Chess Engine Charisma

[Eduard]

Chess Engine Charisma 64-Bit:

Download on my Homepage:
https://solistachess.jimdosite.com/solista-news/

Charisma is a 64-bit UCI chess engine, based on Stockfish developer versions. The code comes from various Stockfish developer versions and Fishcooking test-versions. But some ideas, e.g. the pieces values, are mine. The features, such as the live book, MCTS and the learning code, are identical to Andrea Manzo's great open source engine Brainlearn. Charisma is therefore also open source. I built this engine primarily to analyze openings and practical positions. The engine is a little less selective than Stockfish, but a little more selective than e.g. Corchess. Therefore, the engine can be used universally, e.g. also for playing on various chess servers! It should be noted that the NNUE network is external and in the folder with the engine must be copied. Test result: In my EN-Test 2022, Charisma is currently the leader with 116 out of 120 solutions (60s). In the picture I have shown the settings with which I play on online servers. This includes "MinimumThinkingTime" (100ms) option. This value should only be set on the servers, but should be set to 0 for offline games.

Settings:


Have fun,
Eduard

[Uri Blass]

I am surprised to read about brainlearn with learning code because I thought that after romichess no programmer worked about learning.

I would like to know how much learning help and if solista can improve the result that it get in matches from fixed positions against other engines that do not have learning code.

I wonder what is the reason that almost no programmer release engine with learning code and no testers test learning for rating list.
I would like to see a rating list that every pair of program play 10 games with white and 10 games with black from the same initial position(for example after a line 1.e4 g6 they get 10 games as white and 10 games as black).

If learning is effective then engines that have learning will get better results then engines that do not have learning.
Testers may delete learning after every match of 20 games to have equal conditions because it is not fair that engine A learned from 20 games against B when C learned nothing and later you test A against C in 20 games.

[Eduard]

Brainlearn could always learn. That's why he's called that. I personally don't believe in learning. Some believe in it. It's a nice feature, nothing more. I could also do the engine without learning at all. I prefer to rely on my openings books which I edit often. Nobody has to activate learning. When installing, tick "Read only learning" and the engine will not create a learning file at all. More interesting is MCTS for analysis.

[Eduard]

Update: Charisma NY23
Improved search for longer timecontrols and analyses. New network nn-335a9b2d8a80.nnue.
Downloads:
https://solistachess.jimdosite.com/solista-news/

[supersharp77]

Update: Charisma NY23
Improved search for longer timecontrols and analyses. New network nn-335a9b2d8a80.nnue.
Downloads:
https://solistachess.jimdosite.com/solista-news/

Thank You My Friend!! Great Works!!:)

[Eelco de Groot]

I think I like this one thanks Eduard! I have no idea what all the options do, it's like flying in an alien starship. Don't touch that, it might be the self destruct sequence! MCTS is activated now, that makes it more diferent.

[Eduard]

02 Jan 2023: Charisma Gold

Download on my Homepage:
https://solistachess.jimdosite.com/solista-news/

This engine is a Stockfish derivative, and made primarily for analyzing chess games. The search is a bit slower than Stockfish developer but more accurate. In my EN test 2022, the engine solves 116 out of 120 positions with 75s. The latest NNUE network from 010123 and most Stockfish patches up to 010123 are implemented. It goes without saying that this engine is no Blitz monster. However, the engine is very well suited for games with longer time controls. The engine has no special features like self learning, I wanted a pure analysis tool. The photo shows the settings I use to play blitz and rapid chess games on PlayChess.com.

[Eduard]

There is a new NNUE network again

https://tests.stockfishchess.org/nns

Engine Charisma uses an external network. It's easy to install a new network in Charisma.

[Eduard]

Charisma Diamond 64 Bit:

Download for Windows:
https://solistachess.jimdosite.com/solista-news/

Charisma Diamond is a stockfish derivative with a fast search. The engine is particularly well suited for blitz games and quick analysis. The tactically strong "nn-ad9b42354671.nnue" is used. For those who prefer a more positional approach, I recommend the latest NNUE nets from Linmiao Xu at Stockfish dev. The network is external and it is easy to install new networks. Many Stockfish patches have been implemented, including the latest BUG fix from January 4th 23. In my EN-test 2022, the engine solved 115 of 120 positions. For such a selective engine, that's great. All Brainlearn features are used, such as "Self Learning" (My thanks to Andrea Manzo). MinimumThinkingTime is also implemented for server games without bonus time.

Test:

[Eduard]

Charisma SmallNet released.
(with Patch "Fix stack initialization")

https://tests.stockfishchess.org/nns network "nn-8366015ec235.nnue"

Downloads:

Pixeldrain
https://pixeldrain.com/u/Wkp7UCVq

Filehorst.de
https://filehorst.de/d/eaJiwnGJ

Charisma SmallNet is the little sister of Charisma Diamond. A 12 MB small network is used. In the starting position, the engine as AVX2 version on 20 threads, calculates that much faster:

Charisma SmallNet: 26000 kns after 15s
Stockfish dev: 18000 kns after 15s

In my EN-Test 2022, the engine managed at 60s, great 110 solutions. In the ranking it looks like this:

1) Charisma, Result: 116 out of 120 = 96.6%. Charisma 64-Bit.txt (ZIP)
2) Crystal 5 KWK, Result: 113 out of 120 = 94.1%. Crystal 5 KWK.txt (ZIP)
3-4) Charisma SmallNet (12 MB net), Result: 110 out of 120 = 91.6%. CharismaSmallNet.txt (ZIP)
3-4) Corchess 3 201122, Result: 110 out of 120 = 91.6%. Corchess 3 201122.txt (ZIP)
5) Blue Marlin 15.4, Result: 107 out of 120 = 89.1%. Blue Marlin 15.4.txt (ZIP)
6-8) Dark Sister 1.9a, Result: 106 out of 120 = 88.3%. Dark Sister 1.9a.txt (ZIP)
6-8) Kayra 1.7, Result: 106 out of 120 = 88.3%. Kayra 1.7.txt (ZIP)
6-8) ProteusSF-Piranha 220904, Result: 106 out of 120 = 88.3%. ProteusSF-Piranha 220904.txt (ZIP)
9-10) Stockfish dev 231122, Result: 105 out of 120 = 87.5%. Stockfish dev 231122.txt (ZIP)
9-10) Shashchess 25.3 GoldDi, Result: 105 out of 120 = 87.5%. ShashChess 25.3GoldDigger.txt (ZIP)

All TXT files on my homepage:
https://solistachess.jimdosite.com/testing/

I would like to point out, that the pure engine of Charisma SmallNet is not identical to Charisma Diamond. It's not just the network that's different.

I optimized the engine for practical play on servers. So far I haven't lost a single game on PlayChess.com with the small net! The strength of the small network is the middlegame with tactics. Here the small net is stronger than Stockfish dev. and is faster! Benefits for Stockfish dev. there are special complex positions in the endgame. Possibly also when playing with small books, and complex positions have to be solved at the beginning of the game. In practice, I play with normal theory and with books that already contain the known theory. The actual game usually begins in the middlegame. And here the small fast net cannot be defeated!

The engine contains all Brainlearn features. Thank you Andrea Manzo!

The network is EMBEDDED and the engine does not need a special folder with the network. I wish everyone a lot of fun testing the engine.