TalkChess.com

Sergio's NN used (20200807-1645.bin) [Event "Arena Tournament"] [Site "DESKTOP"] [Date "2020.08.07"] [Round "22"] [White "Stockfish_20080621_x64_modern"] [Black "Lc0 26.0"] [Result "1-0"] [BlackElo "2200"] [ECO "B08"] [Opening "Pirc"] [Time "16:56:02"] [Variation "Classical, 5.Be2 O-O 6.O-O Bg4"] [W...

Thanks for your reply

Version stockfish_15103119_x64_modern
made a record of solving
r2qrb1k/1p1b2p1/p2ppn1p/8/3NP3/1BN5/PPP3QP/1K3RR1 w - - bm e5
at ply 29. So it could be a jump in performance.

/Jarkko

stockfish_15102104_x64_modern has done a new record in my small test.
No losses to Houdini3. 30 matches and SF got 88% of points.
Give it a chance.

/Jarkko

Can someone explain what all these values below mean(values in parenthesis and brackets). Do I understand this correctly? How can this test run show positive improvement of 1.36 ELO even thou there are more losses (11896) than wins (11850)? LLR: 1.36 (-2.94,2.94) [-3.00,1.00] Total: 77733 W: 11850 L...

Test with stockfish_14012720_x64_modern_sse42.exe

I my home test this has beaten Houdini 1.5 most biggest margin.

/Jarkko

Hi,

Can someone say how one can add source code improvement suggestion to Stockfish Framework test queue? Does it require some installations or only registration? I have downloaded the master branch where I think the changes should be done.

/Jarkko

If you use contempt value 3 If I understood it correctly you are only using contempt of 3% of a pawn value, not minor piece. So you should use maximum 50 to have real expirience

You can use Crafty's personalities feature and create your own crafty.cpf text file which crafty loads when it starts. You can can almost any parameters in evaluation code.

use

pers save

to get default file to edit

and

pers list

to what you can change and the current values.

Oh well

Thanks for helping out.
I must continue the quest.

I think we can be certain of these things:

1.) Evaluation function parameters have multiple local optimal points.
2.) Only one global optimum which we probably never reach
3.) By tuning one parameter there is no guarantee that we get out of the local optimum valley.