It is related to the automorphs of course !

gsfs example gives a subpuzzle with

- Code: Select all
`+---+---+---+`

|12.|45.|7.9|

|456|..9|.23|

|7..|...|.5.|

+---+---+---+

|231|.6.|.9.|

|.6.|...|2..|

|8..|2..|...|

+---+---+---+

|31.|...|9.8|

|64.|9.8|...|

|...|.1.|6..|

+---+---+---+ 92 grid solutions.

The mc grid doent have valid a puzzle with a {+2}.

The other grid has one.

The reduced total amount of "different" puzzles inherent in the MC grid is the reason the two grids dont get near sharing a pseudopuzzle.

Red Eds distance distribution stat. for the MC grid, for >500,000 random grids. Average [mode] distance is 43.

- Code: Select all
` 22 : 1`

23 : 1

25 : 1

26 : 4

27 : 5

28 : 15

29 : 29

30 : 49

31 : 124

32 : 205

33 : 446

34 : 856

35 : 1721

36 : 3354

37 : 6598

38 : 12496

39 : 22614

40 : 38944

41 : 61708

42 : 87889

43 : 105053

44 : 93265

45 : 51271

46 : 12600

47 : 747

48 : 4

I think this will be born out if we can [slowly] compute a brief distance distribution stat for a random grid . [Average [mode] distance ~32 ?]

And [perhaps not quite so slowly] for this grid

- Code: Select all
`174852693256931487389764521963417258712589346845623179528396714431278965697145832`

which has automorphism-72.

C