Re: computer-go: About Viking

[Dave Dyer <ddyer@xxxxxxxxxxx>]

>>Either a chain of stones is connected or it is not, and
>>the program should know exactly why.

One fundamental problem is that you're asking the wrong question. The
question is not "are these groups connetable", the question is, "are these
groups connectable while improving my overall position". It's pretty 
common to have groups which CAN be connected, but at an unacceptable cost.

For example it's pretty common to find two groups, either of which
can be saved.  So if you ask "can this group be saved" the answer 
is always yes, but when you try to execute the plan one of the groups
dies, and sometimes both of them.  Or perhaps you can connect 1 stone 
at the cost of allowing a 20 point invasion (where nothing dies).

Another problem is that its completely trivial to reach an incrementally
expanded search tree that is hopelessly too large to solve.  The really
annoying aspect of this is that very frequently the expanded search would
be irrelevant, if you could actually conduct it.  Either a different
node ordering would allow the branch to be removed entirely, or the
result of the search is that the "opportunity" the search was investigating
didn't pan out.