Serious concern about RECKLESS engine

[evqs]

Hello,

I am reporting SERIOUS CONCERN about RECKLESS engine

I have more than THIRTY years of programming experience

and after taking the time to compare both projects, I find the similarities between RECKLESS and Stockfish extremely STRIKING


(ONE)
Make repeated bench runs identical
cj5716Disservin
cj5716
authored and
Disservin
committed
on Jun 9, 2024
Commits on Jun 6, 2024
Simplify evaluation constants
cj5716vondele
cj5716
authored and
vondele
committed
on Jun 6, 2024
Commits on May 24, 2024
Optimise pairwise multiplication
cj5716Disservin
cj5716
authored and
Disservin
committed
on May 24, 2024
Commits on May 21, 2024
Simplify ClippedReLU
cj5716vondele
cj5716
authored and
vondele
committed
on May 21, 2024


Now this individual is coding for RECKLESS chess engine:
Update a comment regarding MAX_MOVES (#734)
cj5716
cj5716
authored
last month
·



(TWO)

Code: Select all

Reckless-dev                       =0  +27/=46/-27 50.00%   50.0/100
Stockfish-dev            		   =0  +27/=46/-27 50.00%   50.0/100

The move count average was almost the SAME, this is definitive PROOFthat RECKLESS engine is a CLONE of well-established Stockfish

I have thirty years of programming experience, and I bothered to inspect SEVERAL Codes from both engines

The similarity is STRIKING !!!

It seems that the plagiarist changed the File Extension from CPP to RS, but the codes look otherwise identical :

Code: Select all

    void operator<<(int bonus) {
        // Make sure that bonus is in range [-D, D]
        int clampedBonus = std::clamp(bonus, -D, D);
        T   val          = *this;
        *this            = val + clampedBonus - val * std::abs(clampedBonus) / D;

        assert(std::abs(T(*this)) <= D);
    }

STOCKFISH

Code: Select all

 fn apply_bonus<const MAX: i32>(entry: &mut i16, bonus: i32) {
    let bonus = bonus.clamp(-MAX, MAX);
    *entry += (bonus - bonus.abs() * (*entry) as i32 / MAX) as i16;
}

RECKLESS


Yes — they are doing the same thing.
Simple explanation
Think of it like this:
• You have a value (entry / *this)
• You want to add a bonus
• But the closer the value gets to a limit (D / MAX), the smaller the bonus becomes
So instead of just doing:

value += bonus
They both do:

value += bonus - (a reduction based on how big value already is)

Step-by-step (plain and simple)
Both versions:
1. Clamp the bonus
→ Keep it between -D and D (or -MAX and MAX)
→ So nothing too extreme happens
2. Reduce the bonus if the value is already large
→ If your value is near the limit, you don’t get the full bonus
→ Prevents overshooting
3. Apply the adjusted bonus


Proof that these codes were plagiarized

Very suspicious


I ask MODS to look into RECKLESS


Regards
Quinn

[Uri Blass]

Reckless is choosing different moves than Stockfish so it cannot be a clone of stockfish.

Reckless did not find 14.Kb1 in the following game and after Reckless blundered by 14.g5 stockfish eval went down and stockfish drew.

https://tcec-chess.com/#div=sf&game=7&season=29

positing some lines of the code that is even not about chess and can be for every thinking game to claim that one is a clone of the other is not convincing for me.

[evqs]

Hello Uri

The Evidence is very clear to me! Houdini also played different moves than Stockfish, it does not prove anything!

I reviewed many Code Files and the similarities are very STRIKING

you can check for yourself, I have found a copy at : https://github.com/codedeliveryservice/ ... /search.rs

[Frank Quisinsky]

1.
Engine Origins, the better TalkChess selection?

2.
What is your goal in posting this in this section? To get more attention as a programmer with so many programming experience, more than THIRTY years of programming experience ... your name I never heard before.

Back:
It is very striking that many programs are now trailing behind Stockfish, but none of them is able to overtake it. I'm not saying that all these programs are probably clones. Among the many people who have contributed to Stockfish, there are usually also programmers who have developed very powerful engines of their own. Unfortunately, that’s often forgotten. In my opinion, it’s actually the engine’s playstyle that matters most these days. That’s what people who watch engine-engine matches enjoy. They want to explore the possibilities available to them. Good ideas should be put to use and improved upon, and I am also quite certain that Stockfish can be surpassed when it comes to measuring playing strength.

