Fabien's open letter to the community

[Laskos]

Hi Adam,

I cannot answer everything right now, it's quite late here.

I used Rybka 1.0 Beta 32bit (2.39MB), at 100 ms. I used Strelka 1.8, in fact all engines at 100 ms. Got a result similar to yours among those two, around 74% similarity. I see nothing wrong with your results. I used a variety of many unrelated or wekaly related engines, therefore this 74% was very high. I am astonished to see Strelka 2.0 B having 92% similarity with Rybka 1.0 Beta 2, it is _extremely_ high. Try tu run a renamed Rybka 1.0 Beta 2, to see the noise of this engine. You got me curious to test those too.

My matrix is here

Code: Select all

  Key:

  1) Critter0.90 (time: 100 ms  scale: 1)
  2) DRybka3 (time: 100 ms  scale: 1.0)
  3) DRybka4 (time: 100 ms  scale: 1.0)
  4) DShredder10 (time: 100 ms  scale: 1.0)
  5) DShredder12 (time: 100 ms  scale: 1.0)
  6) Fruit 2.1 (time: 100 ms  scale: 1.0)
  7) Glaurung 2 (time: 100 ms  scale: 1.0)
  8) Houdini15aa (time: 100 ms  scale: 1.0)
  9) Houdini15bb (time: 100 ms  scale: 1.0)
 10) IvanHoe_B49jA (time: 100 ms  scale: 1.0)
 11) Komodo13 (time: 100 ms  scale: 1)
 12) Naum42 (time: 100 ms  scale: 1.0)
 13) Robbo009a (time: 100 ms  scale: 1.0)
 14) Ruffian (time: 100 ms  scale: 1.0)
 15) Rybka 1,0 beta (time: 100 ms  scale: 1.0)
 16) Spark (time: 100 ms  scale: 1.0)
 17) Stockfish2.0ja (time: 100 ms  scale: 1.0)
 18) Strelka18 (time: 100 ms  scale: 1.0)
 19) Tiger2007 (time: 100 ms  scale: 1.0)

         1     2     3     4     5     6     7     8     9    10    11    12
13    14    15    16    17    18    19
  1.  ----- 53.40 51.68 44.93 46.21 49.32 47.72 50.57 51.43 55.29 48.64 51.82 56
.40 44.93 49.34 46.86 51.32 48.48 42.87
  2.  53.40 ----- 58.25 45.75 48.43 52.05 48.05 53.57 53.88 57.87 51.53 56.55 59
.65 46.56 52.51 48.36 51.37 52.99 44.80
  3.  51.68 58.25 ----- 44.48 47.35 50.51 46.66 53.16 53.65 55.06 49.90 54.53 56
.07 46.22 52.23 47.86 50.30 52.04 44.89
  4.  44.93 45.75 44.48 ----- 51.93 45.93 44.87 41.60 40.85 43.53 44.40 44.66 44
.60 46.02 44.61 43.80 44.40 44.40 42.18
  5.  46.21 48.43 47.35 51.93 ----- 46.73 43.23 44.46 44.87 46.70 45.84 46.73 47
.78 46.37 45.45 46.18 45.68 46.02 44.68
  6.  49.32 52.05 50.51 45.93 46.73 ----- 53.68 44.82 45.36 48.92 48.64 54.96 49
.77 48.05 56.41 47.73 50.45 56.23 43.02
  7.  47.72 48.05 46.66 44.87 43.23 53.68 ----- 41.21 41.67 45.05 45.08 49.75 45
.92 46.98 50.00 44.74 49.85 50.02 41.71
  8.  50.57 53.57 53.16 41.60 44.46 44.82 41.21 ----- 70.37 58.22 46.15 48.47 58
.64 42.36 44.95 45.99 50.11 44.80 43.86
  9.  51.43 53.88 53.65 40.85 44.87 45.36 41.67 70.37 ----- 58.40 46.38 48.41 59
.21 42.73 45.90 46.35 50.06 44.94 43.74
 10.  55.29 57.87 55.06 43.53 46.70 48.92 45.05 58.22 58.40 ----- 49.67 52.44 66
.90 44.53 49.53 48.35 50.42 48.66 44.08
 11.  48.64 51.53 49.90 44.40 45.84 48.64 45.08 46.15 46.38 49.67 ----- 49.21 49
.84 43.93 48.00 46.77 46.95 47.66 42.18
 12.  51.82 56.55 54.53 44.66 46.73 54.96 49.75 48.47 48.41 52.44 49.21 ----- 54
.15 45.82 60.25 47.71 50.70 59.14 44.28
 13.  56.40 59.65 56.07 44.60 47.78 49.77 45.92 58.64 59.21 66.90 49.84 54.15 --
--- 45.21 50.97 48.34 51.32 50.64 44.74
 14.  44.93 46.56 46.22 46.02 46.37 48.05 46.98 42.36 42.73 44.53 43.93 45.82 45
.21 ----- 45.42 43.48 45.10 45.59 44.08
 15.  49.34 52.51 52.23 44.61 45.45 56.41 50.00 44.95 45.90 49.53 48.00 60.25 50
.97 45.42 ----- 46.36 48.65 73.65 41.95
 16.  46.86 48.36 47.86 43.80 46.18 47.73 44.74 45.99 46.35 48.35 46.77 47.71 48
.34 43.48 46.36 ----- 48.43 46.86 43.35
 17.  51.32 51.37 50.30 44.40 45.68 50.45 49.85 50.11 50.06 50.42 46.95 50.70 51
.32 45.10 48.65 48.43 ----- 48.20 44.04
 18.  48.48 52.99 52.04 44.40 46.02 56.23 50.02 44.80 44.94 48.66 47.66 59.14 50
.64 45.59 73.65 46.86 48.20 ----- 41.94
 19.  42.87 44.80 44.89 42.18 44.68 43.02 41.71 43.86 43.74 44.08 42.18 44.28 44
.74 44.08 41.95 43.35 44.04 41.94 -----

