Charisma NY23

[benkrik]

Questions to Eduard.

I am amazed how well most of your engines perform both in blitz tournaments and solving problems. But I am a little bit not sure about all the versions of Charisma, NY23, Diamond, Saphir, Gold, Light, Blitz, 250203, etc

From what I understand: Charisma, Charisma NY23, Charisma Gold are slower and more precise. And the best in solving problems. Is Charisma Gold a next version of Charisma?
I think Blitz, Light 2, 250203 have quicker search and good for Blitz and analysis - but are those the same engine, just different versions?
Also Saphir, Diamond - are trying to be both quick and precise - but again, are some of those better/newer than the others?

And how about Solista Attacks and Leptir - how different are those from Charisma-s.

If you could give us some additional "roadmap" to those engines and their comparison would be appreciated!
Thank you!

[DeletedUser190323]

I have to explain something first. I own 4 PCs.

Ryzen 3900X
Ryzen 2700
Ryzen 5700u laptop
Intel i3 dual core laptop

The i3 dual core only manages 850 kns. With the 5700u I play on the go in energy-saving mode. In mains operation, the 5700u consumes about 25 watts and manages 8000 kns. In the battery saving mode, the CPU consumes only 6 watts and manages 2600 kns.

With 6 watts I can play 6 hours non-stop on the server when I'm on the road.

That's why I'm trying to create the right engines for my computers. I need a very fast engine for my slow i3 dual core. I made Charisma 240223 for this. For the Ryzen 5700u, Charisma 250223 seems optimal to me.

For my Ryzen 3900X I only use engines with a slow search. Leptir 280223 is wonderful for this. I also use it to play Blitz on the server.

Leptir is the best engine for intensive analysis. If you don't care about seconds, you should use Leptir.

I myself am not a fan of learning files. But some of my friends like it. Especially for correspondence chess it can be useful to work with learning files. Solista Attack is my engine that works with a learning file. This engine is kept private because I don't want to publish the learning code.

[benkrik]

Thank you so much (I did test Leptir and Charisma Gold in solving problems - and they are better than any other Stockfish or Stockfish Clone that I tried) So for long analysis of difficult positions/problems Leptir 280223 might be slightly better than Charisma, Charisma NY23 and Charisma Gold?

Eduard and what would be the difference between Charisma Diamond, Light 2 and Blitz (all three should have a fast search?) and which one would you recommend for the best performance during bullet/blitz games (how would they be comparable with Leptir 280223)?

[DeletedUser190323]

Charisma 240223 and Charisma 250223 are the followers of the previous Charisma versions. Therefore only these versions are available for download. I will mark the following versions with a date. Get one of these versions.

Can you tell me what CPU you are using? I would also be interested to know which engine do you think is better and which is worse? Would a Charisma version that searches slower be better than Charisma 250223? I base it on my own testing, and own testing is always subjective. It would be a great help to me if people would say what is good and what is not so good. I see enough downloads, but none of these users are really helping me.

[sarona]

Charisma Gold is my favorite - followed closely by Saphir.

I am more interested in analysis and mate solving, so I prefer a slow search. Both engines are in the top ten in my EloStat cbh database (3975WX/RTX 3090Ti - 60 threads - Peter Martan's 128 position Test Suite, 15"/pos).

I have not yet had the time to test your newer Leptir engines.

[benkrik]

Can you tell me what CPU you are using? I would also be interested to know which engine do you think is better and which is worse? Would a Charisma version that searches slower be better than Charisma 250223?

Eduard, I mostly use Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz with 31.9 GB Memory (8 logical cpu-s)
When I run tournaments/tests I use a single CPU per engine (no ponder). So my average speed for tournaments would be only 1000-1400 kns. For solving problems I use 4-8 cpus and get on average 3000-6000kns.

I have been running some not very serious tests (including 20,000 game (one second per move increments) tournaments for 50 Stockfish clones and few dev versions.) I use balanced Arena 3.5.1 book and realize that some of the results might be randomly affected by a less strong opening. So far after 12,500 games (500 games per engine or so) Charisma Gold - is sharing 1st/2nd place (out of 50!) with Stockfish_23012306. Charisma Light 2 - is the 5th, and Charisma 250223 and Charisma Blitz are sharing 10-12th place with EMAN. The difference between the 1st and 12th place is probably just couple ELOs but all those engines might be up to 5-10 ELO stronger than original 15.1. I realize that not that many games have been played and my tests are not too precise, but so far there is no indication that Charisma Gold in blitz performs worse than 250223 or worse than any Stockfish/clone with a faster search. Other Charismas that I test (Charisma Diamond, Charisma, Charisma Saphir, Charisma 240223 ) - in blitz perform plus-minus the same as 15.1, both Leptir versions for now are slightly better.

As far as a longer time control solving problems go Leptir (both Classic and 280223) is the best, but Charisma Gold and Charisma 250223 are just a notch behind.

Overall (as a universal engine for both blitz and problems) Charisma Gold is my ultimate favorite (followed closely by Charisma 250223 and Charisma Light 2).

[DeletedUser190323]

Thanks!

When I started Charisma, I wanted to release both fast and slow-search versions of it. That's why there were these terms like Blitz or Gold, etc.

Now I've decided to create charisma with just a fast search. Leptir was a private version until a few days ago, and has always had a slow search. Now that you can also download Leptir, I saw no need to continue offering Charisma Gold. However, Charisma Gold has a faster search than Leptir but slower than Charisma 250223.

I'm happy to make Charisma Gold available for download again if desired?

Listing by search speed:
Charisma 240223
Charisma 250223
Charisma Gold
Leptir 280223/Leptir Classic

[DeletedUser190323]

Charisma Gold can be downloaded again from my homepage.

A word about testsuites. I love testsuites. However, position tests have one disadvantage: there is always a special solution. And: 90% of position tests consist of tactical positions. A tactically strong engine will solve more positions. In practice, however, such positions are very rare. Rather, it depends on fine moves to get a lasting advantage. Usually there are even several such moves in sequence. If a tactical monster doesn't do that as well, then his tactics won't do him any good. That's why a tactically weaker engine can be better in practice! The best is when an engine is good both, tactically and positionally. And how do you determine that? You can't tell by testpositions alone. You have to play games. That's why I do both. I test with position tests, but I also play a lot of games on the servers at the same time.

[benkrik]

Thank you for your work! I will test more and let you know, but definitely those are great engines.

[Anton101]

Joaquin26 or Benkrik, someone have sent me screenshots of how you makes fun of Eduard's engines in Outskirts forum, you think you are working with CFish a very outdated engine a real crap actually, you and the Outskirts forum can go to shit. I was expelled from the forum by the imbecil of Kerveros, for matters that have nothing to do with the forum, I also see that Deeds was also expelled... In that stupid forum maybe five or six share material, the rest suck information.

MZ should be ashamed of the shitty nest it has as a forum, actually MZ no longer influences anything, its latest engine is as bad as rotten bread. And for Joaquin 26 do something your CFish is garbage ...https://github.com/Joachim26/CFishNN/releases A GitHub with engines that nobody cares about.