About strelka1.8 sources

[GS]

I got source of this program and was asked to write a message about this program

So was I. I have only studied it for about an hour so far, but I can already say that Strelka 1.8 is definitely not a clone of any chess engine I know. The influence from Fruit is evident, but the similarities are nowhere near big enough to make Strelka deserve the clone label. In fact, I have seen the source code of other closed-source chess engines (which nobody has ever expressed any suspicions about) with much more striking similarities to Fruit, but even these engines cannot be described as clones.

I cannot comment about similarities to Rybka, which I know almost nothing about.

Based on what I know, Strelka looks like a perfectly legitimate engine, and deserves at least the same level of respect as my own program.

Tord

And you are sure what you have seen is really 1.8?
If you haven't compared the compiled source you seem to have
received to the released exe you should be very careful to
announce such things.

Guenther

[Tony]

I'm not saying you don't have a point Chris, I'm saying that evidence seems to show that:
a) Strelka is not a clone;
b) it contains data that has been reverse enginereed from Rybka.
I take the testimonials of many people for point a), and Strelka author's own words for b).
I'm really not sure why you want to discuss the above over and over again, IMO it would be more interesting to take a) and b) for granted and proceeding from there, especially if there is something new that can be added to the lot that has already been said so far.

Well Sergei has sent them he says. Not here yet mind. I take a look and then its over so far as im concerned.

Ive better things to do like look at new engines etc....did you see Garbochess?

I just want this put to bed like you so we need speak of it no more.

I believe I can do that.

Christopher

Are you sure ? Converting dimensions of arrays (maybe splitting up) and converting scores (fe from 100 to 255) can be quite difficult to recognize.

Remember, the experts said it was original, but the "author" later told it was a bitboard Fruit clone.

Tony

I think that it is impossible to write bitboard fruit clone.
If a program is bitboard then you need to write a lot of original code and I simply not consider it as a clone.

You may claim that it has some parts of fruit and it is illegal to release it without the source but even in that case it is not a clone because I consider clone to be something that is the same or almost the same and I think that translating non bitboard program to bitboard is more than 50% change.

Uri

The thing that worries me the most is that the judge and jury here are *clearly* people with axes to grind. The chances that their evaluation will be impartial is zero.

The clone fever madness has gone way over the edge. I actually think that it is just as criminal to accuse without proof as it is to perform the crime in the first place. Chess programming has become host to a hostile field of witch hunters who assume guilt based upon the flimsiest bits of evidence.

On the one hand, I do not want to see people stealing other people's work and taking the credit for it. On the other hand, I do not want chess programmers to face outrageous attacks based upon the phase of the moon.

Sorry Dan, you lost me here.

We are talking about the guy who took Fruit, reverse engineerd the Rybka table in it and then very proudly announce he did this, aren't we ? ( And accused Vas of doing the same )

But now, he has written a new engine and all is well, because this one isn't a clone ?

He has mislead people before ( you were one of them) Ever wondered how he did it ? Ever wondered if he's doing it again ?

But I am the bad guy here. Why, because I don't fall for the same crap twice ?

Tony

Sorry Tony but there is no proof for your words and we do not talk about a new strelka but about the old strelka1.8.

There is no proof that the guy took fruit and there is no proof that he proudly announce he did this.

Someone made these claims in some forum and you automatically assume that the author of the source did it?

It was the author who claimed this.

There may be only a proof that he took part of rybka and it is clear that he did not mean to make his program stronger but only to make it similiar to rybka beta as much as possible.

Uri

Reread your last line a few times.

Using open source is a form of stealing. Using closed source is even worse.

But then again: If he only did it to make his engine similair to Rybka... I guess there is no real problem then.

Tony

[Dann Corbit]

I got source of this program and was asked to write a message about this program

So was I. I have only studied it for about an hour so far, but I can already say that Strelka 1.8 is definitely not a clone of any chess engine I know. The influence from Fruit is evident, but the similarities are nowhere near big enough to make Strelka deserve the clone label. In fact, I have seen the source code of other closed-source chess engines (which nobody has ever expressed any suspicions about) with much more striking similarities to Fruit, but even these engines cannot be described as clones.

I cannot comment about similarities to Rybka, which I know almost nothing about.

