[Computer-go] alarming UCT behavior

[Dave Dyer]

>But you are stopping right at the atari, and then pile on playouts that make it seem work...

Yes, something like that may be the situation that turns the result. 

Suppose the tree stops at a point where there are two moves, a
blunder that leads to a quick end, and another move which leads
to an inconclusive continuation.   

If the tree can't grow, then random playouts will take the "blunder"
branch 50% of time, whereas if the tree were allowed to grow the
non-blunder branch would quickly dominate.

Overall, once the tree is frozen, some percentage of the leaf
nodes will be permanently giving bad feedback.