Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Strelka 1.0 x32:Rybka clone?
1 2 Previous Next  
- - By Uri Blass (*****) Date 2007-05-07 04:24 Edited 2007-05-07 04:28

There are too many similiarities between it and rybka.

1)both cannot underpromote
2)both show mate score as +320.00
3)both do not have chess knowledge about very simple endgames but strelka is even worse than rybka and does not know that king and knight against king is drawn.
4)both have a big exe

I wonder if vasik gave an older beta version of rybka that even did not know that KN vs K is a draw to somebody.

Parent - - By Leto (***) Date 2007-05-07 09:00
I see too many dissimilarities as I have now noted in the thread:
Parent - - By Uri Blass (*****) Date 2007-05-07 09:15
The point is that many dissimiliarities do not prove that it is not based on rybka's code when many similiarities can prove that it is based on rybka's code.

I suspect that a programmer can take the exe file of rybka and change parts of it in order to make it weaker.
Parent - - By Uri Blass (*****) Date 2007-05-07 09:27
I can add that I read that the meaning of the name in russian is a little arrow.

It seems to be a joke about rybka.
Parent - By Graham Banks (*****) Date 2007-05-07 10:03
I must admit that this crossed my mind also.
Parent - - By Kapaun (****) Date 2007-05-07 10:05
That thought crossed my mind, too. We need to state, however, that just some similarities are no proof. So whatever people are doing to ascertain that engines are clones or not, they should do it now.. :)
Parent - - By Carl Bicknell (*****) Date 2007-05-07 10:19
er..I know this might sound like a dumb question but: how did this guy get hold of Rybka's code? You never hear of a Fritz clone!
Parent - - By Vempele (Silver) Date 2007-05-07 10:31
The machine code is obviously all in the executable. How else do you think you'd even be able to use Rybka? The presentation there is of course pretty much unreadable, but you can use a disassembler or decompiler.
Parent - - By Jack-Ireland (**) Date 2007-05-07 11:58
Rybka cannot underpromote??
Regards, Jack
Parent - By Uri Blass (*****) Date 2007-05-07 12:46
The beta version of rybka cannot underpromote(I did not mean to the commercial version)
Parent - - By Carl Bicknell (*****) Date 2007-05-07 12:43
then why aren't other programs cloned? Why aren't a lot of microsoft's programs cloned and why aren't there fritz / hiarcs/ shredder clones all over the place?
Parent - - By Vempele (Silver) Date 2007-05-07 13:21

> then why aren't other programs cloned?

Lack of incentive. It's illegal, remember, so making a profit would be difficult. On the other hand, there are already strong open source engines that are much easier to clone. Besides, there are hundreds of 'clones' where the 'author' merely changed a few strings in a hex editor.

>Why aren't a lot of microsoft's programs cloned

They are pirated instead. The odds that nobody would notice are extremely slim. Microsoft's lawyers would flay the cloner alive. Besides, it's not easy to do.

>and why aren't there fritz / hiarcs/ shredder clones all over the place?

Also, there are ways to hinder the cloning; for example, some compilers can encrypt the executable. Or you can do premature optimization while coding. :)
Parent - - By RFK (Gold) Date 2007-05-08 05:21
Regardless, when it comes to "Strelka" there is no comparison to Rybka in name or game play. I've tried to run the engine in Arena and Shredder-and within five moves it freezes or calls "illegal-move" and shuts down. The only GUI it seems to run smooth in is Scid, and then only if you play black. Enough of that crap. How the hell can it claim to have a elo of 2700-2800? Nuts!
Parent - - By Uri Blass (*****) Date 2007-05-08 05:56
The fact that you do not know how to use Strelka does not mean that it cannot be used.

It cannot be used in fischer time control but it can be used in 40 minutes/20 moves.
I did not try testing it in games because I do not test engines that I suspect to be clones.

Note that Dann Corbit found strange things when he analyzed the exe.
Parent - By RFK (Gold) Date 2007-05-08 14:12
Please, don't patronize me with a curt pedantic didacticism. Instead of berating me with a prefaced insult for not knowing how to use the chess engine-have the graciousness to get to the point and give me the benefit of your knowledge.

Quite, frankly, If I have to figure out how to use the damned thing-its value is already diminished and its return is minimal. In short, it is not worth the hassle. 


