his actions, and not the chess engines, so it's irrelevant if they're clean or not).
The ICGA has always been author-based, not engine-based, and as such, bans are against persons rather than engines. Examples of this include Castillo with El Chinito giving an apology (rather than just fix the problem), and similarly with Reul and LIST. A more recent example is KCC Paduk being denied entry in 2008 due to some incidents that are about a decade old (see the Computer Go section from the Historical Examples of the ICGA wiki). In the Rajlich case, independently of what the Verdict entailed for 2008-10, I expect that the ICGA would minimally have asked to pre-certify his code for a 2011 entry, due to past history.
Uly wrote:Participate how?
One simple way that Rajlich could participate would be to tell the ICGA Panel whether R1/R232a/R3/R4 are good enough approximations to the programs he entered into the ICGA tournaments. Lest this seem facetious, I could note that a foremost Rybka supporter has quite recently claimed that any judge would immediately invalidate the ICGA decision based upon the fact that there was no provenance in this regard [which conveniently ignores the point that Rajlich himself was the basis of this]. This is actually much more a consideration for R3/R4 [which are not "subscription-based"] than with the passel of R1/R2 versions that are available as surrogates for 2006/7. While it seems quite reasonable for a number of reasons to assume that R232a is quite close to the 2007 WCCC entry, it becomes much less clear whether the 2008 entry is that similar to R3, or the 2009/10 entries to R3/R4, especially with cluster capabilities being added. Even something simple, such as whether the 2008 entry used Pradu's bitboard code [as with R4, but not R3] is unanswerable without Rajlich's "participation" in the process. [Note: I'm not saying that using Pradu's bitboard code would be enough for disqualification, but am only using it as the most publicly known example of R3/R4 differences].

Another way Rajlich could assist the ICGA would be to allow an outside person (Thompson or Schaeffer comes to mind) to certify his current source code (as per the previous paragraph). Indeed, I only know about it second/thirdhand, but supposedly when Thompson joined the Panel, upon seeing the evidence almost the first thing he did was suggest (to the Secretariat) that Rajlich simply be asked for his source code.

Post Reply