It walks like a clone, it quacks like a clone ...

[MikeB]

Do you think our esteemed independent rating organization "CCRL" would have the guts to call it a clone, especially in light that authors freely admit it is clone...
Forget about chessbase, forget about Albert, we all have opinions on whether that was legal or not and how that would play out in a civil court - who knows. Until we hit that point, where a court renders it decision , nobody really knows for sure ( although I am sure many think they know).

But one one thing we do know for sure , it is a clone. I honestly think that out of every thing that came out of Fat Fritz 2, this is the one thing that discredits all of the Stockfish developers over the last 15 years or years or so, from Tord to everyone else on the Authors list. I would like hear to your voice on this issue from Tord to Joost and everyone else in between. I would like to ever hear from Albert as well - ALbert do you really think it should not be included with the Stockfish engines and if not why not?

Time for the chess community to speak out and try to get a wrong righted! Bring it on!

CCRL Shame!

[dkappe]

As the author of several of these “clones,” I’m not comfortable with this term. None of them play like each other or like stockfish. Let’s face it, the age of exchangeable networks came with leela and now with nnue has turned most ab engines into nnue “players.” “Clone” was always a pejorative term rather than a diagnostic one and has become even more meaningless with nets. Unless you can say what “clone” actually means, I’d leave the name calling in the past.

[MikeB]

As the author of several of these “clones,” I’m not comfortable with this term. None of them play like each other or like stockfish. Let’s face it, the age of exchangeable networks came with leela and now with nnue has turned most ab engines into nnue “players.” “Clone” was always a pejorative term rather than a diagnostic one and has become even more meaningless with nets. Unless you can say what “clone” actually means, I’d leave the name calling in the past.

Dietrich,
The term "clone" does have some negative implications in the chess community, even so called "legal clones', clones that follow the GPL license to the "T" . Terminology can be negotiated, I would be fine with the less negative connotation term "derivative" , especially in this case where we know the evaluation function embedded in the NNUE is substantially different than current Stockfish NNUE. The crux of my point is that if even we agree on what to call it , say "SF derivative", why should it be kept separate from all the other SF derivatives, some of them, like SugaR , which have changed substantially over the years and are nowhere near Stockfish today in current evaluation ot how Stockfish was in the past.

MikeB

[mvanthoor]

Time for the chess community to speak out and try to get a wrong righted! Bring it on!

The following sentence may not be a very popular opinion, but....

I don't really care anymore. It's almost impossible to know which engine is a clone, or which engine isn't. I mainly disregard any engine that isn't open source (except for testing my own engine; the only thing I care about in that case is the rating), because literally everything can be in there.

To be honest, I've lost interest in the Elo-race at the top. Of course engines are getting stronger all the time; it would be strange if they weren't. They're already so strong that everything over CCRL 3000 is basically godlike. Nowadays, it's easy to have your own chess engine if you want to; just grab one of the 3000+ engines, change a bit here or there, put a network in it to add 300 Elo for free, and re-release it under a new name.

Personally I stick with Stockfish 10 for playing, using the Skill level (11 calibrated against CCRL 1350, which makes the engine, IMHO, too strong for a beginner like my girlfriend, and I don't have enough levels to choose from myself in the end), and Texel 1.07. Maybe some other engines that I like for playing against... such as my own, at some point. The rest of the engines are used for testing, and I don't really mind anymore where they come from or are going to.

There are engines (and developers) I like, and if I can, I will assist with either programming hints/tips, or by testing their engines. A few of them are on this forum; but they're not top 100 engines (yet). I've always wondered where the hundreds of engines come from... we should have 500 active people on this forum. We don't... when looking for engines in the 1600-2100 space, it actually seems most are abandoned/broken/unfindable.

I like that CCRL independently tests my engine, for free, so I can get a sense of where it stands among other engines. If they want to do that with Fat Fritz 2 (and not call it a Stockfish clone), I don't mind. I won't be buying it, because I have the Fritz GUI already, and what I have is good enough for now. There will be people who will buy it just because it's a Chessbase product. Some others will buy it because it sits on the rating list at the top spot with +1 for a month... in an Elo range most of the people buying the program can't even fathom.

Stockfish is open source; it's GPLv3. It can be recompiled (with a different net) and resold for any price; just like any other GPLv3 software. It says so on the FSF website. The only thing necessary is to provide the source, and the credit. If Chessbase neglected to do either one, they could be taken to court, but that is something the Stockfish devs have to decide upon, if it is actually worth it.

The one thing I'm NOT sure about: is the net that Chessbase used open source/free downloadable? Because, Stockfish is GPLv3, and any larger work you create out of it by recompiling its code into it, therefore ALSO has to be GPLv3. If the net is not GPLv3, I think/assume Chessbase is violating SF's license. (They could circumvent this, by making Stockfish load the net from disk, for example, so it's not part of the code.)

