Hi what are the best LCZero settings or the most optimal to use for it. Is the default setting also good?
Greetings, Liem, and sorry for the delay in responding!
So, at least on NCM’s hardware, I think that using the latest T40 network on the GPU servers is going to be your best bet. Some of our testing has suggested that LcZero is going to perform better (relative to Stockfish) at longer think times, e.g., > 10s/move. For shorter think times, Stockfish may have the advantage. Our testing is a little bit sloppy though (it’s a work in progress) so I can’t say that with 100% confidence:
Thank you for this piece of information and also for your hardwork.