[computer-go] On expanding the UCT tree

[Jason House]

On 5/7/07, Peter Drake <drake at lclark.edu> wrote:

> When I
> look through the children of A and count a total of one playout, it
> seems natural that I should update the playout count for A:



I would normally try to save all simulation results, but I can certainly see
the value of forcing invariants that are easy to check (for debugging)

I think the discussion needs to be about both the number of playouts and the
number of wins (losses).  By removing the phantom result from your tree, you
should end up with results that are only based off of searches at your
deepest search depths.  It may also help creation of some other kind of
probability of winning metric (if desired).

I haven't yet implemented my own UCT implementation, so I'm mostly just
basing this off of theory.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20070507/df352b2c/attachment.htm