For non programmers

Not logged inRybka Chess Community Forum

Forum RybkaChess Help Search Register Login

Topic Rybka Support & Discussion / Rybka Discussion / For non programmers (locked)

1 2 3 4 5 6 7 8 9 10

By michiguel (**)

Date 2011-07-30 07:07

>

> Hello Miguel,
>
> Thanks for making a dissenting post while remaining reasonable as usual. :)
>
> I agree that 384 is a huge exaggeration. Probably the best thing to say is that 9 out of 9 tables with 8 weights each were copied, as well as one other value (KnightTrapped is 100 in Fruit, 3200 in Rybka, which is 1 pawn in both programs). There are some number of weights that were not copied (the exact number is hard to say, since after throwing out 0's neither set of weights is a subset of the other). Something like 20. Now, how much information is contained in those numbers? Hard to say. I think it's certainly way more than 1-3-3-5-9--those are _very_ standard numbers, mentioned in countless chess books, etc. The Fruit tables _are_ unique--what other program uses them that isn't acknowledged to have taken them from Fruit? Of course, with the weights, the match isn't exact, so it's a grey area.
>
> A lot of the evaluation is that way--just a fuzzy, but still rather obvious match. It's hard to say whether it crosses the line, and it's mostly going to be a matter of opinion. If the evaluation code was the only piece of evidence, it would certainly be a tough call. The reason why it's so clear to me is the totality of the evidence. It makes me wish I had spent more time writing up reports about the non-evaluation stuff...

The biggest part of the information content here is hidden in the calculation itself and not the numbers (see my reply to Bob); but, it is either common knowledge (e.g. we know the board should have symmetrical values for some pieces) or very easy to pick up by learning about it. In other words, what may be unique about Fruit tables is "how" to initialize them (multiplying ranks, files, weights, etc.), but I do not see this as a problem. That is a concept, not code (not to mention the code is not even in R1 but supposed to be in the compiling stage). You certainly proved with the PSTs that Vas learned from Fruit a lot, and maybe more than he wants to admit, (I certainly applaud you because it was cleverly done), but I personally do not see this alone as a proof of a wrong doing.

Ok, I just wanted to give my perspective about this minute point.

Miguel

By AWRIST

Date 2011-07-30 11:21

Miguel, what frightens me, your sort of naivety that you didnt put any of your briliant obsevations into the panel, sorry if you did, I was not allowed to read. I think you did it, now the whole case will break into pieces. I dont like however how you remain polite towards Zach who started all this as a proxy for Bob Hyatt. This was evil and if now it comes out what you explained then it's not excusable that you waited till now to publish your refutation on CCC. Are you not the least concerned about how Vas might feel over the years of this hate campaign?

By bob

Date 2011-07-30 03:25

Did you look at the _numbers_?? They are not 1, 2, 3, 4, ...

there are numbers like -4167, +379, not exactly numbers that would roll off of one's tongue if he were asked "just say a few numbers out loud."

Of course, the PST data is not the only thing there.

What we are seeing is a pretty poor attempt to refute _any data_. Out of 384 (actually 768 if you look at opening / middlegame which are separate in both) if they can find one number that seems wrong (due to integer truncation) then the other 767 numbers are bad by association. Weak. Very weak.

By michiguel (**)

Date 2011-07-30 06:40

> Did you look at the _numbers_?? They are not 1, 2, 3, 4, ...
>
> there are numbers like -4167, +379, not exactly numbers that would roll off of one's tongue if he were asked "just say a few numbers out loud."

Those are numbers from Rybka!! (actually, what zach derived to find a match) not from Fruit!!

The numbers are very simple. Reordering the tables you will see that some are identical, and I added ".." to show numbers that can be derived by symmetry.


static const int PawnFile[8]   = { -3, -1, +0, +1, .., .., .., .. };
static const int BishopLine[8] = { -3, -1, +0, +1, .., .., .., .. };
static const int QueenLine[8]  = { -3, -1, +0, +1, .., .., .., .. };
static const int KingLine[8]   = { -3, -1, +0, +1, .., .., .., .. };

static const int RookFile[8]   = { -2, -1, +0, +1, .., .., .., .. };
static const int KnightRank[8] = { -2, -1, +0, +1, +2, +3, +2, +1 };
static const int KnightLine[8] = { -4, -2, +0, +1, .., .., .., .. };

static const int KingFile[8]   = { +3, +4, +2, +0, .., .., .., .. };
static const int KingRank[8]   = { +1, +0, -2, -3, -4, -5, -6, -7 };

Six tables can be generated right away with only five numbers on top of 0,1,2,3 (bolded)

// Code follows ============================

int GENERAL_SHIFT = +2;
int GENERAL_CORNER = -1;
int RF_CORNER = +1;
int NL_CORNER0 = -1;
int NL_CORNER1 = -1;

int template[4] = { 0, 1, 2, 3};

for (j = 0; j < 4; j++) { template[j] += GENERAL_SHIFT;}
template[0] += GENERAL_CORNER;

for (j = 0; j < 4; j++) {
  PawnFile  [j] = template[j];
  BishopLine[j] = template[j];
  QueenLine [j] = template[j];
  KingLine  [j] = template[j];
  RookFile  [j] = template[j];
  KnightLine[j] = template[j];
}

RookFile[0]   +=  RF_CORNER;
KnightLine[0] +=  NL_CORNER0;
KnightLine[1] +=  NL_CORNER1;

for (j = 0; j < 4; j++) {
  PawnFile  [7-j] = template[j];
  BishopLine[7-j] = template[j];
  QueenLine [7-j] = template[j];
  KingLine  [7-j] = template[j];
  RookFile  [7-j] = template[j];
  KnightLine[7-j] = template[j];
}

The seventh, I need to break the symmetry (KnightRank)

The ones that do not follow this pattern are the kings. One of them, is linear except the first two squares (KingRank). So I need two numbers on top of 1,2,3 etc.
The one that is "unique" is KingFile (for numbers). Is that unique? kind of very obvious if you fill it up with numbers from 0-4.

