SugaR AI ICCF

[M.Z]

It is a small/simple fix "SugaR AI Experience Tools 1.04"
https://github.com/Zerbinati/SugaR-AI-ICCF/commit/1bb538ae8e48dbdccf9d4d81a98badbd169eb8e4

[M.Z]

https://github.com/Zerbinati/SugaR-AI-ICCF
SugaR AI ICCF 1.70
Restructuring PolyGlot book code and separating Book1 and Book2 options
https://github.com/Zerbinati/SugaR-AI-ICCF/commit/21ec881abb88b024013caa44be970aeac2ae933c
1) Restructuring of PolyGlot book code
2) Separating Book1 and Book2 options
3) Adding separate Book1 Depth and Book2 Depth option
master (#5) ICCF170
@khalid-a-omar
khalid-a-omar committed

Release:
https://github.com/Zerbinati/SugaR-AI-ICCF/releases/tag/ICCF170

[M.Z]

https://github.com/Zerbinati/SugaR-AI-ICCF
SugaR-AI-ICCF 1.80
Crash fix when all available experience book moves are leading to a draw!
https://github.com/Zerbinati/SugaR-AI-ICCF/commit/a1b552f8d4d193fd9cbb652c72753d8252e52004b8bdb
a1b552f
master (#12)
@khalid-a-omar
khalid-a-omar commit

Release:
https://github.com/Zerbinati/SugaR-AI-ICCF/releases/tag/ICCF180

[M.Z]

https://github.com/Zerbinati/SugaR-AI-ICCF
SugaR AI ICCF 1.90
Aligned with:: "Simplify and unify FRC cornered bishop."
https://github.com/official-stockfish/Stockfish/commit/62a0b65ff886f8f4895d854705c0c870e6a2a834
-SugaR AI Experience Tools 2.00
-Improved Learning Code
by Khalid Omar
Changelog:

Experience hashing algorithm is changed to increase the speed of loading (500% faster) and the speed of probing during search (115% faster) which increases the overall nodes/second (NPS) of the engine. This comes at the expense of memory since SugaR now uses around 15% more RAM for experience.

Completely revised the experience from PGN feature to filter PGN games more accurately and avoid mistakes and conflicting evaluations in PGN games, also to detect Black point-of-view which is -unfortunately- different between the different GUI's (example: Arena and CuteChess will always save black evals from black point of view, while ChessBase and InfinityChess store black evals from white point of view).

Completely revised the experience book algorithm to play the best moves based on sophisticated experience move quality calculation (which depends on the next 10 moves from a given position). Any move with negative or zero quality will be excluded from the experience book. If Experience Book Best Move is false then the engine will play one of the top 50% moves (with positive quality) available for a given position, otherwise, the engine will play the best move (maximum quality) always.

Release:
https://github.com/Zerbinati/SugaR-AI-ICCF/releases/tag/ICCF190

[M.Z]

The ICCF version is based on very old code and it becomes more and more challenging to keep it up to date.
Therefore we have decided to close the project.

[peterm]

The ICCF version is based on very old code and it becomes more and more challenging to keep it up to date.
Therefore we have decided to close the project.

Makes things easier of course.
Internal MultiPV and Pure NNUE Eval I'll miss.
Could you maybe implement these two to SugaR AI?
And maybe hash storage too? This one I'd only use with fixed 2G- bug of course.

[M.Z]

The ICCF version is based on very old code and it becomes more and more challenging to keep it up to date.
Therefore we have decided to close the project.

Makes things easier of course.
Internal MultiPV and Pure NNUE Eval I'll miss.
Could you maybe implement these two to SugaR AI?
And maybe hash storage too? This one I'd only use with fixed 2G- bug of course.

Hi peterm
no, the 2 versions were deliberately separate.
the AI version remains optimized for normal play only.

Regards
M.Z

[M.Z]

SugaR AI ICCF had been abandoned a few months ago due to the too old code and lack of time to renew it.
However, a fair number of users have expressed a desire to be able to use this analytics engine again.
So I took some time to work on it and update it.
Here is a first restyling.
SugaR AI ICCF 2.30
Download SugaR AI ICCF 2.30 Release

Regards
M.Z

[peterm]

SugaR AI ICCF had been abandoned a few months ago due to the too old code and lack of time to renew it.
However, a fair number of users have expressed a desire to be able to use this analytics engine again.
So I took some time to work on it and update it.
Here is a first restyling.
SugaR AI ICCF 2.30
Download SugaR AI ICCF 2.30 Release

Thanks a lot for the update, Marco!
"Use NNUE" option isn't there anymore in Fritz GUI, if "classic" eval's wanted, of course nnue file can be deleted, seems to work, but n/s aren't as much lower with loaded net as I would expect.

[peterm]

"Use NNUE" option isn't there anymore in Fritz GUI, if "classic" eval's wanted, of course nnue file can be deleted, seems to work, but n/s aren't as much lower with loaded net as I would expect.

In Shredder GUI the same (no UCI option "Use NNUE"), but this one GUI says "classical evaluation used" at start stop of engine, even with net in UCI option given, so it obviously isn't really loaded. A few test positions I tried tell same story, classical eval's used, high n/s but time to solution not that of NNUE usage.
Missing option "Use NNUE" seems to unload given net regards

[Evo]

Same here in Android version.. my Droidfish also could not read the NNUE EVAL even when I already put the correct path/ name of those Nets.. Sugar ICCF ( Android version ) only ran with classical eval.. with the high speed of kn/s, but failed to read the NNUE net

[peterm]

In Shredder GUI the same (no UCI option "Use NNUE"), but this one GUI says "classical evaluation used" at start stop of engine,

Editing .eng- file in Shredder GUI with
Use NNUE=true
, which is the syntax in AI ICCF 1.90 in Shredder .eng- file, doesn't help neither.
NNUE isn't loaded, "classical evaluation enabled" yet still is Shredder GUI's output regards

[M.Z]

I apologize to all users .. and my mistake!
version 2.30a will be available in a few minutes. xrf03

[M.Z]

Fix stupid bug..

DOWNLOAD SugaR AI ICCF 230a

[Evo]

Fix stupid bug..

DOWNLOAD SugaR AI ICCF 230a

Thx Mr. MZ .. finally one of my fav engines has solved it's problem.. Now Im hoping that one of the "Speed Masters" would like to compile this magnificent engine into Android