We need a new interesting chess Engine project... stockfish and komodo makes no fun
4 posters
Author
Message
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: We need a new interesting chess Engine project... stockfish and komodo makes no fun Thu Apr 01, 2021 1:59 pm
IMO without a new baby we can tune and trigger the next years computerchess is getting boring.
Stockfish or komodo, even with nnue have no chance against LC0 on a normal GPU (talking about my old nvidia 2070).
Even with 8 cores LC0 with j94-130 net is ahead.
Ok i could now increase with more cores. 12 or 16.
But i would prefer testing a new engine that plays different chess. But all the programmers do the usual way, so nothing new under the sun.
Admin Admin
Posts : 2608 Join date : 2020-11-17 Location : Netherlands
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun Tue Apr 06, 2021 8:09 pm
Lc0 plays different chess.
Mclane
Posts : 3022 Join date : 2020-11-17 Age : 57 Location : United States of Europe, Germany, Ruhr area
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun Tue Apr 06, 2021 11:51 pm
Admin wrote:
Lc0 plays different chess.
In this dry and desert computerchess age, i would have given up without LC0.
Completely spammed with rybka, houdini, stockfish and komodo, where you often cant say where which engine begins and which engine ends, LC0 is an alien.
It plays like a human beeing with scores like a human beeing.
TheSelfImprover
Posts : 3112 Join date : 2020-11-18
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun Wed Apr 07, 2021 11:09 am
Mclane wrote:
But i would prefer testing a new engine that plays different chess.
You can train a net to prefer "fun" to "winning".
The clever part will be writing code to score a position for fun. Everything else you can read how to do.
Actually, if you can write code that scores a position for how much fun it is, maybe you won't even need to train a net.
Admin Admin
Posts : 2608 Join date : 2020-11-17 Location : Netherlands
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun Wed Apr 07, 2021 9:34 pm
Mclane wrote:
Admin wrote:
Lc0 plays different chess.
In this dry and desert computerchess age, i would have given up without LC0.
Completely spammed with rybka, houdini, stockfish and komodo, where you often cant say where which engine begins and which engine ends, LC0 is an alien.
It plays like a human beeing with scores like a human beeing.
The wait is for a GPU with 16 cores, I would drop SF immediately.
koolaidsa
Posts : 6 Join date : 2020-11-28
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun Thu Apr 15, 2021 7:40 am
im confused are stockfish and komodo not the better engines on a gpu? are they made to perform at the top only on cpus? lco isnt a top 10 engine on ccrl 40/15
Admin Admin
Posts : 2608 Join date : 2020-11-17 Location : Netherlands
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun Thu Apr 15, 2021 8:33 am
koolaidsa wrote:
im confused are stockfish and komodo not the better engines on a gpu? are they made to perform at the top only on cpus? lco isnt a top 10 engine on ccrl 40/15
Lc0 is not at the top of the 40/15 list because it runs on a CPU. But also look at the 40/2 list, here Lc0 runs on a fast GPU, the platform Lc0 is intended to run. Quite a difference. Hope this helps.
koolaidsa likes this post
Sponsored content
Subject: Re: We need a new interesting chess Engine project... stockfish and komodo makes no fun
We need a new interesting chess Engine project... stockfish and komodo makes no fun