So, I can fill the whole thing with probably ~14 unique numbers, all of them like from -2 to +2 etc.

Miguel

> Of course, the PST data is not the only thing there.
>
> What we are seeing is a pretty poor attempt to refute _any data_. Out of 384 (actually 768 if you look at opening / middlegame which are separate in both) if they can find one number that seems wrong (due to integer truncation) then the other 767 numbers are bad by association. Weak. Very weak.

By bob

Date 2011-07-30 16:19

You say "14 unique numbers". But in a _specific_ order. That is not quite the same. What is the probability of picking 14 numbers from the interval (x...y), and then getting them into the same order? And the fruit rank/file multipliers are not monotonically increasing. They go up by one or two or down by one or two. 16 numbers per PST, plus the constants to multiply by. What's the probability? If we assume that the interval is (-2 to 2). That is 5 values. We need 16 in the same order. (1/5)^16 is right small. My calculator says 0.0 to 9 significant decimal digits...

BTW, Fruits numbers are outside that interval of course. I thought I would make it as favorable as possible. And the probability is still close to zero. With real intervals, it is even closer.

By Rebel

Date 2011-07-30 07:59

>Zach's speech is different than Bob. Zach's evidence, which involved a lot of careful work, is compatible with Zachs' theory that Vas started with Fruit, and kept modifying it until obtaining R1. Note that I said it is compatible, if it is a demonstration, it is another issue (let's not go into that yet). In other words, Zach's evidence and his interpretation is a valid explanation. The key question here should be: Is it the only explanation? That is where the effort should be placed. Bob's speech imply a blatant cut and paste.

A fine observation.

>One thing is what I could or I do believe, another is what I could demonstrate. That line is blurred with this PSTs, at best. As a "jury" a I would disregard it completely.

I agree with your reasoning and conclusion.

Like to add the following, every chess program has PST's. And they all correlate. No program will evaluate a knight on A1 higher than on E4/E5/E6/D4/D5/D6 etc. Every PST in a every chess program has higher values for pawns in the center than on their starting place E2 and D2. Etc. etc. Because of the automatic correlation it's almost impossible to proof copy & paste. I can post mine, another programmer could do the same for demonstration purposes.

I think that in order to keep up the PST copy allegation we need to see both Fruit's and Rybka's PST's next to each other for a fair comparison.

By Homayoun_Sohrabi_M.D. (***)

Date 2011-07-30 05:16

I have another suggestion for supporters of Vas. It would not be unreasonable to get together (figuratively), and draft a RESPECTFUL letter to Mr. Levy (or whoever is the president of ICGA) and outline, point by point and in detail, your concerns about the whole thing, both from a technical and methodology view point as well as from a personal/due process view point. You can of course publish a copy of that letter in this forum. That way you have expressed your concerns in a professional manner and have given Mr Levy an opportunity to respond to your concerns.

You can in fact ask Mr. Levy to consider revising his ruling, based on specific flaws (that you claim) that you have found in their technical reports, etc.

Writing such a letter will involve some real work -and will not be as easy as venting your frustrations on Dr Hyatt- but in the long run, it may be more productive and beneficial to your cause. Thanks.

By Razor

Date 2011-07-30 07:10

By sockmonkey (***)

Date 2011-07-30 09:14

+10

By AWRIST

Date 2011-07-30 11:07

Doctor, yes and no. Good idea but Levy is IMO out of the game. What you are implying tha was all what a true President should have considered in advance. He should have asked

"Folks where is at least the standpoint of a single advocatus diaboli? Do you want to cheat on me with yor report Bob Harvey and Mark? What should this mean that only 14 of 34 had given their votes?? On this data I simply cannot yet go into public with a decision. Excuse me I have dinner now."

By Homayoun_Sohrabi_M.D. (***)

Date 2011-07-30 11:42

Hi Awrist,
what do you mean "Levy is out of the game"?   Isn't he the guy that signed off on the ICGA accusations and banned Vas? You don't want to give him (nicely of course) a piece of your mind?   What does "out of the game" mean? Did he resign his position?

Okay, fine, let's forget about Mr. Levy. You and the other Vas supporters can still compose a detailed statement describing all the faults that you found with the ICGA's report.   You don't have to send it to Levy, but you can still send it to Chessbase, and to Polgar, and to Chess.com, and to Chessvibes, and you can put it on youtube, etc...are you starting to get my meaning?   Better yet, Vas himself can compose this statement since he is the guy who has all the facts.

Anyways, my suggestions were meant only to be helpful.   I know that if I were innocent of what has been alleged, then this is what I would do:   I would state my position in a letter to Levy and whoever else that was willing to listen.   I would also try (with or without the help of a lawyer) to try to defend myself.   You have to fight the fight man, that's how life is. For a software developer, I can not believe that there is anything worse than being accused of copying the work of another software developer.   Why Vas is not defending himself sure beats me.

PS:   I'll answer one question since it seems to be quite a simple one.   The fact that only 14 of 34 gave their vote means that 14 found him guilty and 20 abstained from voting.

By George Speight (***)

Date 2011-08-01 10:03 Edited 2011-08-01 10:09

That should have told you something. And I cringe at the thought of who these 14 who voted actually might be.You got some on here who you could show a page of evidence to and tell them it was a map to Sadam Hussein's buried bullion. The more I see, the less I like it.

gts

Topic Rybka Support & Discussion / Rybka Discussion / For non programmers (locked)

1 2 3 4 5 6 7 8 9 10