From: Chris Fant <[EMAIL PROTECTED]>
> Such guidance has to be fairly subtle, however; it often must take
the form
> of "if he plays here, do this, if there, do that."
> Doesn't the search tree provide such functionality?
It does indeed - but there are often forced sequences which can be predicted
well in advance. Skillful use of such information might lead to a vastly
smaller tree, allow faster and more accurate evaluation, and more time to
devote to other branches.
Suppose both sides are evenly matched, but a group has four liberties in a row,
and no other eyespace. If one's opponent plays in the middle of such an
eyespace, one must connect - unless one has an even larger threat. There are
five-space shapes with a vulnerable point. One should not depend upon
converting a "bulky five" into two eyes; if one can induce one's opponent to
make such a shape, one may win the game. Qualifiers are important - if there is
a sure second eye, the play can and should be ignored; it would be a wasted
play. There are many life-and-death situations which can be similarly analyzed
at a high level; once this analysis is done, there is no need to endlessly
repeat it in many branches of the search tree. Giving such plays a much higher
probability might make good sense.
Going back to what Remi said earlier: the tricky thing about nakade is that,
when a single large eyespace can be destroyed, nakade is very very good; it is
like a bomb detonating the opponent's prospects. When the group has a second
eye elsewhere, the exact same move should be strongly discouraged. Sometimes
one uses such a placement to convert a space into a single eye, then chases the
group in such a way as to build profit while the opposing group makes a bare
minimum of territory, or is killed. Pros are very good at spotting such plays,
and knowing which are good and which are not.
____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page.
http://www.yahoo.com/r/hs_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/
