[sillysaurusx]

Hmm. AlphaGo and AlphaZero were trained using self play. I don't think there was any precomputing, unless you count training as precomputation.

[EGreg]

Precomputation on a huge cluster for a long time