Natch switch (was Re: [Retros] Two little SPG-challenges)

[Pascal Wassong]

Hello,

    Mario wrote:
    M> I would wish the programmers of Natch & Co. would provide a
    M> switch, which forces the detection of shorter solutions.  Is
    M> there any specific reason not to do so?

I don't think that such a switch would be usefull.  For most SPG's in
N moves, if they have solutions in fewer moves, there would be many
cooks in N moves.

There are some very rare exceptions.  But I think that any composer or
solver would recognize these problems.  They involve generally some
tempo play.  It could then be usefull to test them for shorter
solutions.


There is another problem if Natch would search for shorter solution :
it could find many of them.  Let's take for example the problem by
Gerd Wilts from Die Schwalbe 1994, recently awarded a 3rd Honorable
Mention (PDB entry : P0002233).

+---+---+---+---+---+---+---+---+

|*R |*N |   |*Q |   |*B |   |*R |

+---+---+---+---+---+---+---+---+

|*P |*P |*P |   |*P |*B |*P |*P |

+---+---+---+---+---+---+---+---+

|   | . |   | . |*P |*P |   | . |

+---+---+---+---+---+---+---+---+

| . |*K | . |   | . |   | . |*N |

+---+---+---+---+---+---+---+---+

|   | . |   | . |   | . |   | P |

+---+---+---+---+---+---+---+---+

| . |   | . |   | P |   | . |   |

+---+---+---+---+---+---+---+---+

| P | P | P | P | K | P | P | . |

+---+---+---+---+---+---+---+---+

| R | N | B | N | Q | B | . |   |

+---+---+---+---+---+---+---+---+
rn1q1b1r/ppp1pbpp/4pp2/1k5n/7P/4P3/PPPPKPP1/RNBNQB2
a) SPG in 10.0
b) PG in 12.5

When solving b), you would get the unique solution of a), but also
many solutions in 12.0 moves.  You would then have to check all the
solutions to verify that there is only one solution in 12.5 moves.
(To help, Natch could sum up how much solutions were found for each
length.)



    Noam wrote:
    N> I don't have any first-hand experience with Natch, but I
    N> imagine that for most SPG's in N, once Natch has solved it in N
    N> moves it would still take only a fraction of the time to check
    N> for solutions in N-0.5 or N-1.0, and a tiny fraction for even
    N> shorter solutions.  For all I know, it might not even save that
    N> much time for this search to be done together with the
    N> full-length solution.

I think that finding all solutions, even shorter ones, in one time
would take much less time than checking it in several steps.  But this
difference would probably not be of such importance to a human, as
Noam points out.  An exception may be for problems from the
"Slaughterhouse", where Natch is quite weak.  (But in this case it is
probably better to use a program complementary to Natch : Popeye.)


So, I don't think I will include such an option in Natch.  But, as I
always tell to people asking for improvments in Natch, sources are
publicly available, and anybody is welcome to modify them to suit
their needs.


Pascal