coloin wrote:Your initial post was showing patterns with clues in B1245 which have puzzles. any less clues in B1245 will be a maximal pattern which doest have a puzzle.
Not quite true. Removing any clue (clues) in B1245 area will give us invalid pattern, not having valid puzzles, but that pattern may not be maximal.
For example, pattern
- Code: Select all
+-----+-----+-----+
|. . .|. . .|x x x|
|. . .|. . .|x x x|
|. . .|. . .|x x x|
+-----+-----+-----+
|. . .|. . .|x x x|
|. . .|. . .|x x x|
|. . x|. . .|x x x|
+-----+-----+-----+
|x x x|x x x|x x x|
|x x x|x x x|x x x|
|x x x|x x x|x x x|
+-----+-----+-----+
is invalid, but not maximal, because we can add clue in r6c6 cell and get another invalid pattern.
- Code: Select all
+-----+-----+-----+
|. . .|. . .|x x x|
|. . .|. . .|x x x|
|. . .|. . .|x x x|
+-----+-----+-----+
|. . .|. . .|x x x|
|. . .|. . .|x x x|
|. . x|. . x|x x x|
+-----+-----+-----+
|x x x|x x x|x x x|
|x x x|x x x|x x x|
|x x x|x x x|x x x|
+-----+-----+-----+
But maximal patterns are not expandable - adding 1 clue always convert maximal invalid pattern to valid pattern.
coloin wrote:... since 1 clue is needed in B124 we therefore need 2 clues in B1245 ...
2 clues in B1245 not alway guarantee validity of the pattern (see example above).
coloin wrote:Minimal patterns that have puzzles is now a new concept to me.... and indeed there is work that could be done.
Minimal patterns were introduced in the thread Minimal patterns. They have interesting property - to produce minimal puzzles only!
When we are talking about 18-clue patterns, the good examples of minimal 18-clue patterns are 18-clue vertically symmetrical patterns. (All 121 possible 18-clue vertically symmetrical patterns were found years ago.)
I should say, why I switched my attention from maximal patterns to minimal patterns. Both catalogues (of maximal/minimal patterns) are sufficient to describe classes of patterns. For example, all 6016 fully symmetrical patterns can be separated to valid/invalid patterns either by catalogue of maximal patterns (25 patterns) or by catalogue of minimal patterns (49 patterns). Filtering out invalid patterns can be done either by maximal patterns list (for example, by "40 maximal patterns list"), or by minimal patterns list, because any valid pattern must contain some minimal patterns as subsets. But I think catalogue of possible patterns is more suitable to use than catalogue of impossible patterns.
Of course, three 4-boxes partially minimal patterns (Q1, Q2 and Q3 published above in this thread) have very limited practical use. But I am planning to find 2-band (6-boxes) partially minimal patterns. 2-band patterns will be much more useful for filtering, etc.
Serg