will Tcec allow Stockfish with a Leela net to play?

[connor_mcmonigle]

I urge everyone (especially top testers) to not be fooled...don't let this become the blind leading the blind. evaluate the engines, look at the source code, seek impartial advice (and that certainly excludes Andrew and Connor who are pursuing a personal agenda), and make up your own mind.

I'm not trying to improve my position on rating lists by getting Fire removed as you seem to insinuate. Why would I care to jump from something like #30 to #29. I could not possibly care less. I'm not impartial as no one truly is, but I have little to gain or lose here, especially as compared to you.

Andrew's mentioning Seer is rather amusing to me. My training approach is moderately interesting in my opinion, but there are stronger and more interesting engines such as SlowChess and Winter which both further push the boundaries of what's possible with a neural network based evaluation function.

I find the case of Fire irksome as it's extremely derivative and has stolen and continues to steal attention from more original efforts. This hobby dies when everyone just decides to copy Stockfish and not contribute anything of their own as they're so caught up on topping rating lists that they can't be bothered to explore anything new. You write "developers shouldn't be forced to reinvent the wheel with NNUE". Have you ever wondered why people keep reinventing (read: exploring new ideas) search and evaluation functions? If every author copied as much from Stockfish as you have, computer chess would be in quite a sorry state indeed.

[kranium]

Regardless of whether we consider your old evaluation function (copied from Houdini with minimal modification) or the recent NNUE addition (copied from Stockfish/CFish without any modifications - a pretty big overlap) your position doesn't really improve.

Connor, maybe a good idea to look before you make absurd and unfounded accusations
The truth is: Fire's nnue code is significantly modified and optimized.

please see
https://github.com/FireFather/fire/blob ... e/nnue.cpp

' Thanks to Yu Nasu, Hisayori Noda, this implementation adapted from R. De Man
and Daniel Shaw's Cfish nnue probe code https://github.com/dshawul/nnue-probe
modified as follows via clang cpp core guidelines:
declaration and assignment joined
deprecated C++ headers replaced [modernize-deprecated-headers]
macros used to declare a constant changed to 'constexpr' constant [cppcoreguidelines-macro-usage]
parameters made const
c-style casts replaced with C++ casts
name does not match rule 'Global constants'
name does not match rule 'Enum members'
auto used when initializing with a cast to avoid duplicating the type name [modernize-use-auto]
zero constants replaced with nullptr
local variables made const
cast from 'const char *' to 'char *' dropped const qualifier [clang-diagnostic-cast-qual]
redundant C-style casts removed"

[connor_mcmonigle]

The fact that you think those (completely cosmetic) changes constitute "modifications" should be telling.

[connor_mcmonigle]

I hope you realize you've just gone from

The problem with testing Fire with a Stockfish network is that Fire is barely more original than the other Stockfish clones you listed. This is not controversial. You can ask any of the prominent Stockfish contributors for their opinion about the originality of Fire and hear something similar to the above.

to

Many Stockfish developers have voiced their annoyance with Fire and the attention it has received on other forums. Many have stated that they view it as a Stockfish derivative recognizing several of their specific patches in your search code. You'll have to ask them if you don't believe me.

in a very very short time

and now I have to ask them about it?

But the truth is: as of today I'm aware of 1 SF contributer recognizing a piece of code he added to SF as a patch. It's can be found in the engines origins forum.
I don't believe anything else has ever been said or posted.

I'm not comfortable quoting people's comments on other forums here. In any case, "You can ask them and hear something similar" and "They believe Fire is highly derivative. You'll have to ask them if you don't believe me" are entirely consistent. I don't see your proposed inconsistency in any case. Also, both of your quotes from MikeB and Thomas Miller are out of context. Both revised their opinions after they were informed that you copied Houdini's evaluation function.

[kranium]

The fact that you think those (completely cosmetic) changes constitute "modifications" should be telling.

Yet, Fire improved 174 elo when nnue was added...
much more than any Segio V. RL net would provide

[kranium]

Also, both of your quotes from MikeB and Thomas Miller are out of context. Both revised their opinions after they were informed that you copied Houdini's evaluation function.