Thankfully, discussions about clones no longer hinder development today.

Have fun with Reckless.
The move average is 11 moves lesser compared to Stockfish 18 (for every game without adjudication / resign = off). With the result ... much more fun in watching the games in live mode.

But after all:
Stockfish is the best computer chess project I ever saw after my expierence of 50 years computer chess. I started with 9 years. Absolutely no doubt about is the fact: Stockfish without NN, 5 years old, was in my opinion the biggest computer chess sensation if I compare the strength with all the super strong NN engines. Stockfish was miles ahead of its time. And all this is created by a group of many programmers.

Perhaps you helps also with your 30 years computer experience??
Much better as to wrote such messages!

Back again:
The days of using such messages to stir up trouble on TalkChess are over. They’ve been over for many years now. Let’s enjoy all these amazing developments. There are incredible ideas everywhere. It’s amazing what’s available these days. Getting to grips with the available material... that's a task for “Several Lives.”

[Uri Blass]

Hello Uri

The Evidence is very clear to me! Houdini also played different moves than Stockfish, it does not prove anything!

I reviewed many Code Files and the similarities are very STRIKING

you can check for yourself, I have found a copy at : https://github.com/codedeliveryservice/ ... /search.rs

1)There is a difference between clone and derivative.
Clone means the same
Houdini was not a clone and the problem with houdini was that it came without a source code as commercial program when I understood the claim was that it was a derivative of stockfish.

Reckless is a free source code so I see no problem with it.

Using good ideas in the soutce code oif stockfish when you use a different programming language is not a problem and all top engines use ideas from stockfish.

[Frank Quisinsky]

First Houdini was 99% Robbolite, created by Norman Schmidt.
All is mine, was the comment by Houdini programmer in TalkChess after he released v1.
I forgot his name, sorry!

Uri, you know ...
People like that come and go and tend to liven up the scene.
That's often the same situation with other names.
In most of cases are own interest the reason or group building or to slander others.
That's so boring ...

I learn the different between ZuriChess Neuchatel / ZuriChess Nidwalden for some months.
Now, in April I start the test ... I will know that.

A lot of fascinating questions.
I often don't know where to start.

[Ciekce]

So instead of just doing:

value += bonus
They both do:

value += bonus - (a reduction based on how big value already is)

Yes, this is how history works in every modern engine. Keep up.

Programming for longer than I've been alive and still making posts like this on the internet. Impressive

[chrisw]

So instead of just doing:

value += bonus
They both do:

value += bonus - (a reduction based on how big value already is)

Yes, this is how history works in every modern engine. Keep up.

Programming for longer than I've been alive and still making posts like this on the internet. Impressive

1. In SF value is 16 bit, most engines nowadays are the same, so without the reduction the adjust += bonus will overflow, so its kind of forced to apply a limit. If one goes to 32 bit without reduction, there’s still a danger of overflow.
2. Value gets used in history and if it gets uncontrollably large, very large history values will swamp various extension/reduction/pruning decisions. Not a good idea.
3. History variable is supposed to reflect the more recent parts of the search tree, if it is not bounded (+= reduction), then very old contributions to history will not be faded away.
All in all += reduction is eminently sensible and its use can’t be used therefore to “prove” copying.

[Frank Quisinsky]

Conor,

now, you gave other programmers important helps since now a longer time.
People like that to read. I am speaking from TalkChess messages.

But I don't really have a clue how programmers communicate and help each
other on the various platforms these days.

But what I know is, your name is very popular.
Good ...

Best
Frank

[Dann Corbit]

Both are GPL projects so there cannot be a law violation unless SF contains patents.
I think the idea of open source is to share ideas. i don't think it is just a simple translation. And if no ideas of SF were examined, then that would be strange. Everyone else does that.
Now that the evaluation is nothing but a pile of math answers created by brute force, nobody is that sensitive about evaluation, and I guess reckless does not use SF nodes.
I think there is a forum for complaining about cloning. I think that this thread is better situated there.