Why don't strong engines enter WCCC?

[hgm]

Improving Fruit by 400+ elo that time sounds as weird as is like improving Stockfish to 3700 and at the same time changing its move representation system.

I don't know why I have to tell this:
ITS NEXT TO IMPOSSIBLE!

Yet Stockfish is a Glaurung derivative, and Glaurung was also 400 Elo weaker than Stockfish is now. Even though you say it is impossible, many top engines have gained hundreds of Elo in the course of their development. Why would Fruit be different?

I guess you did not study the source code of Fruit very well, because if you had, you would have realized that Fruit is extremely well suited for this. In a sense Fruit is a demo engine, which does everything engines do, but does it in the most simplistic and elementary way. This makes that nearly everywhere there is great room for improvement. The move generator is not particularly fast (plain mailbox in stead of bitboard), the piece-square tables are constructed from simple rank and file contributions, the mobility weights of the pieces have simple 1:2:4:4 ratio rather than something exquisitly tuned like 105:183:367:410, etc.

P.S. Rybka might have taken some ideas from Fruit, but it had many original ideas, and at least was NOT a Fruit clone.

No one accused it of being a Fruit clone. It was accused of being a Fruit derivative, not only taking ideas, but copying actual code and algorithms. So your remark here is about as relevant as saying "But Rybka was not an alien space ship! How could it be banned?".

That's what everyone does, and I'm sure that's what Vas did too!

Sure? How? What source of information do you have that no one else has? Did Vas give you the Rybka source code? Unfortunately there are people that do a lot more.

[Sedat Canbaz]

Now what has web-view stats to do with anything? For all we know people just get there to have a good laugh. I am sure there are many top comedians that have websites that get more hits than yours.

OK, so perhaps you are a good book builder, I cannot judge that. But that thousands of people use your book doesn't make you an expert on what engine programmers do. Even if millions of people would use your book it doesn't make you knowledgeable about that. It is just totally irellevant.

And remarks like "I don't care too much about what programmers say about programming matters, because I have my own view" are pretty efficient in destroying your credibility.

Any news about Winboard GUI, will be a new release ?
Where the UCI engines will be capable to play without adapters ?? ))

Now what does WinBoard have to do with the originality of engines or WCCC participation? Why mention it at all (other than to advertize to the World that you are a slow learner)?

Hey HGM,

Yes...this is true: I am a comedian, but for you: definitely I am not a comedian )))
You are going crazy, mad from my statements and testings...

Relax...take it easy ok...?!)

I say again: you the World's best chess programmer !

And when I say,
I don't care too much...it does not mean I don't care ... )))
I say only too much

About opening books,
No, this is not true: I am very very bad in book making, ok ?)

And all who participated (216 Book Authors) in SCCT,
http://www.sedatcanbaz.com/chess/?page_id=393
They preferred SCCT, just because I am comedian...ok )?

I say SORRY about the question regarding Winboard next release ))

And keep up the good comments over forums!


I wish you all the best,
Sedat

[vittyvirus]

Improving Fruit by 400+ elo that time sounds as weird as is like improving Stockfish to 3700 and at the same time changing its move representation system.

I don't know why I have to tell this:
ITS NEXT TO IMPOSSIBLE!

Yet Stockfish is a Glaurung derivative, and Glaurung was also 400 Elo weaker than Stockfish is now. Even though you say it is impossible, many top engines have gained hundreds of Elo in the course of their development. Why would Fruit be different?

I guess you did not study the source code of Fruit very well, because if you had, you would have realized that Fruit is extremely well suited for this. In a sense Fruit is a demo engine, which does everything engines do, but does it in the most simplistic and elementary way. This makes that nearly everywhere there is great room for improvement. The move generator is not particularly fast (plain mailbox in stead of bitboard), the piece-square tables are constructed from simple rank and file contributions, the mobility weights of the pieces have simple 1:2:4:4 ratio rather than something exquisitly tuned like 105:183:367:410, etc.

P.S. Rybka might have taken some ideas from Fruit, but it had many original ideas, and at least was NOT a Fruit clone.

No one accused it of being a Fruit clone. It was accused of being a Fruit derivative, not only taking ideas, but copying actual code and algorithms. So your remark here is about as relevant as saying "But Rybka was not an alien space ship! How could it be banned?".

That's what everyone does, and I'm sure that's what Vas did too!

Sure? How? What source of information do you have that no one else has? Did Vas give you the Rybka source code? Unfortunately there are people that do a lot more.

First, please have a short break and chill! Don't be that aggressive.
Second, yes, Fruit can be improved a lot upon. But that doesn't mean that Vas used Fruit as initial construction base. You take the twice amount of time to improve engine by 400 elo than to write it from scratch. And changing the board representation system adds much much more time to that...

[Uri Blass]

Improving Fruit by 400+ elo that time sounds as weird as is like improving Stockfish to 3700 and at the same time changing its move representation system.

I don't know why I have to tell this:
ITS NEXT TO IMPOSSIBLE!

