🚀 go-pugleaf

Joost de Heer wrote:

> Not always a complete solution? Could you give examples?

  I had one, once, but I'm afraid I've lost track of it.  It was
in a dual solution  as far as I can recall (I believe it was one of
the problems in English Chess Problems), and I recall reporting it --
probably to Norbert Geissler, one of the Popeye implementors.  I don't
have his reply saved, unfortunately. 

  But in a later exchange he wrote:

	It's right, that in a major solving tourney a problem was tested
	by the judge with Popeye and it found three variations.
	But the solvers in the tourney found four variations, which confused
	the director (he had computer tested it, right ?).
	  Since Popeye doesn't consider defenses which differ only at move three or
	later to the threat they should defend, this happened.
	  You are wrong, that we (the developers) think, that this behaviour is 
	definitely correct.
	  We know, that inplementing this "feature" costs a lot of
	developing time and since such problems are rare, this is 
	not on the top wish list to implement.

That was in November, 1999, so things may have changed since then.

  Actually, I've seen something vaguely similar in Problemiste, although
it was related to short notation ambiguity instead. The position is:

        Andrews:  Q3r3/3q3b/1p6/2p1RpB1/2pk1rpK/6Pp/BP1pR3/8    #3

  The main play of this problem is:  

        1. R5e5 d1=Q  2. Qa4  and then  ... Q7xa4  3. Rd6#
                                or      ... Q1xa4  3. Rd2#

  With long notation selected, both these variants are displayed (in
long notation, of course), but if I have shortened notationed selected,
I get only one of them:
                                        ... Qxa4   3. Rd6#

  I don't expect any problem-solving software to be entirely
without errors. The question is just what types of errors you
can live with. 

-- 
Anders Thulin     ath@algonet.se     http://www.algonet.se/~ath