[computer-go] On expanding the UCT tree

[Erik van der Werf]

On 5/7/07, Peter Drake <drake at lclark.edu> wrote:
> Iterating through the children should not be a significant time hit,
> because (a) UCT trees tend to be quite shallow, rarely more than 5
> moves deep, and (b) the vast majority of nodes are leaves.

Right, but IMO the real reason why you want to examine the children is
transpositions. If you don't use a transposition table it should be
trivial to correct for the fact that the parent was sampled once more
as a leaf.

Erik