Hi dear chess engines fans! Inspired by the interesting @RSilvian experiment about learning,
SINCE I THINK THAT MIXING .EXP FILES BETWEEN STOCKFISH DERIVATIVES ISN'T SO BAD
(after that on 4nd learning run ProteusSF RBE 008a has won it has scored +7 against my @Mehmet1921 's nightmare engine Kayra 1.4 and it has also scored +7 on 5th learning run against the very strong @dorsz 's Swordfish 15.2 using always TCEC 22
Superfinal Jeroen Noomen's selected 50 positions - 100 games repeating every position on reverse side (thanks @im48)

All 100 Kayra 1.4's PGN GAMES (Please analyze them with me!)

All 100 SwordFish 15.2 PGN GAMES (Please analyze them with me!)
(Sorry @dorsz , It's only a test to demonstrate that mixed .exp files work fine. I know that is unfair to compare two engines if the second hasn't an experience file . Without Khalid Omar's learning feature they have the same strength
)
I'VE STARTED THIS FINAL TEST:
5 mins + 2 secs / ALL | Tablebases Syzygy 6 man
ProteusSF RBE 008a + ProteusSF.exp (@Sarona + trained ProteusSF.exp + @Igor 1959 data)
versus
Eman 7.93 Eman 05-05-22 Depth4 original untouched Eman.exp downloaded from @Anton101 FB page.
NOTE: The file not found on Eman config is the Eman.bin book, to be sure that it not interferes.
Since it's a private engine, unfortunately I can run Khalid Omar's and ProteusSF engines only on Windows 11 ARM64 4 CPUs 2 MN/s (supporting only popcnt SSE41). My full respect and admiration to the programmer that with his learning code is changing chess engines history ONCE AGAIN (after Stockfish 1 released as FREE Open Source and NNUE
)
FINAL RESULT: +1 for ProteusSF RBE 00Bb. An Hitchcock thrilling until the end!!!


ALL 100 PGN games ProteusSF RBE 008b versus Eman 7.93
WOW!! Thank you @Sarona & @Igor for giving me your data to compile my ProteusSF.exp milkshake
(I don't want glorify myself as somebody does
) Dear Mr. Khalid Omar I'm available as betatester for Eman 8.xx . Your learning code is sublime. Your engine's source is the only one that I'd like to examine. I don't care about AI, Zeus, Fischerov, Raubfisch and Solidor's clones.
Maestro training method is deprecated!!
(Just kidding, I will apply it rigorously for next September Proteus Aurora 009 release, starting the training from scratch (but releasing also the ProteusSF.exp Milkshake Edition
)

Last edited by AlexChess on Mon 06 Jun 2022, 4:14 pm; edited 66 times in total
SINCE I THINK THAT MIXING .EXP FILES BETWEEN STOCKFISH DERIVATIVES ISN'T SO BAD
(after that on 4nd learning run ProteusSF RBE 008a has won it has scored +7 against my @Mehmet1921 's nightmare engine Kayra 1.4 and it has also scored +7 on 5th learning run against the very strong @dorsz 's Swordfish 15.2 using always TCEC 22
Superfinal Jeroen Noomen's selected 50 positions - 100 games repeating every position on reverse side (thanks @im48)

All 100 Kayra 1.4's PGN GAMES (Please analyze them with me!)

All 100 SwordFish 15.2 PGN GAMES (Please analyze them with me!)
(Sorry @dorsz , It's only a test to demonstrate that mixed .exp files work fine. I know that is unfair to compare two engines if the second hasn't an experience file . Without Khalid Omar's learning feature they have the same strength

I'VE STARTED THIS FINAL TEST:
5 mins + 2 secs / ALL | Tablebases Syzygy 6 man
ProteusSF RBE 008a + ProteusSF.exp (@Sarona + trained ProteusSF.exp + @Igor 1959 data)
versus
Eman 7.93 Eman 05-05-22 Depth4 original untouched Eman.exp downloaded from @Anton101 FB page.
NOTE: The file not found on Eman config is the Eman.bin book, to be sure that it not interferes.
Since it's a private engine, unfortunately I can run Khalid Omar's and ProteusSF engines only on Windows 11 ARM64 4 CPUs 2 MN/s (supporting only popcnt SSE41). My full respect and admiration to the programmer that with his learning code is changing chess engines history ONCE AGAIN (after Stockfish 1 released as FREE Open Source and NNUE

FINAL RESULT: +1 for ProteusSF RBE 00Bb. An Hitchcock thrilling until the end!!!


ALL 100 PGN games ProteusSF RBE 008b versus Eman 7.93
WOW!! Thank you @Sarona & @Igor for giving me your data to compile my ProteusSF.exp milkshake


Maestro training method is deprecated!!



Last edited by AlexChess on Mon 06 Jun 2022, 4:14 pm; edited 66 times in total