Similarity tests

[Frank Quisinsky]

Hi Sedat,

I think I have to gave up too.
Made no sense do discuss with you to that topic.

And you never answers to a point I gave you.
All isn't interesting for yourself with your view ... to big is your heart for Rybka. One of the biggest problems today ... most are looking on Elo only. You are a perfect example with your view.

It is OK for me ...
Vas will be happy!
I am sure you can test a newer version at first

And ...
Never I say Rybka is to 100% a Fruit clone.
Never I say Fire is to 100% original work.

Readme file:
I used the work by Fabian as start point ... thanks Fabian!

That is the most important view to the complete story.
Each programmer should respect the work by an other programmer!
Gentlemen like agreement!

Best
Frank

[bob]

Note:the great Don Dailey's view: +60% sim results can be considered as clones

I doubt that Don ever expressed that opinion.

I don't believe he did either. He and I both agreed that it is a cause for concern, but absolutely not an unequivocal proof. That can only come from actual code comparison.

[Sedat Canbaz]

Note:the great Don Dailey's view: +60% sim results can be considered as clones

I doubt that Don ever expressed that opinion.

I don't believe he did either. He and I both agreed that it is a cause for concern, but absolutely not an unequivocal proof. That can only come from actual code comparison.

Oh...yes, now I remembered,
The great Don Dailey claimed that Ippolit is a clone,
And I fully agree with him...

To be honest,
With +60% I can be mistaken, but as far as I remember it was similar numbers, a long time past...

Actually we should not concentrate especially on 60%, because Twinfish proved us that any clone engine can be released below than 60%

Plus, depending on hardware speeds,
The similarity percentage is not same...especially mp engines produce different results

So my conclusion is that (in case of 55% similarity and above),
-We will need to check the both engine Elo points
-We will need help and advice from neutral engine experts views
-We will need to test only those ones which are released before
-We will need to test only those engines which are 100 Elo stronger
etc

Otherwise,
Our tournaments, especially I can say for Top 20:
-Almost all top participants will be based on Rybkanians

Currently I can't see any other right way...


Best,
Sedat

[Sedat Canbaz]

Hello Frank,

A good decision that finely you gave up too, so we are draw ))

Maybe you missed, but if you check my 1st posting, you will notice that,
Firstly I started with questions...if you answer to my questions, then I will try too, if you will not...I will not too ))

About Rybka 1.0 beta,
Be sure... later I will run sim test with that version too

Btw, you are missing again one important point,
I am a neutral independent tester, where you are not !!!


Best,
Sedat

[Rebel]

Note:the great Don Dailey's view: +60% sim results can be considered as clones

I doubt that Don ever expressed that opinion.

About Don's comment, It was a few years ago, but as far as I remember 60% and above can be considered as non-original engines

Don might have said so in the beginning, I vaguely remember he said something like that (60%) too (public and/or private) but Don definitely became much more careful about his tool later and refrained from such statements.

Nevertheless, you as one of the few are doing the right thing and 60% is an excellent percentage if you want your tournaments as original as possible as can be anno 2014. My praise is for you.

OTOH you have to realize that cloners in time will learn how to deal with SIM, as always there are tricks (the emphasis is on the plural) to fool SIM. I have experimented with that in the April one release of Rebel TNG. It's basically Stockfish that is playing yet SIM reports 48% (or so), removing the anti-SIM tricks I got 65%. which reveals the Stockfish involvement.

[Sedat Canbaz]

Note:the great Don Dailey's view: +60% sim results can be considered as clones

I doubt that Don ever expressed that opinion.

About Don's comment, It was a few years ago, but as far as I remember 60% and above can be considered as non-original engines

Don might have said so in the beginning, I vaguely remember he said something like that (60%) too (public and/or private) but Don definitely became much more careful about his tool later and refrained from such statements.

Nevertheless, you as one of the few are doing the right thing and 60% is an excellent percentage if you want your tournaments as original as possible as can be anno 2014. My praise is for you.

OTOH you have to realize that cloners in time will learn how to deal with SIM, as always there are tricks (the emphasis is on the plural) to fool SIM. I have experimented with that in the April one release of Rebel TNG. It's basically Stockfish that is playing yet SIM reports 48% (or so), removing the anti-SIM tricks I got 65%. which reveals the Stockfish involvement.

Thank you dear Ed,

There are not many people like you !!

Some people can see up to 5 meter under water,
But be sure... you belong to ones, who are able to see up to 500 meters!!

Btw,
What makes me really sad,
Unfortunately some testers (including Frank) are become non-neutral testers
I dislike such things....

For examples,
Frank prefers to test the sons (Fire+Critter), but not allowing the mother (Rybka)
Another example,
He prefer to test the old Shredder, but in the same time he is not allowing the rest old ones

Plus...just 1 day ago, he said that,
We should not test those derivatives too, and today he changed his view, I wonder tomorrow what will be his view, his rules are like barometer)))

In shortly, sorry to say that, but
This is a very bad example for rest testers !

