In Matthew Sadler's book, "The Silicon Road to Chess Improvement", Sadler uses Leela Zero for training with Leela configured to "one node" search (ply). It is a good way to have Leela "plays like a positional genius with flawed tactics".
How do I configure Leela this way? I have Fritz 18, installed Leela, go to parameters, but I can't find this option.
A simple way to play against LC0 with no search is to challenge one of the lichess bots that do this. Lazybot, for instance, has a fairly current net and plays quite well for something that does no search.
In a local chess GUI, I think setting thinking time to zero should also make LC0 play pure policy.
instructions on how to do that are given in the supplementary material that's free with the book: https://www.newinchess.com/engine-training That material describes how to use a really nice tool called Nibbler to play games at 1-node. Nibbler is available here https://github.com/rooklift/nibbler and gives all sorts of additional insights into what neural net-based engines (like Leela but also Stockfish now as well) are thinking. https://youtu.be/kfsBj6CYNis and https://youtu.be/s7CfGpNSalA illustrate how to use it as well.