Nonsense not true
are you going to continue simply making stuff up?

[connor_mcmonigle]

I, personally, didn't really find anything conclusive when I read over the evaluation function code outside of the general layout being fairly similar to Stockfish and I'd agree that many top engines share such evaluation terms.

Within the Stockfish family Fire 8.1's highest depth-1 similarity is 64.21% with Stockfish 8 so I compared its sources with this version and I must agree with you. There are things which are common in many top open source engines like the previously mentioned ones and I would say that a couple of them are much more influenced by Stockfish than Fire is but that's it. If I didn't know Houdini is a Stockfish derivative and Fire is based on Houdini, I wouldn't think Fire's evaluation function could be considered derived from Stockfish at all. If anyone is interested I can make a post with the notes I took during the comparison.

Yes. Houdini's evaluation function was undoubtedly the most original part of the engine (though there is some speculation that it was RE'd from Komodo's evaluation function). Obtaining the Houdini source code is rather difficult now, but if you're able to compare against that, the similarities are striking.

Indeed ...

I'm not lying. Perhaps you missed this?

[kranium]

I, personally, didn't really find anything conclusive when I read over the evaluation function code outside of the general layout being fairly similar to Stockfish and I'd agree that many top engines share such evaluation terms.

Within the Stockfish family Fire 8.1's highest depth-1 similarity is 64.21% with Stockfish 8 so I compared its sources with this version and I must agree with you. There are things which are common in many top open source engines like the previously mentioned ones and I would say that a couple of them are much more influenced by Stockfish than Fire is but that's it. If I didn't know Houdini is a Stockfish derivative and Fire is based on Houdini, I wouldn't think Fire's evaluation function could be considered derived from Stockfish at all. If anyone is interested I can make a post with the notes I took during the comparison.

Yes. Houdini's evaluation function was undoubtedly the most original part of the engine (though there is some speculation that it was RE'd from Komodo's evaluation function). Obtaining the Houdini source code is rather difficult now, but if you're able to compare against that, the similarities are striking.

Indeed ...

I'm not lying. Perhaps you missed this?

I agree with Thomas and Mike completely...Houdini's eval was quite unique
I've said that many times, yet many (especially yourself and Andrew) prefer to continue to spread the falsehood that Houdini is a simple SF clone
And your campaign seems to be working, as Houdini has been discredited removed as competition for Ethereal in TCEC etc

[connor_mcmonigle]

The fact that you think those (completely cosmetic) changes constitute "modifications" should be telling.

Yet, Fire improved 174 elo when nnue was added...
much more than any Segio V. RL net would provide

Lol. Yes Norman, clearly changing the qualifiers on some types and replacing typenames with "auto" was worth a lot of elo /s

Surely you're not serious?

[connor_mcmonigle]

I, personally, didn't really find anything conclusive when I read over the evaluation function code outside of the general layout being fairly similar to Stockfish and I'd agree that many top engines share such evaluation terms.

Within the Stockfish family Fire 8.1's highest depth-1 similarity is 64.21% with Stockfish 8 so I compared its sources with this version and I must agree with you. There are things which are common in many top open source engines like the previously mentioned ones and I would say that a couple of them are much more influenced by Stockfish than Fire is but that's it. If I didn't know Houdini is a Stockfish derivative and Fire is based on Houdini, I wouldn't think Fire's evaluation function could be considered derived from Stockfish at all. If anyone is interested I can make a post with the notes I took during the comparison.

Yes. Houdini's evaluation function was undoubtedly the most original part of the engine (though there is some speculation that it was RE'd from Komodo's evaluation function). Obtaining the Houdini source code is rather difficult now, but if you're able to compare against that, the similarities are striking.

Indeed ...

I'm not lying. Perhaps you missed this?

I agree completely...Houdini's eval was quite unique
I've said that many time, yet many prefer to continue the falsehood often re-iterated by yourself and Andrew that Houdini is a simple SF clone

"..., the similarities are quite striking"
"... Indeed"

Thomas Miller later asked to edit the post you're now quoting in that very thread as well...