mclane wrote: ↑Mon Feb 16, 2026 5:09 pm
And today we mainly see engines similar to stockfish, the strongest open source engine .
And then the engines produce draw scores of 70% ! And people conclude: chess is finished. Stockfish plays like god.
Lets change the game to Go or whatever more complex game.
You clearly make an interesting point here, Thorsten, if we assume that most successful developers follow a similar path as Stockfish, which feels plausible.
But shouldn't pure ELO of said engines somehow balance this?
Actually I feel your point is the issue with the Kittinger engines back then in the past, they were just weaker than the Lang ones. Problem is that the Lang engines were zillions, and the Kittinger engines few. And the mathematics indeed worked against them given the way SSDF used to select opponents, so the difference in ELO was probably exagurrated.
Even worse. Vancouver 68020 beat vancouver 68000 (1/2 the speed) and almeria16 bit lost to almeria32 and both to to vancouver32 and so the latest lang engines got many points alone by playing slower versions and predecessors of themselves.
We had these effects in early rating lists, and we have similar effects today.
The best would be if the engines were all closed code and all engines were VERY different from each other and each engine is only participating ONCE.
But often we see stockfish 18 versus stockfish 17 versus stockfish 16 versus clones of stockfish.
And then you wonder that draw rate goes high ??
If i let vancouver versus almeria versus portorose versus dallas versus roma versus amsterdam i also get very strange results because its lang versus lang versus lang.
So if you do rating lists, use different engines.
And do not test zillions of iterations of the same engine.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
The best would be if the engines were all closed code and all engines were VERY different from each other and each engine is only participating ONCE.
No.
Sounds narrow-minded and dictatorial to me. I don't know where you did get the idea, that the people who write chess-engines have to do it in a manner that pleases you or matches your imagination of such programs (stylewise or else).
You're free to ignore btw. The lists, the games etc.
cpeters wrote: ↑Mon Feb 16, 2026 6:49 pm
No.
Sounds narrow-minded and dictatorial to me. I don't know where you did get the idea, that the people who write chess-engines have to do it in a manner that pleases you or matches your imagination of such programs (stylewise or else).
You're free to ignore btw. The lists, the games etc.
Of course he is.
But he adresses a relevant issue IMHO. I have no idea about engines like Reckless (the new shit in town I suppose), Viridithas or Horsie. You could say "I ignore them", but this would be wrong, too. I simply don't care enough to look them up, although they might +very+ well deserve much better treatment from me.
This is a bit like what is discussed in real life working market. Users may have become rarer than programmers, to take it a little too far and to be taken with a +huge+ grain of salt and humour.
New engines get the CCRL guys and Stephan Pohl ( if strong or aggressive enough) for download - this is like 10 users?
I understand that most programmers just do engine programming for sake of internal competition and to be an esteemed part of the developer community - but isn't it still a valid question what users would be interested to see as new developments?
mclane wrote: ↑Mon Feb 16, 2026 5:09 pmToday we have a very high draw rate because of ??
Certainly one reason is that everyone and his dog is cloning Stockfish. A more important reason though is that chess is a draw, so winning requires a mistake from the opponent - and as engines get stronger, fatal mistakes become less likely. The same applies to human players: the draw ratio rises with strength.
Human chess is partly about tactics and strategy, but mostly about memory
cpeters wrote: ↑Mon Feb 16, 2026 6:49 pm
No.
Sounds narrow-minded and dictatorial to me. I don't know where you did get the idea, that the people who write chess-engines have to do it in a manner that pleases you or matches your imagination of such programs (stylewise or else).
You're free to ignore btw. The lists, the games etc.
Of course he is.
But he adresses a relevant issue IMHO. I have no idea about engines like Reckless (the new shit in town I suppose), Viridithas or Horsie. You could say "I ignore them", but this would be wrong, too. I simply don't care enough to look them up, although they might +very+ well deserve much better treatment from me.
This is a bit like what is discussed in real life working market. Users may have become rarer than programmers, to take it a little too far and to be taken with a +huge+ grain of salt and humour.
New engines get the CCRL guys and Stephan Pohl ( if strong or aggressive enough) for download - this is like 10 users?
I understand that most programmers just do engine programming for sake of internal competition and to be an esteemed part of the developer community - but isn't it still a valid question what users would be interested to see as new developments?
Well. It's late stage capitalism, I suppose. Abundance in digital (chess-) goods now vs the sparsity like in the nineties. But as there seems to be plenty to chose from, the demand for closed source from a consumer-standpoint just for the sake of whatever (purity?) is nuts. Just restrain yourself then, and let others (me) enjoy the activity of the avantgarde.
There are thousends of chess-related projects on github alone to explore and play with - it's not all Stockfish.
The argument that high draw rates prove "everyone is copying Stockfish" has no evidence. If that were true, unique engines would draw less. They don't. Leela is the perfect counter-example: distinct code, same high draw rate.
I also find it ironic that people are dismissing new engines (Reckless, Viridithas, Horsie) without knowing anything about their internals. How can you claim they are copying Stockfish if you haven't looked at the code?
We should also remember the lesson of Houdini. For years, the community believed it was a unique masterpiece simply because the source was closed. It turned out to be a derivative. Closed source does not guarantee originality; it often just hides the similarities.
Finally, regarding the discussion on Chess System Tal: The real difference between it and these newer engines is simply transparency. We scrutinize the open-source engines because their code is visible and their methods are public. We give closed-source engines a pass because they are black boxes. We should appreciate developers who play with open cards, rather than assuming the worst.
How do you know that lc0 is not using stockfish as sparring partner?
And therefore creates nets that handle stockfish good ?!
... Because the development is completely open source and if you want to you can look up exactly what went into it. Also, are we really gonna argue whether Leela is different from Stockfish?
mclane wrote: ↑Tue Feb 17, 2026 10:43 am
How do you know that lc0 is not using stockfish as sparring partner?
And therefore creates nets that handle stockfish good ?!
To see that you're seemingly to lazy to educate yourself is disappointing.
