>
> It does this by generating random legal moves. A string of legal moves, to
> the end, is one "playout."
>

OK, now I understand it generates a sequence of moves, all the way to the
game end; which means a playout typically contains 200 (from middle game) ~
300 (from opening) moves, and the so-called 'evaluation' becomes just
counting the stones.

Very interesting!

Then it doesn't need to know anything about those secondary, or 3rd-ary,
4th-ary ... concepts like 'eyes', 'connections', ...; it only need to know
the very basic game rules that any stone un-capturable is alive. Haha. "God
create integers, all the rest are just work of man" :-)

No wonder it plays so well at 9x9, because the max length of playout is only
81, it can 'see' what the board look like when the game ends. This is kind
of similar to the exhaustive search approach (depth-first in this case).

This also explains that when I read the games MoGo against GNUGo, toward the
end of the game, GNUGo would play PASS, but MoGo would continue to play at
some very uncommon positions that a normal player would never consider.



> Playing randomly like that shouldn't work, but when you play Mogo et al,
> you see that intelligent behaviour emerges.
>

Although interesting, I would hardly call that 'intelligence' :-)  And I
suspect how far it can achieve on 19x19 board (compare to human players of
course).
_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to