I am using the latest sim03.exe, it takes longer, but the previous one was noisy and unreliable (too few positions, many of them bad).

For hierarchical cluster analysis I chose the Complete Linkage algorythm: update s_pq by min(s_pr, s_qr) if S is a similarity matrix and s_ij is the similarity measure between cluster i and cluster j. Then perform that itteratively. For measure I have to dig deeper, I did that in SPSS, no clear idea.

Ok, hope that helps.

Kai

[Laskos]

I recall a thread in which you said this and I called you delusional. Is that the one?

I don't really remember this thread, it was probably another, and in the future be more careful with wording, I have to repeat to your non-random mind that it is you who is paid by Chessbase. These were not my words, I was just astonished to hear that about a fervent advocate of a Chessbase product.

As far as some morals go. And some education, I would say. That is learned at the age of 5, not even 15.

Kai

[Adam Hair]

Hi Adam,

I cannot answer everything right now, it's quite late here.

I used Rybka 1.0 Beta 32bit (2.39MB), at 100 ms. I used Strelka 1.8, in fact all engines at 100 ms. Got a result similar to yours among those two, around 74% similarity. I see nothing wrong with your results. I used a variety of many unrelated or wekaly related engines, therefore this 74% was very high. I am astonished to see Strelka 2.0 B having 92% similarity with Rybka 1.0 Beta 2, it is _extremely_ high. Try tu run a renamed Rybka 1.0 Beta 2, to see the noise of this engine. You got me curious to test those too.

My matrix is here

Code: Select all

  Key:

  1) Critter0.90 (time: 100 ms  scale: 1)
  2) DRybka3 (time: 100 ms  scale: 1.0)
  3) DRybka4 (time: 100 ms  scale: 1.0)
  4) DShredder10 (time: 100 ms  scale: 1.0)
  5) DShredder12 (time: 100 ms  scale: 1.0)
  6) Fruit 2.1 (time: 100 ms  scale: 1.0)
  7) Glaurung 2 (time: 100 ms  scale: 1.0)
  8) Houdini15aa (time: 100 ms  scale: 1.0)
  9) Houdini15bb (time: 100 ms  scale: 1.0)
 10) IvanHoe_B49jA (time: 100 ms  scale: 1.0)
 11) Komodo13 (time: 100 ms  scale: 1)
 12) Naum42 (time: 100 ms  scale: 1.0)
 13) Robbo009a (time: 100 ms  scale: 1.0)
 14) Ruffian (time: 100 ms  scale: 1.0)
 15) Rybka 1,0 beta (time: 100 ms  scale: 1.0)
 16) Spark (time: 100 ms  scale: 1.0)
 17) Stockfish2.0ja (time: 100 ms  scale: 1.0)
 18) Strelka18 (time: 100 ms  scale: 1.0)
 19) Tiger2007 (time: 100 ms  scale: 1.0)