[MikeB]

Time for the chess community to speak out and try to get a wrong righted! Bring it on!

The following sentence may not be a very popular opinion, but....

I don't really care anymore. It's almost impossible to know which engine is a clone, or which engine isn't. I mainly disregard any engine that isn't open source (except for testing my own engine; the only thing I care about in that case is the rating), because literally everything can be in there.

To be honest, I've lost interest in the Elo-race at the top. Of course engines are getting stronger all the time; it would be strange if they weren't. They're already so strong that everything over CCRL 3000 is basically godlike. Nowadays, it's easy to have your own chess engine if you want to; just grab one of the 3000+ engines, change a bit here or there, put a network in it to add 300 Elo for free, and re-release it under a new name.

Personally I stick with Stockfish 10 for playing, using the Skill level (11 calibrated against CCRL 1350, which makes the engine, IMHO, too strong for a beginner like my girlfriend, and I don't have enough levels to choose from myself in the end), and Texel 1.07. Maybe some other engines that I like for playing against... such as my own, at some point. The rest of the engines are used for testing, and I don't really mind anymore where they come from or are going to.

There are engines (and developers) I like, and if I can, I will assist with either programming hints/tips, or by testing their engines. A few of them are on this forum; but they're not top 100 engines (yet). I've always wondered where the hundreds of engines come from... we should have 500 active people on this forum. We don't... when looking for engines in the 1600-2100 space, it actually seems most are abandoned/broken/unfindable.

I like that CCRL independently tests my engine, for free, so I can get a sense of where it stands among other engines. If they want to do that with Fat Fritz 2 (and not call it a Stockfish clone), I don't mind. I won't be buying it, because I have the Fritz GUI already, and what I have is good enough for now. There will be people who will buy it just because it's a Chessbase product. Some others will buy it because it sits on the rating list at the top spot with +1 for a month... in an Elo range most of the people buying the program can't even fathom.

Stockfish is open source; it's GPLv3. It can be recompiled (with a different net) and resold for any price; just like any other GPLv3 software. It says so on the FSF website. The only thing necessary is to provide the source, and the credit. If Chessbase neglected to do either one, they could be taken to court, but that is something the Stockfish devs have to decide upon, if it is actually worth it.

The one thing I'm NOT sure about: is the net that Chessbase used open source/free downloadable? Because, Stockfish is GPLv3, and any larger work you create out of it by recompiling its code into it, therefore ALSO has to be GPLv3. If the net is not GPLv3, I think/assume Chessbase is violating SF's license. (They could circumvent this, by making Stockfish load the net from disk, for example, so it's not part of the code.)

Excellent comments ...

[towforce]

As the author of several of these “clones,” I’m not comfortable with this term.

If the code you take is on a source control site like Github or Bitbucket, and your copy also goes onto such a site with the same licence, it would be better to call it a "fork" IMO.

[towforce]

...we all have opinions on whether that was legal or not and how that would play out in a civil court - who knows.

I am NOT reopening that dispute (if anyone wants to do that, please do so in that thread, not this one!), but looking back, I've realised that, while it's understandable that people have different opinions about what a court would decide, one aspect of it makes the whole thing almost laughable: for CB to be allowed to keep their net private:

1. Nets would have to be able to be copyrighted
2. A net would have to be able to be excluded from the GPL

Lots of argument about points (1) and (2) individually, but in actual fact, very few, if any, actually said "yes" to both - so no idea why people (including myself) fought so hard about the two individual points!

btw another reason why it's not important, as stated several times, people who want the strongest chess program will download the latest SF, people who buy FF2 will be getting a lot more than SF + new net.

[dkappe]

As the author of several of these “clones,” I’m not comfortable with this term.

If the code you take is on a source control site like Github or Bitbucket, and your copy also goes onto such a site with the same licence, it would be better to call it a "fork" IMO.

A fork of my own engine, a0lite julia? Seems peculiar. Or a fork of a network? It’s time to retire the “Schimpfwort.”

Also, my latest networks are all released under a GPL incompatible license.

[towforce]

A fork of my own engine, a0lite julia? Seems peculiar.

Maybe call that a "version".

Or a fork of a network? It’s time to retire the “Schimpfwort.”

Yes - actually: if you make a "plug compatible" network, it wouldn't be a crime to call it a fork. You'd probably call it a "new network" though.

Also, my latest networks are all released under a GPL incompatible license.

Probably not critically important under most circumstances.

[Koivisto]

I've always wondered where the hundreds of engines come from... we should have 500 active people on this forum. We don't... when looking for engines in the 1600-2100 space, it actually seems most are abandoned/broken/unfindable.

Almost no discussion that is relevant to development of chess engines goes on in this forum, why should we be active here? Talkchess really is not the centre of the chessprogramming world.