The New Sudoku Players' Forum

[daj95376]

Code: Select all

+-------+-------+-------+ # AI Etana (Snail) by ArtolI
| 1 . . | . . 7 | . 9 . | # extreme nightmare
| . 3 . | . 2 . | . . 8 |
| . . 9 | 6 . . | 5 . . |
+-------+-------+-------+
| . . 5 | 3 . . | 9 . . |
| . 1 . | . 8 . | . . 2 |
| 6 . . | . . 4 | . . . |
+-------+-------+-------+
| 3 . . | . . . | . 1 . |
| . 4 . | . . . | . . 7 |
| . . 7 | . . . | 3 . . |
+-------+-------+-------+


Code: Select all

+-------+-------+-------+ # AI Etana (Snail) by ArtolI -- reordered
| 1 . . | . . 7 | 9 . . | # extreme nightmare
| . 3 . | . 2 . | . 8 . |
| . . 9 | 6 . . | . . 5 | # asymmetry seems crucial to difficulty level
+-------+-------+-------+
| . . 7 | . . . | . . 3 |
| . 4 . | . . . | . 7 . |
| 3 . . | . . . | 1 . . |
+-------+-------+-------+
| 6 . . | . . 4 | . . . |
| . 1 . | . 8 . | . 2 . |
| . . 5 | 3 . . | . . 9 |
+-------+-------+-------+


Code: Select all

+-------+-------+-------+ # AI Etana (Snail) by ArtolI -- reordered, symmetrical
| 1 . . | . . 7 | 9 . . | # extreme
| . 3 . | . 2 . | . 8 . |
| . . 9 | 6 . . | . . 5 |
+-------+-------+-------+
| . . 7 | . . . | . . 3 |
| . 4 . | . . . | . 7 . |
| 3 . . | . . . | 1 . . |
+-------+-------+-------+
| 6 . . | . . 4 | 3 . . |
| . 1 . | . 8 . | . 2 . |
| . . 5 | 3 . . | . . 9 |
+-------+-------+-------+


[gsf]

Code: Select all

# asymmetry seems crucial to difficulty level


the second puzzle is symmetric on the antidiagonal
the first and second require X cycles (simple coloring) in the proposition constraints in my solver
haven't seen any like that yet
the third requires swordfish in the proposition constraints

[tarek]

it would be interesting to see what the ratings of all puzzles would be after you update your program gsf !

tarek

[gsf]

it would be interesting to see what the ratings of all puzzles would be after you update your program gsf !
tarek

do we have a comprehensive list of the hardest in one place?

I'll update in a few days
until then you can do it by the command line
for the puzzles that -qhardest says are not-solvable try:

Code: Select all

-B -q-G -Z'FNP(FNBT2H2)V(6)||FNP(FNBTHW)V(7)||FNP(FNBTHWX)V(8)||FNP(FNBTHWXY)V(9)'

with ' quoting appropriate for your shell/command-line interface

the three puzzles should rate 99920 99921 99668 respectively with the above -Z expression

the first two puzzles solve with proposition constraint expression 8

Code: Select all

FNP(FNBTHWX)V(8)

where V(8) is the label which shows up as V8 in the -f%Q output
this means that batched single candidate guessing requires X cycles (simple coloring)
the third puzzle solves with proposition constraint expression 6
this means that batched single candidate guessing requires naked/hidden pairs

[tarek]

thanx gsf,

I've just used the older version of the solver with the expressions you've mentioned to screen some of my puzzles & landed on a 99960 rating puzzle.....

Looking for backdoors also.... I found several with (backdoor size 2) under the V9 constraints.......From what you've mentioned in your site & in other threads.... these should be Rare !!!

tarek