         1     2     3     4     5     6     7     8     9    10    11    12
13    14    15    16    17    18    19
  1.  ----- 53.40 51.68 44.93 46.21 49.32 47.72 50.57 51.43 55.29 48.64 51.82 56
.40 44.93 49.34 46.86 51.32 48.48 42.87
  2.  53.40 ----- 58.25 45.75 48.43 52.05 48.05 53.57 53.88 57.87 51.53 56.55 59
.65 46.56 52.51 48.36 51.37 52.99 44.80
  3.  51.68 58.25 ----- 44.48 47.35 50.51 46.66 53.16 53.65 55.06 49.90 54.53 56
.07 46.22 52.23 47.86 50.30 52.04 44.89
  4.  44.93 45.75 44.48 ----- 51.93 45.93 44.87 41.60 40.85 43.53 44.40 44.66 44
.60 46.02 44.61 43.80 44.40 44.40 42.18
  5.  46.21 48.43 47.35 51.93 ----- 46.73 43.23 44.46 44.87 46.70 45.84 46.73 47
.78 46.37 45.45 46.18 45.68 46.02 44.68
  6.  49.32 52.05 50.51 45.93 46.73 ----- 53.68 44.82 45.36 48.92 48.64 54.96 49
.77 48.05 56.41 47.73 50.45 56.23 43.02
  7.  47.72 48.05 46.66 44.87 43.23 53.68 ----- 41.21 41.67 45.05 45.08 49.75 45
.92 46.98 50.00 44.74 49.85 50.02 41.71
  8.  50.57 53.57 53.16 41.60 44.46 44.82 41.21 ----- 70.37 58.22 46.15 48.47 58
.64 42.36 44.95 45.99 50.11 44.80 43.86
  9.  51.43 53.88 53.65 40.85 44.87 45.36 41.67 70.37 ----- 58.40 46.38 48.41 59
.21 42.73 45.90 46.35 50.06 44.94 43.74
 10.  55.29 57.87 55.06 43.53 46.70 48.92 45.05 58.22 58.40 ----- 49.67 52.44 66
.90 44.53 49.53 48.35 50.42 48.66 44.08
 11.  48.64 51.53 49.90 44.40 45.84 48.64 45.08 46.15 46.38 49.67 ----- 49.21 49
.84 43.93 48.00 46.77 46.95 47.66 42.18
 12.  51.82 56.55 54.53 44.66 46.73 54.96 49.75 48.47 48.41 52.44 49.21 ----- 54
.15 45.82 60.25 47.71 50.70 59.14 44.28
 13.  56.40 59.65 56.07 44.60 47.78 49.77 45.92 58.64 59.21 66.90 49.84 54.15 --
--- 45.21 50.97 48.34 51.32 50.64 44.74
 14.  44.93 46.56 46.22 46.02 46.37 48.05 46.98 42.36 42.73 44.53 43.93 45.82 45
.21 ----- 45.42 43.48 45.10 45.59 44.08
 15.  49.34 52.51 52.23 44.61 45.45 56.41 50.00 44.95 45.90 49.53 48.00 60.25 50
.97 45.42 ----- 46.36 48.65 73.65 41.95
 16.  46.86 48.36 47.86 43.80 46.18 47.73 44.74 45.99 46.35 48.35 46.77 47.71 48
.34 43.48 46.36 ----- 48.43 46.86 43.35
 17.  51.32 51.37 50.30 44.40 45.68 50.45 49.85 50.11 50.06 50.42 46.95 50.70 51
.32 45.10 48.65 48.43 ----- 48.20 44.04
 18.  48.48 52.99 52.04 44.40 46.02 56.23 50.02 44.80 44.94 48.66 47.66 59.14 50
.64 45.59 73.65 46.86 48.20 ----- 41.94
 19.  42.87 44.80 44.89 42.18 44.68 43.02 41.71 43.86 43.74 44.08 42.18 44.28 44
.74 44.08 41.95 43.35 44.04 41.94 -----

I am using the latest sim03.exe, it takes longer, but the previous one was noisy and unreliable (too few positions, many of them bad).

