On discord, ryokousha has posted a couple more spanning more boxes:
- Code: Select all
* . . | . * . | . * .
. * . | . * . | . * .
. . . | . . . | . . .
------+-------+------
. . . | . . . | . . .
* * . | . * . | . . .
. . . | . . . | . . .
------+-------+------
. . . | . . . | . * .
* * . | . . . | * . .
. . . | . . . | . . .
- Code: Select all
* . . | . * . | . . .
. * . | . * . | . * .
. . . | . . . | . . .
------+-------+------
. . . | . . . | . . .
* * . | . * . | . . .
. . . | . . . | . . .
------+-------+------
. . . | . . . | . . .
. * . | . . . | * . *
. . . | . . . | . * .
We're looking at generation possibilities, limiting by number of cells rather than boxes.
Also wondering whether it makes sense to start classifying these by number of houses with full triplets. For example, a common structure in the 10 cell patterns is 2 rows, 2 columns, and 3 boxes - perhaps we can say something about these as a class.
In the first, we can see that r1c2 is r4c1 (b1c1) is r5c5 (r4b5) and this breaks b2. In the second, we have the same equivalent cells, just using a different row/box combination (b4r5 instead of r4b5) for the last. Obviously the two patterns are different in the isomorph sense, but logically they are very similar.