But however, Frank is expert...and I wish him good luck !



Best,
Sedat

[Henk]

Still don't know where can I download the e-book "How to clone Rybka 3.0 and pass the similarity test in 24 hours". Maybe we should write one.

This is the way to create many strong "original" chess engines. Users should like that. Or can users smell the difference between an original engine and a clone.

[Sedat Canbaz]

Dear Frank,

See here please...even 1st Rybka is below than 55 % than Fruit !!!!!!

Now I hope you will change your view....

Best,
Sedat


sim version 3
------ Rybka 1.0 Beta (time: 100 ms scale: 1.0) ------
64.68 Strelka 2.0 B (time: 50 ms scale: 1.0)
61.39 Naum 4.6 (time: 100 ms scale: 1.0)
58.76 Murka 3 x64 UCI (time: 100 ms scale: 1.0)
54.87 Rybka 3 (time: 100 ms scale: 1.0)
54.05 RobboLito 0.085g3 w32 (time: 100 ms scale: 1.0)
53.84 Fruit 2.1 (time: 100 ms scale: 1.0)
52.63 Stockfish 1.5 JA 64bit (time: 100 ms scale: 1.0)
52.50 Elektro 1.0 (time: 100 ms scale: 1.0)
51.04 BlackMamba 2.0 x64 (time: 100 ms scale: 1.0)
50.81 Bobcat 3.25 (time: 100 ms scale: 1.0)
50.79 Stockfish 1.7.1 JA (time: 100 ms scale: 1.0)
50.74 Critter 0.90 64-bit SSE4 (time: 100 ms scale: 1.0)
50.66 Equinox 3.20 x64mp (time: 100 ms scale: 1.0)
50.63 Fire 3.0 x64 (time: 100 ms scale: 1.0)
50.16 MinkoChess 1.3 x64 (time: 100 ms scale: 1.0)
50.15 Houdini 4 x64 (time: 100 ms scale: 1.0)
50.13 Octochess revision 5190 (time: 100 ms scale: 1.0)
50.11 Stockfish 2.1 JA 64bit (time: 50 ms scale: 1.0)
49.89 Critter 1.6a 64-bit (time: 100 ms scale: 1.0)
49.56 Stockfish 2.1 JA 64bit (time: 100 ms scale: 1.0)
49.53 Protector 1.7.0 (time: 100 ms scale: 1.0)
49.43 Stockfish 1.7 JA 64bit (time: 100 ms scale: 1.0)
49.38 Senpai 1.0 (time: 100 ms scale: 1.0)
49.28 Daydreamer 1.75 JA (time: 100 ms scale: 1.0)
49.00 Komodo64 2.03 DC (time: 100 ms scale: 1.0)
48.31 Cyrano 0.6b17 (time: 100 ms scale: 1.0)
47.92 cheng4 0.36a (time: 100 ms scale: 1.0)
47.61 Stockfish 140614 64 SSE4.2 (time: 100 ms scale: 1.0)
46.83 Gull 3 x64 (time: 100 ms scale: 1.0)
45.27 Rodent 1.4 (build 2) (time: 100 ms scale: 1.0)
44.87 TwinFish 0.07 (time: 100 ms scale: 1.0)
44.71 Vajolet 2.48 (time: 100 ms scale: 1.0)
42.22 Igorrit 0.086v8_x64 (time: 100 ms scale: 1.0)
40.26 Booot 5.2.0(64) (time: 100 ms scale: 1.0)
28.39 Arasan 17.2 (time: 100 ms scale: 1.0)

[Sedat Canbaz]

Btw,

Even Critter 1.6 is more close to Rybka 3
See please Critter 0.90

Richard Vida is great programmer !!!


sim version 3
------ Critter 1.6a 64-bit (time: 100 ms scale: 1.0) ------
62.59 Houdini 4 x64 (time: 100 ms scale: 1.0)
61.68 Fire 3.0 x64 (time: 100 ms scale: 1.0)
60.56 RobboLito 0.085g3 w32 (time: 100 ms scale: 1.0)
59.98 Elektro 1.0 (time: 100 ms scale: 1.0)
59.72 BlackMamba 2.0 x64 (time: 100 ms scale: 1.0)
57.66 Rybka 3 (time: 100 ms scale: 1.0)
56.43 Equinox 3.20 x64mp (time: 100 ms scale: 1.0)
55.71 Critter 0.90 64-bit SSE4 (time: 100 ms scale: 1.0)

[Rebel]

Btw, What makes me really sad, Unfortunately some testers (including Frank) are become non-neutral testers I dislike such things....

For examples, Frank prefers to test the sons (Fire+Critter), but not allowing the mother (Rybka) Another example, He prefer to test the old Shredder, but in the same time he is not allowing the rest old ones

Well, some prefer strength over originality, even CEGT and CCRL do. Others don't, you, CSVN and SSDF as far as I can tell. Anno 2014 the CC community (as a whole) is best served this way, two sides of the coin.