Parent - By Silvian (***) Date 2007-05-11 09:14 Edited 2007-05-11 09:37
Parent - - By Vasik Rajlich (Silver) Date 2007-05-07 12:21
I already got some emails about this. Of course, I have no idea what the author has done, but my first impression is a pretty good one. Sure, the engine is totally raw and needs a lot of nuts-and-bolts work, but that's something that any competent programmer can do. The author shows a clear sense of priorities by concentrating on his ideas rather sidetracking himself with all the little things like underpromotions, distances-to-mate, obscure endgame rules, etc.

When you start in computer chess, the first thing you need to do is to figure out if your ideas will work. (Unless you have no new ideas, in which case the result is predictable and any order is fine.)

Re. mate scores, +320.00 is perfectly logical. It's the largest round value which will fit into a signed 16-bit hash entry.

Parent - - By Eduard (***) Date 2007-05-07 13:03
Dear Vasik!

Please read my message here in "Deutschsprachiger Bereich":

My Example 4 (Beispiel 4) is very courious then this position I played against Rybka 1.2f.
(Strelka need play long time the same line Qxf3???)

My Example 3 (Beispiel 3) is interesting too, then it was played in CSS Online Tournament.
(Strelka need play long time the same line).

Parent - - By Werner (***) Date 2007-05-07 16:53
I have at least 2 examples with Zugzwang positions, where you get the same (wrong) output with both engines:

8/2B1Q3/8/6pp/7k/7P/6P1/2q3K1 w - - 0 1

Strelka and Rybka 1.0 beta want to move 18:
1.Kg1-f2 Dc1-f1+ 2.Kf2xf1
  =  (0.00)   Tiefe: 12   00:00:01  198kN

or here
2q5/8/6p1/B6p/7k/7P/R5PK/8 w - - 0 1

1.Ta2-f2 Dc8-b8+ 2.Kh2-h1 Db8-b1+ 3.Kh1-h2 Db1-b8+ 4.Kh2-h1 Db8-b1+ 5.Kh1-h2 Db1-b8+
  =  (0.00)   Tiefe: 14   00:00:04  1197kN

but to be fair - I also found positions where there are differences like here:

8/3p4/p1bk3p/Pp6/1Kp1PpPp/2P2P1P/2P5/5B2 b - - 0 1

Rybka 1.0 beta
1...Lc6-b7 2.Lf1-e2 Kd6-c6 3.Le2-f1 d7-d6 4.Lf1-g2 Kc6-d7 5.Kb4-a3 Kd7-e6 6.Ka3-b2
  µ  (-0.75)   Tiefe: 26   00:00:36  17668kN

Strelka 1.0 beta

8/3p4/p1bk3p/Pp6/1Kp1PpPp/2P2P1P/2P5/5B2 b - - 0 1

Analysis by Strelka:

1...Lc6xe4 2.f3xe4 f4-f3 3.e4-e5+ Kd6-d5 4.Kb4-a3 Kd5xe5 5.Ka3-b4 Ke5-e4 6.Kb4-c5 Ke4-e3 7.Kc5-b6 Ke3-f2 8.Kb6xa6 Kf2xf1 9.Ka6-b7 f3-f2 10.a5-a6 Kf1-g2 11.a6-a7 f2-f1D 12.a7-a8D Kg2xh3 13.Kb7-c7 Df1-f4+ 14.Kc7xd7 Kh3xg4
  -+  (-3.87)   Tiefe: 21   00:00:01  2105kN
So for me it is not clear if this is a clone. Maybe a joke like Vas wrote :-))

Parent - - By Werner (***) Date 2007-05-08 08:07
...and there are different max search dephts:
Strelka 71.00   0:00    0.00   1.Kb1 Kg8 2.Lf2 Kh8 3.Lg1 Kg8 4.Lh2 Kh8 5.Ka1 Kg8 6.Lg1 Kh8 7.Lh2 (2.531) 7
Bester Zug: Ka1-b1 Zeit: 0:00.531 min  K/s: 7.909  Knoten: 2.531

Rybka 1.0 beta  69.01   0:00    0.00   1.Kb1 Kg8 2.Lf2 Kh8 3.Lg1 Kg8 4.Lh2 Kh8 5.Ka1 Kg8 (1.899) 60
Bester Zug: Ka1-b1 Zeit: 0:00.172 min  K/s: 61.440  Knoten: 1.920

Parent - - By Vasik Rajlich (Silver) Date 2007-05-09 10:53
These types of things are complex. New engines, especially strong ones, get a lot of scrutiny. I am sure that somebody will sit down, snoop around a bit in the binary, etc. There isn't really much I can add without a detailed investigation.

