Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Support / Minority Report 3, PST + other lookup tables are they legal?
1 2 Previous Next  
Parent - - By Ugh (*****) Date 2011-07-27 20:28
May I interject a zeroth item to the agenda?

0. What should we do with the WitchFinder General and his entire process when he is caught lying over such a serious matter, saying the PST tables in both programs were identical except for a multiplier constant? A very damaging allegation against Vas which is now shown to be completely untrue?
Parent - By Nick (*****) Date 2011-07-27 20:36

> "saying the PST tables in both programs were identical except for a multiplier constant"

Do you have a reference for where the accusers said this?  (URL, document page number?)

I am assuming that reviewers like Ken Thompson did not do their own hexrays/Ida Pro disassembly and just looked at the available evidence.  It would be good to get their opinions if that evidence has changed.
Parent - By Mark (****) Date 2011-07-27 23:14
Somehow, I knew we'd never make it to point 1!  :grin:
Parent - By Fulcrum2000 (****) Date 2011-07-27 19:38
Best post of today! :grin:
Parent - By Mark (****) Date 2011-07-27 20:59

> for non-programmers, this piece of code is basically adding 0 (zero) 64 times into the PST for the king, opening game phase. Very funny. And totally purposeless.
> I do not believe Vas would write code to add zero 64 times in Rybka.

Several of the weights in Fruit are zero:

KnightBackRankOpening = 0;
QueenCentreOpening = 0;

I guess at the time Fruit was programmed, it wasn't known what values might be best, and it turned out that zero was best.
Parent - - By Richard Vida (**) Date 2011-07-27 21:19 Edited 2011-07-27 21:29

> 384 numbers, chosen one at a time out of a potential sample pool of 2^64 possible 64 bit integers.

Sorry for interrupting, but no sane programmer would want to give a static bonus of 21474836 pawns for a knight on any square ... (and that is still only 32bits and with unit of pawn=100). Your 2^64 is pure BS (sorry). PSQ-table bonuses/penalties out of the range of +- 2 Pawns are just ridiculous and will not work.

Edit: reasonable value is something like 2^8. That is with completely random numbers. Considering common sense (i.e. centralisation idea - knights on center squares are better than knights in the corners) the sensible range becomes considerably tighter. Maybe we are down from number of atoms on earth to number atoms in your glass of whiskey :)
Parent - - By bob (Gold) Date 2011-07-27 22:41
256 is not enough.  Rybka already uses numbers up in the thousands.  So let's go 16 bits to be safe, -32768 to +32767.

384 numbers, 2^16 choices for each.  which is 6 x 2^6 x 2^12 = 3 x 2^18,  So the probability of picking one set is 1 in 600,000 or so (1 / (3 x 2^18) to be exact).  The probability of two people picking the same weights is 1 / (9 x 2^36) which is getting way on down near zero.
Parent - - By Richard Vida (**) Date 2011-07-27 23:12
I am afraid you didn't get my point. I can even use 256-bit numbers (1 pawn = 2^240), that does not make the given range more sensible choice than 8-bit.

Besides the range, these numbers are not random.

Give me a sample 9 sets of 384 _random_ numbers + 1 set of _real_ piece square tables. Any amateur chess programmer will tell you in a few second which is the real one.
Parent - - By bob (Gold) Date 2011-07-28 00:02
Want to bet?  I have some old piece square tables generated by Cozzie's simulated annealing automatic tuning program.  They look like nothing you have ever seen.  And they were no better or worse than the default ones I was already using either.  But no two numbers were the same, and there was no nice symmetry around the center or anything else.  I'll run them down and post just one...

I still consider it impossible for two different programmers to come up with the _same_ set of values...
Parent - By Mark (****) Date 2011-07-28 01:26
If you find them, could you post them on your website?  I've been experimenting with different piece square tables and would like to try them for fun.

I have a feeling that it's not that helpful to just copy the tables.  The optimum tables for one program are probably very different than the optimum tables for another program.
Parent - By bob (Gold) Date 2011-07-27 16:29
Because I have read Zach's report which has what Chris claims has _not_ been shown.  Simple, eh?
Parent - - By bob (Gold) Date 2011-07-26 18:36
As always, you miss the point.  No surprise there.  And then you want to hyperbolize in an effort to shift attention.  Feel free to produce random collections of PGN, until you happen to hit on "Fischer's 100 greatest games" and see if you are subject to copyright infringement action then.  This is not nearly as complicated as you want to make it appear to be. 

I think "obfuscation" is the process you want to pull off here...

But it will only work on those that don't have the technical savvy to understand what is going on.

BTW, do you know how many potential piece/square tables there are, when you have 6 pieces x 2 sides x 64 squares x 2^32 (on 32 bit hardware, 2^64 for 64 bit hardware) possible values?  Let's see.  2^64 x 2^6 x 2 X 6 comes out to 2^72 x 3 different tables.

Feel free to start enumerating those.  And _please_ don't come back until you have them all ready.  I'm not sure where you are going to store them, since that number is pretty close to the number of atoms in planet earth, so you might have a small problem.

But then, math never was your strong point, I take it???

However, as I clearly pointed out, none of what you enumerate is copyrightable by itself.  Now if you somehow come up with a scheme to produce just 12 X 64 values, and you invest significant effort into creating just that set of values, you can claim copyright.  But you missed that point, completely, and go off into the bizarre end of things...

Parent - - By rarchinio (*) Date 2011-07-27 05:52
There are about 2^50 atoms in our planet. You're off by a mere factor of 2^24... About the same as saying that a gram of helium contains a single atom.
Parent - By bob (Gold) Date 2011-07-27 16:39
I believe there are 10^50 atoms on planet earth.    not 2^50.  Actually an accurate estimate is 1.33 x 10^50

The point being that if you could store one number per atom, a _real_ trick, you need a _really_ big storage device.  Just a tad bigger than what could be built here.
- By Homayoun_Sohrabi_M.D. (***) Date 2011-07-28 17:25
Are the members of the forum willing to put their money where their mouth is?  I am -and I have no clue as to which side is right!  We are not helpless.  Why don't we get some money together and hire a TOP INDEPENDENT EXPERT to come up with an independent conclusion.  The expert can do his/her own independent testing as needed.  Vas may have to cooperate and provide code as needed under guarantee of absolute confidentiality.  I am willing to contribute up to 2000 dollars.  The best part will be once the investigation is complete, some of us from both sides can get together, announce the findings, and toast a few drinks.   This is a serious proposal, anybody willing to put their money where their mouth is?   Best regards.
- By RivertonKnight (*) Date 2011-07-30 11:43
I hazed through this thread in 5 minutes ... I gather Fruit and Crafty are as strong as Rybka ...because code numbers are the same? :eek: :twisted:
I think stronger engines are here for mankinds benefit and we should move past all this muddling.
Up Topic Rybka Support & Discussion / Rybka Support / Minority Report 3, PST + other lookup tables are they legal?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill