Curiosities

For fans of Killer Sudoku, Samurai Sudoku and other variants

Curiosities

Postby urhegyi » Wed Feb 24, 2021 12:10 pm

What happens here with all those wings? Is this really a SE 8+ JSWX or is there another way of solution not supported by the program?
Code: Select all
#1/WX/B4,JS
7.......3....................61.92.....9.3.....32.48....................5.......6
112222255111222555111125555666649999666444999766444998777434888777333888773333388

Code: Select all
Eliminated candidates per Method and per Sudoku

Method   \  Sudoku |   SER |     1
                   |-------|------
Not counted elims  |     0 |   137
Naked Singles      |   0.1 |    58
Hidden Singles     |   0.2 |    27
Naked Pair    [2]  |     3 |     8
Naked Triple  [3]  |   3.6 |     2
Naked Quad    [4]  |     5 |     3
Hidden Quad   [5]  |   5.4 |    13
Hidden Triple [6]  |     4 |     1
Locked Singles[2]  |   2.8 |    68
Locked Singles[3]  |   2.9 |     4
Locked Singles[4]  |     3 |     7
XYZ Wing      [3]  |   4.4 |     1
VWXYZ Wing    [5]  |   6.3 |     1
UVWXYZ Wing   [6]  |   6.6 |     2
TUVWXYZ Wing  [7]  |   7.5 |     4
STUVWXYZ Wing [8+] |     8 |     1
XY-Wing       [3]  |   4.1 |     1
                   |-------|------
Eliminated Cand's  |   338 |   338
Sum(SER * Cand's)  | 420.6 | 420.6

Initial Candidates :   338
Maximum SER rating :     8 - Approach
Labour rating      : 420.6 - Experimental rating
urhegyi
 
Posts: 539
Joined: 13 April 2020

Re: Curiosities

Postby urhegyi » Wed Feb 24, 2021 12:22 pm

It seems when you limit the length of wings to max 4 (WXYZ-wing) it still solves with a lower rating:
Code: Select all
Eliminated candidates per Method and per Sudoku

Method   \  Sudoku |   SER |     1
                   |-------|------
Not counted elims  |     0 |   140
Naked Singles      |   0.1 |    55
Hidden Singles     |   0.2 |    33
Naked Pair    [2]  |     3 |     5
Naked Quad    [4]  |     5 |     3
Hidden Quad   [5]  |   5.4 |    13
Hidden Triple [6]  |     4 |     1
Locked Singles[2]  |   2.8 |    71
Locked Singles[3]  |   2.9 |     4
Locked Singles[4]  |     3 |     7
XYZ Wing      [3]  |   4.4 |     5
XY-Wing       [3]  |   4.1 |     1
                   |-------|------
Eliminated Cand's  |   338 |   338
Sum(SER * Cand's)  | 373.8 | 373.8

Initial Candidates :   338
Maximum SER rating :   5.4 - Approach
Labour rating      : 373.8 - Experimental rating
urhegyi
 
Posts: 539
Joined: 13 April 2020

Re: Curiosities

Postby urhegyi » Wed Feb 24, 2021 12:31 pm

When limiting subsets to length max 3 (no quads) we have a solution with locked singles[5]
Code: Select all
Eliminated candidates per Method and per Sudoku

Method   \  Sudoku |   SER |     1
                   |-------|------
Not counted elims  |     0 |   140
Naked Singles      |   0.1 |    54
Hidden Singles     |   0.2 |    36
Hidden Triple [6]  |     4 |     5
Locked Singles[2]  |   2.8 |    82
Locked Singles[3]  |   2.9 |     4
Locked Singles[4]  |     3 |    10
Locked Singles[5]  |   3.1 |     1
XYZ Wing      [3]  |   4.4 |     5
XY-Wing       [3]  |   4.1 |     1
                   |-------|------
Eliminated Cand's  |   338 |   338
Sum(SER * Cand's)  |   333 |   333

Initial Candidates :   338
Maximum SER rating :   4.4 - Approach
Labour rating      :   333 - Experimental rating
Last edited by urhegyi on Wed Feb 24, 2021 12:38 pm, edited 3 times in total.
urhegyi
 
Posts: 539
Joined: 13 April 2020

Re: Curiosities

Postby urhegyi » Wed Feb 24, 2021 12:33 pm

When further limiting subsets to length max 2(no quads, no triples) we have a solution with locked singles[5] and an additional WXYZ-wing.
Code: Select all
Eliminated candidates per Method and per Sudoku

Method   \  Sudoku |   SER |     1
                   |-------|------
Not counted elims  |     0 |   136
Naked Singles      |   0.1 |    56
Hidden Singles     |   0.2 |    33
Naked Single  [1]  |   2.5 |     1
Naked Pair    [2]  |     3 |     3
Hidden Pair   [7]  |   3.4 |     1
Locked Singles[2]  |   2.8 |    83
Locked Singles[3]  |   2.9 |     4
Locked Singles[4]  |     3 |    11
Locked Singles[5]  |   3.1 |     1
XYZ Wing      [3]  |   4.4 |     5
WXYZ Wing     [4]  |   5.5 |     2
XY-Wing       [3]  |   4.1 |     2
                   |-------|------
Eliminated Cand's  |   338 |   338
Sum(SER * Cand's)  | 348.4 | 348.4