Parent - - By Vertu Date 2007-05-09 16:54
Sergey Markov (SmarThink) got source codes of Strelka from Yuri Osipov, compiled them and says it is not a clone.

If you can read Russian, see
Parent - - By Vertu Date 2007-05-09 22:19
Sergei Markov wrote:

I have examined it's source code today. It's like no one of open source engines and also it is not contains any encrypted binaries. There are no reason to think that it is a clone. I have spoken with Osipov for a long time, he is expirienced enough in computer chess programming topics. Finally we exchanged with our sources: now he has SmarThink src and I have Strelka src. I hope we will cooperate in our research. Also you can see photos of me and Osipov:

About Rybka cloning. Of course, I never seen Rybka sources (but Vasik seen my). I think there are no any reason to think that Vasik presented Rybka's source to Osipov. Reverse engeneering of Rybka to make structured C-sources is really a Sisyphus work. Theoretically possible, but practically incredible.
Parent - - By Vasik Rajlich (Silver) Date 2007-05-11 12:27
This looks quite interesting. Does somebody have a better translation than this?

Author seems to be more of an 'idea' guy than a nose-to-the-grindstone developer.

Also, Strelka uses bitboards, so the current 32-bit release is just a weakened version.

Parent - - By RFK (Gold) Date 2007-05-11 18:55

Not the best translation-but, a tad clearer. I used ""

[I really did not expect such reception, and from it I feel today not so comfortably. As to the brief description of algorithm, algorithm really absolutely usual. I am ready to meet any person who understands chess algorithms, to bring initial texts - we them together we shall look, откомпилируем and we shall check up on tests. I am assured, that any expert, having looked at source codes, will tell: well also what in them such?
If to speak more in detail about algorithm of the Arrow in it it is laid an essential emphasis on tactics. The algorithm was debugged on a set of test positions in which there is not trivial tactical decision. To achieve a finding of such decisions, it was necessary to bring in the program courageous enough heuristics of increase in depth of search or trimming of unpromising variants. Besides effective enough methods of optimization of parameters of estimated function were used simple, but. To tell about it is more detailed, a lot of place and is necessary for time, therefore let's leave it on then.

The history of creation of the Arrow can be begun from apart. Approximately in 1986 I for the first time have become interested in game algorithms. Any special ideas was not, simply it would be desirable to try to make playing program. I have chosen for the beginning simple tic-tac-toe (five in a number), and on this game for the first time have tested main principles of a minimax and an alpha-бета. In four years I have solved замахнуться on a chess, and have for the beginning written something absolutely simple. This program played very badly - simulated game more likely. Gradually game of the program was improved, but then I have thrown it.
And then, in 1996 there was a next splash in interest to a computer chess. Modest Фритц, written for PC, in the next world championship unexpectedly for all became the champion, left such monster, as Deep Blue. I have understood, that there has come time when practically any qualified programmer on enough modest technics can achieve significant successes.
But I would not like to come back to boring enough an alpha-бета to algorithm any more. There was an idea to realize something absolutely original, and there was a corresponding algorithm. The idea of algorithm is simple enough - we reduce a chess problem to a problem of search of an extremum of the function set on set of points, and instead of search we try to search for this extremum in the most perspective direction (something of type of a method of descent). The trouble has appeared (as well as for all such algorithms), in constant сваливании in local extrema of which I and could not get out.
I undertook numerous attempts something to correct, but so anything and has not left.
And then I have decided to return to old kind an alpha-бета to algorithm and to finish it to mind. By this time a plenty of the various information on it on the Internet, including the open initial texts has collected very much. Having studied all this, I have found out, was how much improved for last years an alpha-бета. And I have sat down to write, using as a matter of fact that was known to all. The Arrow which is absolutely unexpected for me as a result was born has appeared enough strong, though also very imperfect
(i.e. examination) me completely arranges the second variant. Moreover, I simply burn with the desire, that such experts have addressed to me, and I shall give them everything, that at me is. I hope, that such examination will allow to remove or even to weaken charges in cloning.
As to free выкладывания source codes I also was going to make it. And most likely, in due course and I shall make. However, yet I do not wish to hurry up with it. It is necessary to weigh consequences of such step very carefully.

Of the further plans did not think yet. For the beginning it is necessary to realize that new situation in which I am unexpected for myself it has appeared. I can tell only, that my interest to an alpha-бета to algorithm for today is satisfied, and I any more do not wish to come back to it. Me quite would arrange, if there was an enthusiast (or some enthusiasts) to which it till now is interesting and to which I could transfer the work made by me.
To commerce I concern with greater doubt - I do not think, that on chess it is possible to earn something essential.

