"Apache" <apa### [at] hotmailcom> wrote in message
news:4057361a@news.povray.org...
> > Hmmm, sounds like genetic algorithms would be better in this case.
> Not if we want *prove* that there aren't any better solutions to the
puzzle.
> Setting up a small script that will find a *very good* solution isn't a
big
> deal, but getting the set of best solutions is something else.

What about calculating a probability weight for each board assuming 86 is
the shortest path?  For example at 0 depth, the pegs are on opposite sides
so that's the aboluste worst and the weight is 0%.  The closer you get to 86
moves you calculate the percentage of pegs on their proper sides.  So if you
are 10 moves deep, and your weight isn't close to 12% (10/86) you stop
searching that branch.  I'm not sure how many paths this would cut out, but
it might cut out some.  Of course I'm assuming the best solution doesn't
involve too much backward travel.

AW

Post a reply to this message