...you already can do much better than just selecting purely based on visits by incorporating Q in a conservative way, such as via Leela Zero's LCB method (which KataGo also uses). Also LCB helps a lot at low visits too.
I briefly researched this LCB method, and found some leela-zero discussiones:
In an email to me, David Wu wrote:
I briefly researched this LCB method, and found some leela-zero discussiones:
https://github.com/leela-zero/leela-zero/issues/860 https://github.com/leela-zero/leela-zero/pull/883
Read through those links and implement the LCB method.