Leptir
-
- Posts: 202
- Joined: Thu Jul 06, 2023 11:47 am
- Location: Germany
- Has thanked: 7 times
- Been thanked: 293 times
- Contact:
Leptir
New engine by Solista: Leptir 090124. This is my first engine with two networks. Stockfish developers have now implemented two networks. Leptir is now based on this code. All Stockfish updates up to 080124 have been implemented. Further: I used the CorChess code by I. Ivec too. Other features include the Random Op. MultiPV code by Marco Zerbinati as well as MinimumThinkingTime and Slow Mover. This engine is ideal for analysis! With default settings (Random Op. MultiPV=2): In my ENET-2023 position test, the engine managed 107 out of 110 solutions at 60s on 6 cores. For quick analyzes the value should be set to 1 and for extremely long analyzes to the value 5. Engine Settings:
For gaming: For Lichess I recommend setting the slow mover to 115 when playing with Popyglot Book at Blitz 3 to 5 minutes. For example in BotLi
https://github.com/Torom/BotLi settings. For PlayChess with CTG book I recommend a value of 95. Some testers believe that every engine has to play well in every test. These are either people who can't do anything, or they do it on purpose to harm others. This engine is primarily made for analysis. If you want to play with it, you should be able to use at least 6 cores. If you don't have that, you should use Stockfish dev.
Downloads Win 64 Bit embedded networks (avx2, avx512, bmi2, sse41, modern, source code):
https://mega.nz/file/iUUWyAob#ItqwTkJgt ... BhzWAnQTJ4
https://pixeldrain.com/u/pCJHDkJ6
Have a nice day!
https://solistachess.jimdosite.com/solista-news/
Eduard
For gaming: For Lichess I recommend setting the slow mover to 115 when playing with Popyglot Book at Blitz 3 to 5 minutes. For example in BotLi
https://github.com/Torom/BotLi settings. For PlayChess with CTG book I recommend a value of 95. Some testers believe that every engine has to play well in every test. These are either people who can't do anything, or they do it on purpose to harm others. This engine is primarily made for analysis. If you want to play with it, you should be able to use at least 6 cores. If you don't have that, you should use Stockfish dev.
Downloads Win 64 Bit embedded networks (avx2, avx512, bmi2, sse41, modern, source code):
https://mega.nz/file/iUUWyAob#ItqwTkJgt ... BhzWAnQTJ4
https://pixeldrain.com/u/pCJHDkJ6
Have a nice day!
https://solistachess.jimdosite.com/solista-news/
Eduard
-
- Posts: 106
- Joined: Thu Jul 06, 2023 12:30 pm
- Has thanked: 1056 times
- Been thanked: 245 times
Re: Leptir
Thank you very much Eduard !
I wish you good health and happiness to yourself and your family.
I wish you good health and happiness to yourself and your family.
-
- Posts: 202
- Joined: Thu Jul 06, 2023 11:47 am
- Location: Germany
- Has thanked: 7 times
- Been thanked: 293 times
- Contact:
Re: Leptir
11. Jan 2024: Leptir 100124. Engine Update! I have now modified the engine and increased the threshold for handover to the small network. This means that the two positions --> read more from here: viewtopic.php?p=2141#p2141 are solved immediately. The engine solves 108 of 110 positions in my ENET-2023 position test on 6 cores with 60s/pos!!
Download Win 64 Bit embedded networks (avx2, avx512, bmi2, sse41, modern, source code):
https://pixeldrain.com/u/sMqwgyYA
More here:
https://solistachess.jimdosite.com/solista-news/
Eduard
Download Win 64 Bit embedded networks (avx2, avx512, bmi2, sse41, modern, source code):
https://pixeldrain.com/u/sMqwgyYA
More here:
https://solistachess.jimdosite.com/solista-news/
Eduard
-
- Posts: 106
- Joined: Thu Jul 06, 2023 12:30 pm
- Has thanked: 1056 times
- Been thanked: 245 times
Re: Leptir
Hi Eduard!
I congratulate you for the modification in evaluate.cpp without a doubt your engine find better movements, I really like Leptir, but it must be clarified first of all, this code was not created by MZ, but rather by Vondele and David for a branch called drawMe.
(https://github.com/vondele/Stockfish/commits/drawMe/ ), which is designed to play suboptimal moves in the opening to get out of theory as fast as possible and make it harder for humans to draw against the engine.
So the credits are for Vondele and David.
Best regards;
I congratulate you for the modification in evaluate.cpp without a doubt your engine find better movements, I really like Leptir, but it must be clarified first of all, this code was not created by MZ, but rather by Vondele and David for a branch called drawMe.
(https://github.com/vondele/Stockfish/commits/drawMe/ ), which is designed to play suboptimal moves in the opening to get out of theory as fast as possible and make it harder for humans to draw against the engine.
So the credits are for Vondele and David.
Best regards;
-
- Posts: 202
- Joined: Thu Jul 06, 2023 11:47 am
- Location: Germany
- Has thanked: 7 times
- Been thanked: 293 times
- Contact:
Re: Leptir
Thanks Anton for this correction. I really thought it was from MZ. That's why I put his name as the author.
-
- Posts: 202
- Joined: Thu Jul 06, 2023 11:47 am
- Location: Germany
- Has thanked: 7 times
- Been thanked: 293 times
- Contact:
Re: Leptir
Leptir Ultra by Solista: Try another way! Special version for ultra long analysis and correspondence chess. Download (avx2, avx512, bmi2, sse41, modern, source code): https://pixeldrain.com/u/rjxd5DLq
I tested the engine carefully, with analysis of course. I played Blitz on PlayChess, and even Bullet 1+0 offline against Eman 9.80 (with Ponder on and 5000 kns / engine).
Crazy result 1min + 0s:
I tested the engine carefully, with analysis of course. I played Blitz on PlayChess, and even Bullet 1+0 offline against Eman 9.80 (with Ponder on and 5000 kns / engine).
Crazy result 1min + 0s:
-
- Posts: 106
- Joined: Thu Jul 06, 2023 12:30 pm
- Has thanked: 1056 times
- Been thanked: 245 times