I would like to leave this question without comments. The pseudonym has the real prototype, and I would not wish anything to speak about it. It is very difficult to try to predict the future. I think, the future behind those to whom in a head interesting ideas by a lucky chance will come. And it is absolutely unimportant, whether it works on Chessbase or simply sits at home.
And here what to do then - to create a command or to go on bow to large manufacturers - depends on the person. If there is a desire to earn on ideas - stop to program and start to be engaged in the organization if money you interest a little - continue to work.

I both hands for a full transparency. Also I shall be the first who will send initial texts of the program on the championship. I am afraid only, that I shall be not only the first, but also the only thing that it will make.

I already spoke, that of the further plans did not think yet. However I consider, that in the international competitions to participate it is necessary. And business not only in ambitions or prestige. Only so it is possible to develop and move forward.]
Parent - - By Vasik Rajlich (Silver) Date 2007-05-13 10:15
The author reminds me a bit of Fabien. Computer chess for Fabien was a tool, not an end.

Parent - - By Banned for Life (Gold) Date 2007-05-13 19:52
The author implies that he has developed a technique for morphing a tree search into an N-dimensional surface search (he then indicates that he was unable to get past issues with local maxima). If he has really done this, it would seem to be a revolutionary method of performing the search. Of course most revolutions are failures, but this would be an interesting development, nevertheless.

Parent - By Vasik Rajlich (Silver) Date 2007-05-15 10:58
To tell the truth, I didn't really follow that. I get the impression that he has some wild but unclear ideas.

There is definitely some sort of a tradeoff between working incrementally and trying bigger steps. Too-small steps are slow and will eventually get you to a local maximum. Too-big steps will get you nowhere.

Parent - - By sharpnova (***) Date 2008-08-19 06:33
this is not a difficult transformation actually. and the overhead of searching for global maxes far outweighs any possible benifit. (i see none. the area of the surface is equal to the size of the tree)

so this really seems like a BS idea made to confuse readers and try to convince people he's legitimate.
Parent - - By Linus (***) Date 2008-08-19 07:00
Thanks for clearing that up after more than a year. Now I can sleep better :)
Parent - By sharpnova (***) Date 2008-08-19 07:32
hardy har har :p i knew it was old as hell but i noticed no one had pointed out that spreading the tree out into a high dimensional surface improves nothing and just makes management more tedious and involve horrendous resource allocation for traversal and search.
Parent - - By Roland Rösler (****) Date 2007-05-14 02:19
Perhaps you are right. But it don´t invent so much time in Yuri Ossipov. But Fabien was (is!) great! When you have a look to the readme site of Fruit 2.1 you can see that he is very accurate. Nevertheless he wasn´t quite sure and gave his source to the public. He asked the forum for faults but he earned Toga. Such is life! Fabien was sour and showed with Fruit 2.2.1 that he is better (just for the ego); last Fruit was built in an weekend (+60 Elo). After that, he only was disappointed and do other things. But there are the "Patzer" in the foren and praised Toga.
What we will learn about this? Never sources to the public. The public is almost stupid.
PS: Fabien was also dreaming of an own GUI.
Parent - By Vasik Rajlich (Silver) Date 2007-05-15 11:02
It's not clear what happened to Fabien. I guess he found something better to do. I can't help but think that it's a pity, but there are many things in this world besides computer chess.

Parent - - By noctiferus (***) Date 2007-05-13 21:28
another translation (seems more clear)
Parent - - By RFK (Gold) Date 2007-05-13 21:36 Edited 2007-05-13 21:42

You wouldn't happen to know the http for that article?

Parent - By noctiferus (***) Date 2007-05-13 21:59 Edited 2007-05-13 22:06
I simply got it from google translator, clicking on Vas' link.

