[Retros] Programs for Solving Retro Problems

Mario Richter mri_two at t-online.de
Sun Aug 24 14:16:17 EDT 2008


Alain> What i would really look forward is not a program that
Alain> can handle retro analysis, only a program that allows
Alain> the user to unplay a position.

'Retractor' has an edit history, that can be (mis-)used for
that purpose ...

mario> [...]
mario>> -- most severe (and at least to some extent a k.o. criteria):
mario>> some *legal* positions are judged by 'Retractor' as *illegal*.

>

Theodore> Do you have an example of this? If not, could you let me know next time
Theodore> you come across such a position? As designed, this is not supposed to
Theodore> happen.

A double-check, which can be explained by a capturing
pawn that promotes.
This also illustrates the big dilemma for all programmers
of RA-programs: on the one hand, rules should be as general
as possible, on the other hand, one has to be very careful,
that rule doesn't match situations, where its application
leads to wrong results.
Here the rule seems to be something like "If a king is in check
by two enemy pieces, and none of this enemy pieces can interfere
the checking line of the other piece, than the position is illegal."
This works in most positions, but forgets the case, that one
of the enemy pieces may change his shape while retracting -
by unpromotion.

e.g. 5QN1/8/5k2/8/8/8/8/7K (w: Qf8 Ng8 Kh1 b: Kf6)
or 5R1B/8/5k2/8/8/8/8/7K (w: Rf8 Bh8 Kh1 b: Kf6)

'Retractor' rejects the above positions with the message:
"Error: This position places your king in an illegal
double check."



Regards,

mario





More information about the Retros mailing list