For hierarchical cluster analysis I chose the Complete Linkage algorythm: update s_pq by min(s_pr, s_qr) if S is a similarity matrix and s_ij is the similarity measure between cluster i and cluster j. Then perform that itteratively. For measure I have to dig deeper, I did that in SPSS, no clear idea.

Ok, hope that helps.

Kai

Thanks for the quick response, Kai.

I've gotten that result with Strelka 2.0 and Rybka 1.0 Beta 2 several
times. That's why I believe Beta 2 may actually be a Conkie joke; it
may actually be Strelka.

I am also using sim03.exe. Also, I am using a diverse group of engines.
I only posted the results that were relevant to my question. It is part of
a preliminary set of tests that I have been running. I hope to present my
data in the near future.

[tomgdrums]

man the silence on the Rybka forum is deafening!!!!!!

I posted there HOURS ago asking the Rybka team to say something! Anything!

As I posted on another forum, I will not go to bed tonight as a hypocrite SO if the Rybka team can not even say ANYTHING about this, then Rybka 3 and 4 are off my computer before I turn in for the night. Which sucks but I won't feel right if I don't.

[Albert Silver]

I recall a thread in which you said this and I called you delusional. Is that the one?

I don't really remember this thread, it was probably another, and in the future be more careful with wording, I have to repeat to your non-random mind that it is you who is paid by Chessbase. These were not my words, I was just astonished to hear that about a fervent advocate of a Chessbase product.

As far as some morals go. And some education, I would say. That is learned at the age of 5, not even 15.

Kai

It has nothing to do with manners. Sean suggested I was being paid by CB to post here, you expressed astonishment and made a comment on my character, and I said you were delusional to believe such a ridiculous story. More specifically, as my memory is pretty decent: one had to be delusional to believe that CB would give any importance to what was discussed in this forum or other, much less pay me or anyone to post here.

In a nutshell: neither Chessbase, nor anyone, pays me to post here, nor do I believe that they care one iota about the discussions in this forum. Why should they?

[Alexander Schmidt]

I don't understand the fuss.

VR used Fruit code. Noone who looked at the facts can disagree. Maybe everything is rewritten, but that would still be a GPL violation.

What I wrote many times is: We cannot say if this was a GPL violation or not as long as Fabien doesn't complain.

Now he complained, now we know Rybka is illegal.

[De Vos W]

After Vas stated he talked to Fabien, and Fabien had no issues with Rybka regarding Fruit code. Yes, this is not news I guess if it is a given that Vas is a pathological Liar .

Could somebody please give a link to where Vas stated this. Peter Skinner asked earlier in the thread, but got no response.

He is what I would like to know Mr. Banks. Where is your INDIGNATION
of Rybka. We had to take your suppression and censorship of other programs you deemed a threat to Rybka for using so called stolen code and ideas.

Now Rybka has a direct accusation from the author of Fruit that Vas used code from Fruit in Rybka. Where is your indignation of Rybka. We have more proof against Rybka, then you ever had against the other programs.

I want to know...

When will all version of Rybka be removed from your CCRL Rating List. Since you have stated that no program that is not original will be rated on CCRL Mr. Hypocrite.

I guess this only applies to programs you deem a threat to Rybka.

Personally, I'm waiting for this to play out further before making any big decisions.

Once Fabien says that he has examined all the facts for himself and states that Rybka is undeniably nothing more than a Fruit ripoff and which versions this applies to (statements that he would be prepared to defend in a legal sense), I will be perfectly happy to personally stop testing those versions and to advocate for their removal from the CCRL rating lists (although I'm only one of a dozen or so testers, so that decision would be a group one).

However, there are always two sides to every story and it's incredibly annoying and frustrating that Vas does not say more on this issue.
Perhaps FSF action would be a great way to end this debate once and for all.

Meanwhile, I do think that the issue should be discussed without resorting to spreading false information or making personal attacks.

I've seen members post that the most recent Loop was a Toga ripoff and that the most recent Naum was a Rybka ripoff, so who knows where all this madness will end?
Trouble is that Rybka seems to the only target.

