Deriving pUCT from first principles
What do you do when there are too many possibilities?
Demystifying multi-layer inference.