I have been looking for quite some time without any luck.
Can anyone recommend an engine which is both *open source* and UCI compatible that has variable strength that dips into the beginner ranges? I'm talking about 800-1400 elo, preferably with a high degree of granularity. Ideally an engine that is fun to play against.
All of the engines I have looked at so far have a minimum of about 1600elo (including stockfish), which are not helpful for my project.
I copied this question over to the discord channel and got some feedback, so it's not urgent.
Stockfish lowest is 1350. Also you can try to play a non-rated game against my bot, it will adapt to the opponent's strength.
You can use Leela Chess Zero with a weak network weights file (like me) and/or configure parameters to make it artificially weaker (e.g. having one node per move like @LeelaInstinct or @bomtot). If you're feeling adventurous, you can also train your own network weights.
I also like the PyChess engine. It has 20 difficulty levels and it feels like playing a human.
Lastly, if you have a Linux machine, you can use cpulimit to artificially weaken your engine by limiting its CPU usage. A nice side effect of this is being able to play many games at once without putting too much stress on the CPU.
Thanks for responding @high_five and @LeelaWeak I'll look into each of those options.
I noticed stockfish went down to 1350, but I was hoping for something that could go down even lower, and perhaps play more human-like moves than a tuned down stockfish which can play somewhat erratically.
I really like the idea with training leela to be weak, which is actually very much in line with the end goal of my project. Once I get some groundwork done I'll definitely go that direction. In the mean time, all those other tips will probably fit the bill.
Look at the end of the engines rating ccrl.chessdom.com/ccrl/404/rating_list_all.html
There are some weak engines
This topic has been archived and can no longer be replied to.