PaulIQ164 wrote:The actual ins and outs of generating puzzles are, I think, not really supposed to be discussed here, since it's how Pappocom makes its money. Though it's largely irrelevant as I don't know the first thing about it anyway.
I think the confidential part of Pappocom's software is the RATING of puzzles.
Generating a valid puzzle is easy
- pick a random completed grid
- delete clues randomly (or systematically) testing at each stage that the resulting grid can ONLY be completed back to the original grid (ie has just one solution)
- optionally add clues to make it symmetric or make it easier
The tricky part is to then ensure that the resulting puzzle is of reasonable difficulty and reasonably interesting for solvers.
So as not to step on Pappocom's toes, I will not even speculate on how this might be done, except to note that one of the recent articles in either New Scientist or American Scientist (I forget which) said that "15 to 20 different factors were involved".
Gordon