SPCC: Testrun of Ethereal 12.62 finished

[Vinvin]

Maybe we will see an Ethereal NNUE engine in the near future.

Nope. NNUE is a one way ticket to kill the originality of the engine.
I'll stick to hand picked terms, slightly adjusted by a minimal NN.

Thanks for the tests. +34 in your testsing; +26 in our regression testing.
Good to see that the self-play gains are inline with the general gains.

Do you know which change(s) gives so many points ?

[AndrewGrant]

Maybe we will see an Ethereal NNUE engine in the near future.

Nope. NNUE is a one way ticket to kill the originality of the engine.
I'll stick to hand picked terms, slightly adjusted by a minimal NN.

Thanks for the tests. +34 in your testsing; +26 in our regression testing.
Good to see that the self-play gains are inline with the general gains.

Do you know which change(s) gives so many points ?

You could go through the repo history https://github.com/AndyGrant/Ethereal/commits/master
15 commits since Ethereal 12.50, each commit message will show SPRT results for the content of the commit.

SPRT, especially with a broad opening book, generally over estimates the elo gains.
However, you can gauge which patches had more weight than others.

But the answer is these 3:
https://github.com/AndyGrant/Ethereal/c ... f359399567
https://github.com/AndyGrant/Ethereal/c ... 5ccd38c171
https://github.com/AndyGrant/Ethereal/c ... 675669cecd

[Rebel]

Maybe we will see an Ethereal NNUE engine in the near future.

Nope. NNUE is a one way ticket to kill the originality of the engine.

I am not so sure that's true.

Here is a sim-test at depth=1 with the current NNUE engines.

http://rebel13.nl/dump/mysim.html

I expected 70-80% but.....

Didn't t you show some week ago that minicnnue is 86% like sf while minic is only 36%? It was just for opening position maybe?

Yes, I remember. I will look into it, hoping I still have the data.

I still have the data (one of my PC's gave up last week) but I am unable to reproduce the result because it's unclear which Sergio nn.bin I used for SF. I am sorry.

[pohl4711]

Thanks for the tests. +34 in your testsing; +26 in our regression testing.
Good to see that the self-play gains are inline with the general gains.

These results match perfectly. Mention, in my testruns, Ethereal 12.50 was popcount-binary and Ethereal 12.62 was new avx2-binary, which is more than 8% faster, than popcount-compile. So, my result of Ethereal 12.62 should be a little bit better (compared to Ethereal 12.50), than your selfplay-testings. And that is exactly what we see...