A van Kessel wrote:
01010101010101010101010101010101
00000000000000001111111111111111
11111111111111110000000000000000
IMHO they are exactly the same and should be as such.
At the start of every simulation (before a 0 or 1 is reported)
, the situation is (should be) exactly the same.
So there is no difference with flipping a coin repeatedly.
If it occurs more often than expected by coincidence, there
is som hidden state in the playout board code (such as a over-
active result hash/cacheing, or heuristics with memory)
You seem confused.
UCT itself is not static or stateless. Such sequences can and will
result from UCT discovering that one move leads to a won or lost
variation. It's not only in the playout leaves that you have to consider
this but also in the UCT tree.
--
GCP
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