Based on what I know, Strelka looks like a perfectly legitimate engine, and deserves at least the same level of respect as my own program.

Tord

And you are sure what you have seen is really 1.8?
If you haven't compared the compiled source you seem to have
received to the released exe you should be very careful to
announce such things.

Guenther

I played a match against Movei, which Strelka won, so it is very strong at least. It would be a pretty incredible hoax to produce another such very strong engine. Here are the node counts and outputs for the default version and the version that I compiled. The node counts, ce and pvs are identical (which is something of a surprise, since I am using a different compiler).
[/code]
Analysis from c:\epd\orangutan.epd
10/1/2007 3:44:20 PM Level: 60 Seconds
Analyzing engine: Strelka

1) -
Avoid move:
Best move (Strelka): Ng8-f6
Not found in: 01:00
2 00:00 108 0 +0.21 Nb8c6 Bc1a3
3 00:00 286 0 +0.22 Nb8c6 Nb1c3 Nc6xb4
4 00:00 1.243 0 +0.12 Nb8c6 b4b5 Nc6d4 Nb1c3
5 00:00 3.831 0 +0.20 Nb8c6 b4b5 Nc6b4 Nb1c3 Ng8f6
6 00:00 10.564 0 +0.07 Ng8f6 Ng1f3 Nb8a6 b4b5 Na6c5 Nb1c3
7 00:00 26.874 866.903 +0.09 Ng8f6 Ng1f3 Nb8a6 a2a3 c7c5 Nb1c3 c5xb4
8 00:00 56.385 895.000 +0.09 Ng8f6 Ng1f3 d7d6 b4b5 Nb8d7 Nb1c3 Nd7e5 Nf3xe5 d6xe5
9 00:00 127.557 904.659 +0.09 Ng8f6 Ng1f3 Nb8c6 b4b5 Nc6b4 Nb1c3 d7d6 d2d3 Bc8f5
10 00:00 276.105 841.783 +0.08 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 Bc8f5 Bc1f4
11 00:01 595.228 827.855 +0.09 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 Qd8d6 Qd1d3 Bc8g4
12 00:02 1.133.640 843.482 +0.07 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 Qd8d6 Qd1d3 Bc8g4 Bc1g5
13 00:05 2.803.239 842.319 +0.08 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 e7e6 Qd1d3 Bf8d6 Nf3g5 Bc8d7
14 00:12 5.996.770 773.776 +0.09 Ng8f6 Ng1f3 a7a6 e2e3 Nb8c6 b4b5 a6xb5 Bf1xb5 e7e6 Nb1c3 Bf8c5 d2d4 Bc5b4 Bc1b2
15 00:27 16.104.433 809.634 +0.09 Ng8f6 Ng1f3 e7e6 b4b5 d7d5 Nb1c3 Bf8d6 Bc1b2 Nb8d7 Nf3d4 Nd7e5 e2e3 00 f2f4 Ne5g4
10/1/2007 3:45:23 PM, Time for this analysis: 00:01:00, Rated time: 01:00

0 of 1 matching moves
10/1/2007 3:45:23 PM, Total time: 12:01:03 AM
Rated time: 01:00 = 60 Seconds

--------------------------------------------------------------------------------

Analysis from c:\epd\orangutan.epd
10/1/2007 3:44:20 PM Level: Blitz 5/2
Analyzing engine: Strelka_1.8_UCI