the original should be in

I'm quite interested in such a problem formulation: long time ago, I wrote some papers about global optimization (29 years ago; good grief, time is running fast...)
Parent - By albitex (***) Date 2007-05-12 20:53
Can You help me? I have download Strelka, but I don't load it in my shredder GUI . I also download the w2buci adaptator for
do winboard on UCI but no,no load it. Can you write me that made the W2bUci and load Strelka?
Thank Alberto
Parent - - By albitex (***) Date 2007-05-12 20:49
werner i have download Strelka. But I don't load in shredder GUI . I have also download w2buci adaptator for conversion
of winboard to UCI but I have not do null. I have see that there are + version of W2bUci adaptator online. That want I donload?
Can you help me?
Thank Alberto
Parent - - By albitex (***) Date 2007-05-13 00:54
Now I have load the engine strelka. Now ok load.I don't can delete the up my message because off time.
But I now play only analisy mode whit Strelka or depth analisy control time mode , but don't game in normal time control. L'engine
move only if I push the space key. The engine no run in pemament mode, but it think only your time.
( I have made whit W2bUci tool). Control time is ko.Because?
Parent - By Graham Banks (*****) Date 2007-05-13 18:00
You can only use x moves in y minutes in any GUI
Parent - - By onursurme (***) Date 2007-05-10 14:44
can anyone test strelka against rybka ? any estimation about its Elo ?
I couldn't do it because it is very problematic to run Strelka without error.
I've tried Arena and Scid. Strelka doesn't obey time rule.
I've managed only 1 game between them, Strelka generally did it well without a move, which is a big blunder.
I repeated that position and Strelka didn't change its mind.
Parent - By Vempele (Silver) Date 2007-05-10 15:32
1: Strelka                          336,5/700
2: Rybka 2.2 32-bit           71,0/100  111111011111==1=101101011=110001=1=11=0=111=11011=1001=0==1====1=11===1===011=111=11111=1111101=1101
3: Rybka1.0 beta 32-bit    62,0/100  010=1===10100=0101=0=1101111==11=1=====11=00=111000===11=1011111=111101=110111=1=01=0==111=101=10100
4: Toga GrapeFruit            56,5/100  ==0001000001====11=0=0==110101==11=1000=11=10111=111===10110==01=101001101====1101100010==11111==1=1
5: Toga II 1.2.1a               49,5/100  010000=1=00110=====11=100000110100===1===1110=11011==01=00=10=010001=1=01011=1=0==000=101==1==110101
6: GambitFruit beta 4bx    45,0/100  0101=0010001===01=0101=01=100111=00=01=101==0=0===000=110=00=1=10001=10001=1=0=0101010=1==0=0101=1=0
7: Naum 2.0                      40,0/100  00010=01000=1000100=01011011111=1101010=01=110=1=0000=0001=0=00=01101100100=00==0000000=0011==01==01
8: Spike 1.2 Turin              39,5/100  0=00=1011=010=0=00100101=11100=10000==0==101110101010001000100010=000100=10=0==0000111=00101==1000==

Core 2 Duo T5600 (1.83 GHz), 60 moves/1 minute, 24 MB hash (Strelka 46 MB because it can't be changed, Toga GrapeFruit 16 MB because Arena is buggy), Sherwin50.pgn test suite. No tablebases.

Its current CEGT rating is 2777 after 182 games, and I think my results indicate about the same.

Strelka won't work unless you set the time control to x moves in y seconds or fixed depth. It simply doesn't support any other tc.
Parent - By Graham Banks (*****) Date 2007-05-10 18:29
Kirill (CCRL) has been doing this. I'll alert him to your post.

Regards, Graham.
Parent - - By Silvian (***) Date 2007-05-12 17:11 Edited 2007-05-12 17:19
Parent - - By -valdimor- (**) Date 2007-05-12 18:54
i just downloaded strelka and loaded it into arena but it doesn't play :s

do you guys know how to make it move?
Parent - - By RFK (Gold) Date 2007-05-12 20:04

Set a depth limitation and/or move/time limit. See if that breathes life into that scalene, unsymmetrical, disproportionatingly weighted-chatoyant, of a chess engine

Parent - - By -valdimor- (**) Date 2007-05-12 21:38
this engine is not that good as everybody said in the engine room,actually it's worthless!!!so i guess they where just renamed rybkas
Parent - - By Uri Blass (*****) Date 2007-05-12 21:43
This engine is good and is simply not for you.
It is not programmed to play in the engine room so it does not support fischer time control.

You can use it with time control of x minutes/y moves.

Parent - - By RFK (Gold) Date 2007-05-12 22:03
For whom is the engine good for? I would like to see Strelka ultimately give Rybka a run for her money. But, judging from the comments of Yuri Osipov it doesn't sound like he is too concerned either which way it goes.
Parent - - By Michael Waesch Date 2007-05-13 03:47
There are still people who think that its only a virus or at least malware, no engine at all. I don´t say that this is true. I only say that I don´t care at all. I never felt the urge to get that engine. I´d rather download Hiarcs 0.1 which would at least function.

Up Topic Rybka Support & Discussion / Rybka Discussion / Strelka 1.0 x32:Rybka clone?
1 2 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill