On Thu, 25 Aug 2005, Simon Tatham wrote: [...] > Hmm. I suppose it could be seen that way, although that's not how I > saw it. I saw it as a special case of a rather different pattern, > which I currently describe as `mutual neighbour analysis'. Rather > than being a chain of arbitrary length, this pattern involves > finding two non-adjacent squares and a bunch of their mutual > neighbours, and observing that placing a particular number in one of > the end squares forces all the neighbours to take values which cause > a contradiction in the other end square. [...] > If the 45 square has a 4 in it, [...] a problem. Hence, the 45 square > can't be a 4, so must be a 5.
Tell me again how this is distinct from "backtracking"? Basically, your method is
1. Fill in a value for cell X. 2. Fill in a value for cells Y, Z,... dependent on X. 3. If at any point you find a contradiction, return to step 1.
That looks a lot like backtracking to me!
(Not that backtracking is bad; I just don't see why this particular class of backtracking applications is interesting.)
-Arthur, Sudoku is crosswording for the illiterate