1) -
Avoid move:
Best move (Strelka_1.8_UCI): Ng8-f6
Not found in: 01:00
2 00:00 108 0 +0.21 Nb8c6 Bc1a3
3 00:00 286 0 +0.22 Nb8c6 Nb1c3 Nc6xb4
4 00:00 1.243 0 +0.12 Nb8c6 b4b5 Nc6d4 Nb1c3
5 00:00 3.831 0 +0.20 Nb8c6 b4b5 Nc6b4 Nb1c3 Ng8f6
6 00:00 10.564 0 +0.07 Ng8f6 Ng1f3 Nb8a6 b4b5 Na6c5 Nb1c3
7 00:00 26.874 866.903 +0.09 Ng8f6 Ng1f3 Nb8a6 a2a3 c7c5 Nb1c3 c5xb4
8 00:00 56.385 909.435 +0.09 Ng8f6 Ng1f3 d7d6 b4b5 Nb8d7 Nb1c3 Nd7e5 Nf3xe5 d6xe5
9 00:00 127.557 911.121 +0.09 Ng8f6 Ng1f3 Nb8c6 b4b5 Nc6b4 Nb1c3 d7d6 d2d3 Bc8f5
10 00:00 276.105 884.951 +0.08 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 Bc8f5 Bc1f4
11 00:00 595.228 907.359 +0.09 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 Qd8d6 Qd1d3 Bc8g4
12 00:02 1.133.640 896.158 +0.07 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 Qd8d6 Qd1d3 Bc8g4 Bc1g5
13 00:04 2.803.239 906.317 +0.08 Ng8f6 Ng1f3 a7a6 a2a3 Nb8c6 Nb1c3 d7d5 d2d4 e7e6 Qd1d3 Bf8d6 Nf3g5 Bc8d7
14 00:10 5.996.770 884.348 +0.09 Ng8f6 Ng1f3 a7a6 e2e3 Nb8c6 b4b5 a6xb5 Bf1xb5 e7e6 Nb1c3 Bf8c5 d2d4 Bc5b4 Bc1b2
15 00:24 16.104.433 914.556 +0.09 Ng8f6 Ng1f3 e7e6 b4b5 d7d5 Nb1c3 Bf8d6 Bc1b2 Nb8d7 Nf3d4 Nd7e5 e2e3 00 f2f4 Ne5g4
10/1/2007 3:46:30 PM, Time for this analysis: 00:01:00, Rated time: 01:00

0 of 1 matching moves
10/1/2007 3:46:30 PM, Total time: 12:02:10 AM
Rated time: 01:00 = 60 Seconds

Code: Select all

[Uri Blass]

I also found that the analysis is the same as strelka1.8 that I have.

Uri

[Dann Corbit]

I got source of this program and was asked to write a message about this program so I decided to do it.

Who Uri?

Who sent or gave you the source?

Who asked you to write something about it?

And......why are you writing about it?

What for?

Christopher

Sergei Markoff sent me the source.

Note that the main similiarity in analysis is to rybka but I do not have the source of rybka so I compared the code only with the open source programs and I remember claims that strelka is based on fruit so it is logical to compare with fruit.

Note that strelka is clearly smaller than fruit(except some big table of numbers) so I guess it is more easy to understand it than understanding fruit.

Uri

This obnoxiously long url:
http://computerchess.org.uk/ccrl/4040/ ... ibrate=no
Shows that Strelka's eval is much more similar to Hiarcs than to Rybka. Unfortunately, there does not seem to be any data for Fruit and derivatives against Strelka, so that does not seem to be a good indicator for that matchup.

[Dann Corbit]

I got source of this program and was asked to write a message about this program so I decided to do it.

Who Uri?

Who sent or gave you the source?

Who asked you to write something about it?

And......why are you writing about it?

What for?

Christopher

Sergei Markoff sent me the source.

Note that the main similiarity in analysis is to rybka but I do not have the source of rybka so I compared the code only with the open source programs and I remember claims that strelka is based on fruit so it is logical to compare with fruit.

Note that strelka is clearly smaller than fruit(except some big table of numbers) so I guess it is more easy to understand it than understanding fruit.

Uri

This obnoxiously long url:
http://computerchess.org.uk/ccrl/4040/ ... ibrate=no
Shows that Strelka's eval is much more similar to Hiarcs than to Rybka. Unfortunately, there does not seem to be any data for Fruit and derivatives against Strelka, so that does not seem to be a good indicator for that matchup.

It appears that the 64 bit version of Rybka 1.0 has about the same correlation to Strelka 1.8 as Hiarcs does (a bit over 70% ponder hits).

