Future of NNUE is Dimension 3072 network!

[Eduard]

One last gift before you go: https://cdn.discordapp.com/attachments/ ... 37a51e601&

This posting alone shows that some people abuse the forum for senseless provocation. If Viren thinks it's a great achievement to achieve something like this, keep it up. You don't know anything about me. These people like Viren are not interested in constructive content. But other people here also enjoy sharing nonsensical philosophical content. These people know nothing about chess. Keep it up. The forum gets the kind of people it deserves. I know that chess players are becoming fewer and fewer here. I won't write much more here either. I correspond with many chess players and many are GMs. Why don't these people post here? But they write me emails and encourage me. Thanks.

[Draude]

You don't know anything about me.

You don't know anything about me. People like you are not interested in constructive content. You also share nonsensical opinions apparently. You know nothing about chess. Keep it up!

I make good clones, share all my testing work, and post it in this forum. Then you call me a fool and an arrogant fake news poster. If you want to prove it, download Fat Fritz 2 and Stockfish dev. yourself and test it on the position I posted, and then call me out for saying crap, but not before please! Thank you!

[Draude]

I correspond with many chess players and many are GMs.

Then I am now under the impression that engine developers who correspond with many GMs are arrogant fools indeed. You have no basic idea of how to test the engine properly. My test positions are clearly superior to all the testing you do! I have thousands of them! Can you solve even half of them? My dimension 4096 engines are the future of chess!

[Eduard]

Then do it and release it here , and don't talk nonsense here. So far you've only written nonsense and can only fake.

[Draude]

Have you released any of your engines? Why did you delete them from your website, if they are so good? Your engine Leptir-Big-Ultra scores very badly against SF dev:
Score of Stockfish 16.1 vs Leptir Big-Ultra-bmi2: 637 - 394 - 969 [0.561]
... Stockfish 16.1 playing White: 632 - 3 - 365 [0.815] 1000
... Stockfish 16.1 playing Black: 5 - 391 - 604 [0.307] 1000
... White vs Black: 1023 - 8 - 969 [0.754] 2000
Elo difference: 42.4 +/- 10.9, LOS: 100.0 %, DrawRatio: 48.4 %
2000 of 2000 games finished.

Tested at 20 minutes + 5 seconds, 4 threads, Perfect 2021 book.

(Your website) From now on I will pass on all Stockfish versions with my own parameters to my friends only.

I'm your friend, pass me your Stockfish version please

[smatovic]

...

@Stockfish guys:

Maybe start to listen to some of your precious users? Tuning Stockfish to win TCEC/CCCC is just one part of the game, if people use derivatives to solve positions resp. for analysis, there might be a reason for it? It should be no hard task to add different .epd testsuites to your testing framework, also as a kind of regression test....

Fishtest is a distributed task queue for testing chess engines through self-playing. The main instance for testing the chess engine Stockfish is at this web page: https://tests.stockfishchess.org

https://github.com/official-stockfish/fishtest/wiki

--
Srdja

[Draude]

...

@Stockfish guys:

Maybe start to listen to some of your precious users? Tuning Stockfish to win TCEC/CCCC is just one part of the game, if people use derivatives to solve positions resp. for analysis, there might be a reason for it? It should be no hard task to add different .epd testsuites to your testing framework, also as a kind of regression test....

Yes! This is very good suggestion! I think Stockfish devs need to listen to it. In my opinion, a test suite with at least 6000 positions, to test the mate finding ability of SF dev. versions! It has been getting worse and worse at solving my difficult test suites. Many of the position it fails are complicated checkmate sequences.

Any Stockfish devs here, please construct such a test suite, thank you!

[smatovic]

...

Alright kid, at least I tried.

--
Srdja

[Viren]

@talkchess guys:

Maybe start to use your brain? We already have a test suite for mates that is used to reject patches:
https://github.com/official-stockfish/S ... 2002504682

Test suites cant be used to measure small differences accurately because they have a small sample size of positions, which may also be biased. They can also be wrong: All moves could actually draw, or multiple moves could actually win.

[smatovic]

@talkchess guys:

Maybe start to use your brain? We already have a test suite for mates that is used to reject patches:
https://github.com/official-stockfish/S ... 2002504682

Do you use this mate .epd regression test after every passed patch on fish-test?

Test suites cant be used to measure small differences accurately because they have a small sample size of positions, which may also be biased. They can also be wrong: All moves could actually draw, or multiple moves could actually win.

Then be smart*, it's not meant to measure Elo gain, therefore you have your SPRT self-play.

*testsuites are a moving target.

--
Srdja