Graham Banks wrote:
Personally, I'm waiting for this to play out further before making any big decisions.
Once Fabien says that he has examined all the facts for himself and states that Rybka is undeniably nothing more than a Fruit ripoff and which versions this applies to (statements that he would be prepared to defend in a legal sense), I will be perfectly happy to personally stop testing those versions and to advocate for their removal from the CCRL rating lists (although I'm only one of a dozen or so testers, so that decision would be a group one).

I quote "Prima"
Prima wrote:
Which versions this applies to....and.... stop testing those versions??!!! Is this suppose to be a clause to still protect and support Rybka at all cost? This is hilarious in itself.

These guys just won't stop with the endless and unsupported excuses to keep using and testing Rybka. They even make a version-clause - indirectly calling Mr. Fabien Letouzy a liar and untrustworthy (my interpretation here). Why should Fabien Letouzey have to prove 'which Rybka version' is Fruit-based and, I quote here;
(statements that he would be prepared to defend in a legal sence)?

But in a similar situation where Vas made the accusation that Ippoli and RobboLito is a Rybka 3 clone, the likes of this poster did not ask Vas to prove which version of Ippolit/RobboLito were Rybka 3 clone. Nor did they 'encourage' Vas to make accusing statements that he would be prepared to defend in a legal sense.

Also, the excuse that Rybka 3 source code was lost (or stolen?) is unbelievable. My personal opinion is that this is a ruse to give the impression that Rybka 4 is ' yet another original Rybka engine', or totally unrelated to its previous Fruit-based versions. I strongly believe that Fruit 2.1 was/is in Ry*ka 1.0 beta, therefore Fruit's concepts has to be transferred to Rybka 2, Rybka 3, and then to Rybka 4's versions. It may be tweaked but nevertheless, Fruit is present in all Rybkas.

To surmise, based on the logic of these guys or Rybka-fan boys, it's okay for Vas to take an open source GPL code and increase his Rybka many hundred ELOs, then makes it close source/not release source - a direct violation of GPL license. He then goes commercial with the GPL-obtained code to improve subsequent Rybka versions without releasing the source codes but he's not asked to defend his ethics, in relation to the Fruit-Rybka GPL license in court, despite numerous and respectable programmers' conclusion that Rybka is definitely a Fruit-clone. I could go on here.....

It's funny that now that Rybka is on the defensive side here, this poster is willing to "Personally wait for this to play out further before making any big decisions." Interesting concept. Where was this civil concept when Vas made an unfounded clone claim that Ippolit and RobboLito are Rybka 3 clones? Where was this civil gesture from the chessbase Sysops/ certain-forum-moderators before censoring and banning people from both forums and playchess server?

Are other sane people seeing these and the double standard hypocrisy going on here?

[hgm]

The proof is the author of Fruit himself. "Fabien's open letter to the community". And Vas statement that he claimed that Strelka 2.0 is a clone of Rybka 1.0. This linked Fruit code with Rybka code.

Read it again, then. As many times as it needed to register. He says "legally 'there is no issue". Which means no GPL violation, i.e. no copying of code.

Now I understand of course you take Fabien for an idiot, so that you can ignoe what he writes completely, and just want to use the fact that he says anything at all as a good opprtunity to shoot off your mouth aganst those that you dislike.

But I take Fabien kind of seriously. "No copying of code, but a translation of the algorithm".

So 'poof' goes your 'proof'...

[De Vos W]

...

Personally, I'm waiting for this to play out further before making any big decisions.

Once Fabien says that he has examined all the facts for himself and states that Rybka is undeniably nothing more than a Fruit ripoff and which versions this applies to (statements that he would be prepared to defend in a legal sense), I will be perfectly happy to personally stop testing those versions ...

"those versions" only ?

Matthias.

I'm quite sure Graham will do the honorable thing and immediately stop testing Rybka 1.0 and 2.0.

[Houdini]

CCRL has never claimed authority in cloning issues. Surely, CEGT neither.

That doesn't alter the observation that CCRL and CEGT have been instrumental in allowing Rybka to ban a number of engines without actually providing any proof or beginning any legal action.
If the CCRL and CEGT members are not aware of this role, you've been artfully manipulated...

Robert