OpenChess

minutka15 wrote: Wed May 29, 2024 3:36 pm

OrgZ wrote: Wed May 29, 2024 3:11 pm C_Dragon v3.0 Private continuation from 2022, now With Random OP. MPV code by MZ
https://pixeldrain.com/u/FMLx8z5n

Nets:
nn-c721dfca8cd3.nnue
nn-baff1ede1f90.nnue

Thank you Tanick!
Would be nice if you could make an engine with small net (around 15-20 mb) to use for blitz games
That is always i am looking for

You are welcome.
The last engine that i derived and gave a small net of 6mb was Neon Tetra a book engine. Very weak but effective for book tests. Unless i missed something but nets of 15-20 mb i haven't seen. Latest nets are huge 60mb 3072 dimensions...older ones are about 40-50 mb that is 2560 dimensions.

Thanks, Tanick.

Thank you very much for your contributions, master. I hope one day you will make an engine with the possibility to modify certain parameters externally and create personalities and styles for human engine training games.

Here is my repack of latest Tanick Ramz engines
All credit goes to master - Tanick Ramz !

Only for chessbase programs as fritz 10-19
https://pixeldrain.com/u/szPUADwL

Thank you very much!

minutka15 wrote: Thu May 30, 2024 5:43 pm Here is my repack of latest Tanick Ramz engines
All credit goes to master - Tanick Ramz !

Only for chessbase programs as fritz 10-19

Tanick_Ramz.png

https://pixeldrain.com/u/szPUADwL

Thank you minutka15.

@capachess
Thanks, well so far i have written move by depth (i use this specifically for my books) code, and time per move (human adaptation purposes)...with time i'll merge them all in one engine.

On Raid and Tactical gear
i have Written Gear shift and Raidlimit (polished by JaRHeads)

Gear Shift (Which runs on 3 stages, which need one to adjust in the code itself not through the ucioption. thus the default stage for Raid 3.6 is 2 ):
to randomly shuffle and order possible moves based on previous evaluations
that yielded higher scores at a depth greater than 13. If no
historical evaluations are available, it reverts to the engine's
current evaluations to avoid missing strong moves. Additionally,
the "RaidLimit" is a time management system designed to minimize
the computational time spent on the Gear Shift process, ensuring
efficiency in move randomization and reordering.

Thank you for your explanation, and I look forward to the engine, in which we can customize and humanize more, with other modifiable parameters. Thank you very much master, sorry for my bad English

TEST by MR Adrian Oliva

C_Dragon v3.0_avx2 vs. Steel Fighter 2-avx2

Processor AMD A10 – 9700 R7,10 Compute Cores 4C + 6G - 3.50 GHz - RAM 8.00 GB - 64-bit operating system - x64 processor - Windows 10 Pro - 1024 mb hash Fritz 16 GUI – Permanent Brain: OFF – Both Engines Play with 4 Threads and Large Pages - Opening Book: Strongbook 240220 ECO + 6ply ctg - Final Tablebases: Syzygy 3-4-5 pieces – C-Dragon v30: nn-c721dfca8cd3.nnue and nn-baff1ede1f90.nnue - Steel Fighter 2: nn-ea6ac98d29d4.nnue and nn-321bbccb5bcf.nnue

Book Settings:

1.Tournament ON
2.Min games 3-30
3.Variety of play 235
4.Influence of learn 0
5.Learning strength 0

DESKTOP-TORM2PL, Blitz 3min+1seg 0


1 C_Dragon v3.0_avx2 +3 +20/=464/-16 50.40% 252.0/500
2 Steel Fighter 2-avx2 -3 +16/=464/-20 49.60% 248.0/500

Games in PGN: https://www.mediafire.com/file/cmudm70h ... 6.pgn/file
https://chess.massimilianogoi.com/forum ... 2659#p2659

Congratulations OrgZ!

Homayoun wrote: Thu Jun 06, 2024 9:56 pm

+1