SPCC: Testrun of Lc0 805123 finished

[pohl4711]

NN-testrun of Lc0 0.30dev (same binary as in TCEC) with net 805123 finished.


https://www.sp-cc.de


(Perhaps you have to clear your browsercache (press <STRG>+<SHIFT>+<DEL>) or reload the website))

[ernest]

You wrote : "Next ratinglist- and VLTC-regression-testruns: Stockfish 221004"

Well, let's see : NCM (https://nextchessmove.com/dev-builds) found quite a bad score for that version !

(of course NCM is not SPCC )

[Jouni]

Yes rating is now +1.48 ELO over SF15. Minimal progress in 6 months! Is there chance, that Dragon reach SF level now?

[pohl4711]

Yes rating is now +1.48 ELO over SF15. Minimal progress in 6 months! Is there chance, that Dragon reach SF level now?

Perhaps a restart with this (best so far in my testings) version could make sense:
Stockfish 220817 has 3814 SPCC-Elo, means +12 Elo to SF 15,
but this is not my decision, of course.

[mehmet123]

NCM results doesn't show the real power of Stockfish. Stockfish 7 is very weak according to Stockfish 16 Dev versions. If the oppponent is Komodo Dragon we could reach more accurate results.

[pohl4711]

NCM results doesn't show the real power of Stockfish. Stockfish 7 is very weak according to Stockfish 16 Dev versions. If the oppponent is Komodo Dragon we could reach more accurate results.

NCM is nonsense, SF 7 is much too weak as opponent for latest SF devs.
I refer only to my own testings..

[Ozymandias]

Perhaps a restart with this (best so far in my testings) version could make sense:
Stockfish 220817 has 3814 SPCC-Elo, means +12 Elo to SF 15,

How much is it above this one, in your testing?

Author: Michael Chaly
Date: Mon Feb 7 17:30:35 2022 +0100
Timestamp: 1644251435

[pohl4711]

Perhaps a restart with this (best so far in my testings) version could make sense:
Stockfish 220817 has 3814 SPCC-Elo, means +12 Elo to SF 15,

How much is it above this one, in your testing?

Author: Michael Chaly
Date: Mon Feb 7 17:30:35 2022 +0100
Timestamp: 1644251435

1 Stockfish 221004 avx2 : 3807 8 8 7000 69.0% 3664 61.4%
2 Stockfish 220927 avx2 : 3804 8 8 7000 68.7% 3664 62.1%
3 Stockfish 15 220418 : 3802 7 7 8000 69.5% 3655 60.7%

[Ozymandias]

I see 221004 and 220927, but not 220207.

[pohl4711]

I see 221004 and 220927, but not 220207.

Oh. 220207 is long ago...
That was before release of SF 15...

If I look in my old diagrams (with Zoom), I can say, I did not test 220207, but 220205: 3804 SPCC-Elo