Yet Stockfish is a Glaurung derivative, and Glaurung was also 400 Elo weaker than Stockfish is now. Even though you say it is impossible, many top engines have gained hundreds of Elo in the course of their development. Why would Fruit be different?

I guess you did not study the source code of Fruit very well, because if you had, you would have realized that Fruit is extremely well suited for this. In a sense Fruit is a demo engine, which does everything engines do, but does it in the most simplistic and elementary way. This makes that nearly everywhere there is great room for improvement. The move generator is not particularly fast (plain mailbox in stead of bitboard), the piece-square tables are constructed from simple rank and file contributions, the mobility weights of the pieces have simple 1:2:4:4 ratio rather than something exquisitly tuned like 105:183:367:410, etc.

P.S. Rybka might have taken some ideas from Fruit, but it had many original ideas, and at least was NOT a Fruit clone.

No one accused it of being a Fruit clone. It was accused of being a Fruit derivative, not only taking ideas, but copying actual code and algorithms. So your remark here is about as relevant as saying "But Rybka was not an alien space ship! How could it be banned?".

That's what everyone does, and I'm sure that's what Vas did too!

Sure? How? What source of information do you have that no one else has? Did Vas give you the Rybka source code? Unfortunately there are people that do a lot more.

My comments:
1)Fruit is not extremely slow and even if you can do it faster by translating it to bitboard then you may earn maybe 50 elo and certainly not something close to 400 elo.

2)I also do not believe that you can earn a lot by changing mobility weights in fruit.

Based on my knowledge
the main advantage of the free source detrivative of fruit (Toga) is search and not evaluation.

3)If we talk about copying algorithms and not about copying code then
it is clear that part of the algorithms that people use are going to be the same.

When it is obvious that there is only one best algorithm to do something then you can expect different intelligent people to find it independently even if they do not copy from other people so there is no point in objection to copying of algorithms when you cannot prove if people copied the algorithm or found it by themselves.

[hgm]

Second, yes, Fruit can be improved a lot upon. But that doesn't mean that Vas used Fruit as initial construction base.

Indeed, it does not mean that. But of course the Rybka verdict was never motivated by "There is a stong engine around that can be improved a lot. You have an engine that is a lot better, so you must have used it as a starting point!". People painstakenly reverse-engineered Rybka, to make a detailed comparison. In absence of source code, this is the only way it can be done. Looking at how it plays, how much Elo it has, staring Vas in the eye to measure how long it takes before he blinks... That is all just totally irrelevant.

I don't really want to go too much into the Rybka case, as it is only tangential to the topic of this thread. I have not examined Rybka (I don't even have it), so I cannot be sure whether it is a clone or not. (And I could not care less.) But I am getting a bit tired of people that claim they are sure, for a completely non-sensical reason, who have not made any effort to investigate the matter properly.

You take the twice amount of time to improve engine by 400 elo than to write it from scratch. And changing the board representation system adds much much more time to that...

This just is not true. And there is no way you could know that, as you are still struggling with your first move generator. I don't want to discourage you, but it will take a very long time before you even have an engine that can beat micro-Max (~2000 Elo), if you are going to write it from scratch. Starting from a 'skeleton engine' like Fruit will give you an enormous boost, timewise. (Not to mention the 700 extra Elos...)

[Sedat Canbaz]

This is my last posting over this topic !)

Furthermore continuing... it's pointless to discuss !)

And I should point out this too,
Still there is no any true about proving that Rybka is a clone/derivative or not...!

But there is one true:
- Rybka is +400 Elo stronger than Fruit !!
- Rybka was unbeatable for almost 6 years !
- We lost a great programmer (Vasik Rajlich)


And once I wish to say,
How many years is past...and if we still discuss Rybka...and this would not be exaggeration, if I say:
- Vasik Rajlich is one of the most talented chess programmers!
- Rybka chess engine will remain in my heart always as KING!!!


Many thanks to all, for your interest and comments...



Best Regards,
Sedat

[Adam Hair]

1)I started by copying some names of variables and name of functions from tscp but at that point
I had no working code(only variables and empty functions and few arrays when I needed to add code for it).

1)I will not say that Movei is created via copying.
Most of the code of movei is certainly not in other engines and
I did not start from the code or from most of the code of some chess program.

2)tscp has not the same status as fruit or stockfish (it is not GPL) and it is clearly legal to start from it and make the program close source if you do enough changes and get the permission of the author.

I did not start with tscp and the author of tscp has no objection to what I did so it certainly does not create a legal problem unlike using fruit's code or stockfish's code.

Thanks for your clarification....

I know that you very honest programer!

