but i see [ as usual] despite the great advances the enormity of the task still remains
The 665/665 divide and conquer was excellent
Although its not clear how that is performed !!! [ ? you set a limit to the clues per band]
Might it be possible to think of a few more partitions which might speed it up ?
There is another division in the distribution - i toyed with earlier in thread ...
Every 17 pattern has to have at least 10 clues in one of the representative B12347 boxes [ 7 clues in B5689] - see below
The majority of 17 puzzles have the 10 plus 7 in one of the 9 representatives
some don't have a 10 plus 7 [but these will have a 11 plus 6 or 12 plus 5]
- Code: Select all
+---+---+---+ +--+--+--+
|1..|2..|...| |.7|.9|.9|
|.3.|...|4..| +--+--+--+
|...|...|...| |10|.9|10|
+---+---+---+ +--+--+--+
|...|.4.|.3.| |10|11|10|
|5..|...|..1| +--+--+--+ [85 clues / 5 = 17][ adding up the 9 ways to have a representative B12347 = 5 whole grids]
|...|89.|...|
+---+---+---+
|..4|...|86.|
|.7.|..1|...|
|...|5..|2..|
+---+---+---+ 7 clues in B12347
If there were a pattern without a B12347 representative more than 9 - it wouldn't have 17 clues
- Code: Select all
+--+--+--+
| 9| 9| 9|
+--+--+--+
| 9| 9| 9|
+--+--+--+
| 9| 9| 9|
+--+--+--+ 81/5 < 17
Maybe blue can incorporate ...?