Thank you for the Unix script! I will try it and see if the results
produce a distribution that's akin to the same distribution for optimal
solutions.
Patsolve is very good at finding short solutions in a very reasonable
amount of time, but they are not really optimal; at least in the latest
version of FcPro.
I recently wrote an atomic-move BFS solver that produces (near-)optimal
solutions; i.e., optimal much of the time. Unfortunately, no matter how
difficult the deal, it takes around three minutes to reach a solution on
my P-III/450 PC. That's why I'm looking for a way to only have it tackle
the more difficult deals.
FYI: Here's an instance where PatSolve takes 86 atomic-moves (FcPro
Mode=T) when there exists a 41 atomic-move solution
#290086 Attempt: 1 NumFcs=4 (BFS Solver) 86/41 moves
72 7d 74 54 5c 7b 74 7a 57 b7
c7 57 87 37 57 35 2c 6b 65 25
a6 c5 3c 31 3a 35 85 28 18 13
25 3h 41 6d 63 d3 4d 42 46 4c
82
Again, thanks for your response and suggestions!!!
Shlomi Fish wrote:
> On Tue, 2 Sep 2003, Danny A Jones wrote:
>
> > My wording lacked clarity. I'm sorry.
> >
> > I should have probably defined "difficult" as requiring 20% more moves
> > for a solution than a median/"typical" problem.
> >
> > For example, in problems 1-500, some deals require only 28 moves to
> > solve; whereas others require 58 moves. The median is 41 moves. Problems
> > requiring 49 moves or more are what interest me. (These numbers are
> > subjective, but should still give reasonable insight into my
> objectives.)
> >
>
> Do you mean the number of moves that the solver emits as a solution? Or
> the minimal number of moves required to solve the board?
>
> If the former, then the following script should do:
>
> #!/bin/sh
> first_game=32001
> last_game=32100
> moves_num=150
> for I in `seq $first_game $last_game` ; do
> m="`pi-make-microsoft-freecell-board $I |
> fc-solve -l good-intentions |
> grep ====== |
> wc -l`"
>
> if test $m -gt $moves_num ; then
> echo $I
> fi
> done
>
> It's a UNIX script which has to be adapted to Windows NT.
>
> If you wish to find the length of the optimal solution, then Freecell
> Solver can't help you much because it cannot do this yet. Patsolve can,
> however, only it uses atomic (one-card) moves, which may or may not be
> what you are looking for.
>
> Regards,
>
> Shlomi Fish
>
> > WKRfresno_at_aol.com wrote: <snip>
Received on Wed Sep 03 2003 - 16:22:59 IDT