Btw, It seems you learned some ideas exception TSCP, also from Fruit, right ?

sim version 3
------ Fruit 2.1 (depth: 2, scale: 1.0) ------
78.87 Loop 13.6 (depth: 2, scale: 1.0)
74.79 Gambit Fruit 1.0 Beta 4bx (depth: 2, scale: 1.0)
73.56 Deep Onno 1-2-70 (depth: 2, scale: 1.0)
68.84 Alaric 707 (depth: 2, scale: 1.0)
68.60 ECE 11.01 (depth: 2, scale: 1.0)
68.06 Toga II 1.3x4 (depth: 2, scale: 1.0)
67.72 Strelka 2.0 B (depth: 2, scale: 1.0)
67.70 Belka 2.0.0 (depth: 2, scale: 1.0)
63.73 GarboChess v2.20 x86 (depth: 2, scale: 1.0)
63.55 Naum 4 (depth: 2, scale: 1.0)
62.21 Rotor 0.6 (depth: 2, scale: 1.0)
61.94 Cyclone 1.0 (depth: 2, scale: 1.0)
61.88 Daydreamer 1.75 JA (depth: 2, scale: 1.0)
61.86 N2 0.4 (depth: 2, scale: 1.0)
61.77 Twisted Logic 20080404x (depth: 2, scale: 1.0)
59.65 Pupsi2 v0.08 (depth: 2, scale: 1.0)
59.47 Glaurung 2-epsilon/5 (depth: 2, scale: 1.0)
58.85 Stockfish 1.4 JA (depth: 2, scale: 1.0)
58.80 Ayito 0.2.994 (depth: 2, scale: 1.0)
58.57 Atlas 2.90 (depth: 2, scale: 1.0)
58.34 Umko 1.2 i686 (depth: 2, scale: 1.0)
58.08 Gull 1.0a (depth: 2, scale: 1.0)
57.26 Komodo32 1.0 JA (Doch32) (depth: 2, scale: 1.0)
56.76 Doch32 1.2 JA (depth: 2, scale: 1.0)
56.58 Murka 2 w32 UCI (depth: 2, scale: 1.0)
56.34 Atlas 3.14b (depth: 2, scale: 1.0)
56.28 Philou Version 3.6 (depth: 2, scale: 1.0)
56.19 Naum 4 (depth: 3, scale: 1.0)
55.84 bright-0.2c (depth: 2, scale: 1.0)
55.73 Movei00_8_438 (depth: 2, scale: 1.0)

I would like to point out that a similarity percentage by itself is relatively meaningless. It is only in the context of the percentages from all other engine pairs that we can determine if the percentage is high or low. I have seen enough to know that 78.87% for Fruit 2.1/Loop 13.6 at depth 2 is abnormally high. But nobody I know of has done any analysis to show that a percentage of 55.73% for an engine pair is abnormally high.

[Sedat Canbaz]

Yet Stockfish is a Glaurung derivative, and Glaurung was also 400 Elo weaker than Stockfish is now. Even though you say it is impossible, many top engines have gained hundreds of Elo in the course of their development. Why would Fruit be different?

Hey HGM,

Sorry that I can't keep my word,
But after reading such comments.... I could not to keep myself )))

You are really very naive, but believe me you are a such person !!

And once more I will try to explain you,
- After Fruit release, Vasik Rajlich managed to improve Rybka with more than +400 Elo

-Stockfish team managed to improve their engine in similar Elo points, but after Ippo release

So I guess, without Ippo release,
Stockfish would be at least 100-200 Elo weaker...

And If you still can not see the difference,
Don't worry...to be naive it's a natural... )) !!

So...in other words,
Mr. Vasik Rajlich's Elo improving record is leading...
I mean, in case of comparing the Elo differences !!

And my credits are going especially for Mr. Marco Costalba,
otherwise Stockfish would not be so strong as it nowadays!!

[hgm]

Indeed, very unfortunate. I rejoiced for nothing.

I don't see your point, if indeed you have one. It does not matter whether the ideas came from Ippolit or Rybka or elsewhere. Fact is that an existing engine which was close to the top at the time when it was made, could later be improved by about 400 Elo.

It doesn't matter the slightest if you have to be a genius programmer to do it without outside help / inspiration, or need Ippolit if you are a cloner. The engine you used as starting point would still be stolen code if it wasn't your own.

It does raise an interesting question, though. So far this discussion assumed that the 400 Elo to bridge the gap between Fruit and Rybka must have been added by genius programming. But of course it could just as easily have been stolen from yet another engine that contained tricks that were not in Fruit, as you claim most of the Glaurung -> Stockfish progress was. "Steal once, always a thief", as the saying goes. And let's not forget Vas did clone Crafty before he made Rybka, I have seen no one contesting that. So that would make him a repeat offender...

[Sedat Canbaz]

I hope the below table will be useful:

Code: Select all

CCRL (40/4) 2009 year (before Ippo release)
Rybka 3 64-bit 4CPU	        3163 Elo
Stockfish 1.5.1 64-bit 4CPU	3026 Elo

CCRL (40/4) 2011 year (after Ippo release)
Stockfish 2.1.1 64-bit 4CPU	3191 Elo
Rybka 3 64-bit 4CPU	        3163 Elo

CCRL (40/4) 2014 year
Stockfish 5 64-bit 4CPU	    3370 Elo
Rybka 3 64-bit 4CPU	        3163 Elo	

Note (after Ippo release 2009),
Exception Stockfish, Houdini, Komodo, Gull too and plus many engines are become stronger !