Initial Candidates :   338
Maximum SER rating :   5.5 - Approach
Labour rating      : 348.4 - Experimental rating
urhegyi
 
Posts: 539
Joined: 13 April 2020

Re: Curiosities

Postby Hajime » Wed Feb 24, 2021 2:00 pm

SiSeSuSo does per full method a full sweep over the puzzle.
So all hidden/naked subsets before the next method
So all generalized intersections before the next method
So all wxyz-wings are searched before the next method
etc....
The ordering is NOT according to the SER rating per sub-method
User avatar
Hajime
 
Posts: 733
Joined: 20 April 2018
Location: Netherlands

Re: Curiosities

Postby urhegyi » Tue Mar 09, 2021 6:23 pm

When finding JSWX layouts you need a portion of luck too. While a few days ago I found one which resulted in a high number of 8 clues puzzles, today's one results only in very easy puzzles. I think it's layout dependent. Thisone is to symmetrical from design. Even when I generate with the forcing net option on the highest rated puzzles I can get are:
Code: Select all
#1/WX/B4,JS
....9.....1.....9....964......3.7...............8.6......135....5.....4.....5....
117777722111777222114476662114555622344555662334555699344486699333888999338888899
#1/WX/B4,JS
..1...5...1.....9.....6.......3.7...............8.6.......3.....5.....4...9...4..
117777722111777222114476662114555622344555662334555699344486699333888999338888899

Is it through the number of solving methods in SiSeSuSo that need to be augmented with some essential ones like AIC's?
The generator needs forcing nets so that the solver can use turbot-fishes as solving method.
Without those nets all generated puzzles will result in maximum pointing/claiming and very rarely in additional generalized intersection moves.
I'm interested if generating with JigsawExplainer will confirm this maximal rating for this layout.
You will have to generate a lot of puzzles to find a few harder ones.
urhegyi
 
Posts: 539
Joined: 13 April 2020

Re: Curiosities

Postby Hajime » Wed Mar 10, 2021 2:39 pm

urhegyi wrote:When finding JSWX layouts you need a portion of luck too. While a few days ago I found one which resulted in a high number of 8 clues puzzles, today's one results only in very easy puzzles. I think it's layout dependent. This one is to symmetrical from design. Even when I generate with the forcing net option on the highest rated puzzles I can get are:
Code: Select all
#1/WX/B4,JS
....9.....1.....9....964......3.7...............8.6......135....5.....4.....5....
117777722111777222114476662114555622344555662334555699344486699333888999338888899
#1/WX/B4,JS
..1...5...1.....9.....6.......3.7...............8.6.......3.....5.....4...9...4..
117777722111777222114476662114555622344555662334555699344486699333888999338888899


Me too. It stops with Naked/Hidden Subsets and Pointing/Claiming.
When generating from a fully solved puzzle as input, randomly some (symmetrical) cells will be emptied and then SiSeSuSo tries to solve it with the checked methods.
A next try will randomly empty different cells, and mayby you are lucky.
If you try not symmetrical (In Gen_Params) some progress is made (your first puzzle/solution):
Code: Select all
#1/WX/B4,JS
...49......2......3............178............9..4.3..........6......2...........
117777722111777222114476662114555622344555662334555699344486699333888999338888899

Response to the rest in blue:
Is it through the number of solving methods in SiSeSuSo that need to be augmented with some essential ones like AIC's?
Also X-Chain, XY-Chains and Fish are rare. AIC is on top of my TODO-list, together with "grouped" methods, but I find it hard to program.
The generator needs forcing nets so that the solver can use turbot-fishes as solving method.
Without those nets all generated puzzles will result in maximum pointing/claiming and very rarely in additional generalized intersection moves.
No, but if you check all methods, it is not necessary that the resulting puzzle needs all checked methods
I'm interested if generating with JigsawExplainer will confirm this maximal rating for this layout.
You will have to generate a lot of puzzles to find a few harder ones.
Sad but true

Other generators can create a lot faster a puzzle (from a fully filled grid or from a complete empty grid), but checking if the puzzle is valid is done by BruteForce/Backtrack. If there are 2 or more solutions, this is an invalid puzzle. So you can generate 100's of puzzles pretty quick (task A).
But at the end all those puzzles need to be solved with logical methods (task B) before it can be decided which methods must be used and/or what the SE Rating is.
In SiSeSuSo task A and B are combined and manually.
User avatar
Hajime
 
Posts: 733
Joined: 20 April 2018
Location: Netherlands


Return to Sudoku variants