Each of these program pairs has closer ponder hit ratios:
# Pair Ponder hit Moves counted
1 OBender 3.1.0 – Micro-Max 4.8 80.6 883
2 Lime 62 – Micro-Max 4.8 78.3 387
3 Feuerstein 0.4.4.2 – Micro-Max 4.8 78.0 943
4 Clueless 1.4 – Micro-Max 4.8 77.9 783
5 Feuerstein 0.4.51 – Micro-Max 4.8 77.7 897
6 Monarch 1.7 – Micro-Max 4.8 76.4 1379
7 Loop 13.6 32-bit – Fruit 2.2.1 75.8 1186
8 Loop 10.32f – Fruit 2.2.1 75.6 1018
9 Toga II 1.2.1a – Glaurung 2 epsilon/5 64-bit 75.6 336
10 Delphil 1.6c – Arion 1.7 74.7 352
11 Loop M1-T 64-bit 2CPU – Toga II 1.3.1 72.7 1532

[nczempin]

It appears that the 64 bit version of Rybka 1.0 has about the same correlation to Strelka 1.8 as Hiarcs does (a bit over 70% ponder hits).

Each of these program pairs has closer ponder hit ratios:
# Pair Ponder hit Moves counted
1 OBender 3.1.0 – Micro-Max 4.8 80.6 883
2 Lime 62 – Micro-Max 4.8 78.3 387
3 Feuerstein 0.4.4.2 – Micro-Max 4.8 78.0 943
4 Clueless 1.4 – Micro-Max 4.8 77.9 783
5 Feuerstein 0.4.51 – Micro-Max 4.8 77.7 897
6 Monarch 1.7 – Micro-Max 4.8 76.4 1379
7 Loop 13.6 32-bit – Fruit 2.2.1 75.8 1186
8 Loop 10.32f – Fruit 2.2.1 75.6 1018
9 Toga II 1.2.1a – Glaurung 2 epsilon/5 64-bit 75.6 336
10 Delphil 1.6c – Arion 1.7 74.7 352
11 Loop M1-T 64-bit 2CPU – Toga II 1.3.1 72.7 1532

Well, there you have it. They are all uMax 4.8 clones

[Tord Romstad]

And you are sure what you have seen is really 1.8?
If you haven't compared the compiled source you seem to have
received to the released exe you should be very careful to
announce such things.

The source code I have uses Windows-specific functions, and therefore doesn't compile on my Mac. It's possible that doing the necessary changes would be easy, but I haven't tried. Anyway, even if I were able to compile it, I wouldn't be able to make any comparison, because I can't use the Windows binary version of Strelka.

Anyway, I don't see why it is necessary that I do such a comparison, as other people have already done so. Of course it is theoretically possible that the source code I was sent is different from the one sent to Dann, Uri and others, but this seems to silly to be plausible.

Tord

[swami]

And you are sure what you have seen is really 1.8?
If you haven't compared the compiled source you seem to have
received to the released exe you should be very careful to
announce such things.

The source code I have uses Windows-specific functions, and therefore doesn't compile on my Mac. It's possible that doing the necessary changes would be easy, but I haven't tried. Anyway, even if I were able to compile it, I wouldn't be able to make any comparison, because I can't use the Windows binary version of Strelka.

Anyway, I don't see why it is necessary that I do such a comparison, as other people have already done so. Of course it is theoretically possible that the source code I was sent is different from the one sent to Dann, Uri and others, but this seems to silly to be plausible.

Tord

Yes,Tord.

Author wouldn't have sent the source code to many people if the compiled code exe from the source was different from the released exes.

Regards.

[GS]

And you are sure what you have seen is really 1.8?
If you haven't compared the compiled source you seem to have
received to the released exe you should be very careful to
announce such things.

The source code I have uses Windows-specific functions, and therefore doesn't compile on my Mac. It's possible that doing the necessary changes would be easy, but I haven't tried. Anyway, even if I were able to compile it, I wouldn't be able to make any comparison, because I can't use the Windows binary version of Strelka.

Anyway, I don't see why it is necessary that I do such a comparison, as other people have already done so. Of course it is theoretically possible that the source code I was sent is different from the one sent to Dann, Uri and others, but this seems to silly to be plausible.

Tord

Well, _no one_ has it yet compared to the released version!
It was much time to clean it up meanwhile to let it look
less cloney, even if I don't know why someone should do this,
because the 'author' himself claimed he has cloned parts of Fruit
and Rybka... The whole new discussion seems fruitless to me,
except Sergei wants to use the source now for his own commercial
SmarThink, which would raise a lot of questions and issues.

Guenther