The New Sudoku Players' Forum

[urhegyi]

Sisesuso versions 11 &12 2023: It seems they can't find this xyz-wing and solve this jigsaw with an AIC while it is not needed.

Code: Select all

#1//B4,JS
..5.2.3...2.....1....315..............86.37..............276....5.....4...6.9.1..
111111233141412223444445233748545223778555233788565263778566666788896969778999999


[Hajime]

The implementation is according to the document: http://www.sudokuwiki.org/sudoku/Generalizing_Sudoku_Strategies.pdf
SiSeSuSo will check all types of houses (boxes,Asterisk,Girandola,Windoku) against rows and columns where applicable. So not 2 jigsaws...
But in the example of urhegyi above it is a jigsaw and a row. So we examine this (in row box situation).

General: house X and Y intersect. S is subset of cells of X and T is subset of Y. S and T do intersect also in subset C.

Below the cells without a minus are the (W) XYZ cells and in cells with a minus all mentioned candidates can be eliminated.

Type A=(if the selected cells in S and T outside C share no values, then no value in one of these cells is valid in other cells in X or Y, outside the cells of union S and T )
Example type A:

Code: Select all

-1234  1234  1234 | 12 -12 -12 | -12 -12 -12
-34   -34    34   |
-34   -34   -34   |

Type B=(if all pairs of cells in S or T outside C have at most one value v in common, then v is not valid in other cells in the intersection X and Y outside C)
Example type B:

Code: Select all

-1 -1 123 | 12 . . | . . .
.  .  13  |
.  .  .   |

Type C: If candidate 1 is not part of the intersection cells, then also candidate 1 can be eliminated from cells that "see" all candidate 1 cells of the union.
Example type C:

Code: Select all

-1 234 23  | 12 . . | . . .
 .  .  134 |  -1
 .  .  .   |

Note: This type C is added in sisesuso, but is not in the pdf.

Now a new situation (type D) is here

Code: Select all

-1 -1 123 | 12 . . | . . .
.  .  13  | -1?
.  .  .   |

the cell with the ? has the focus.
Why is that valid???? Well it is not.

But with jigsaw that cell will "see" all of the XYZ cells, which can not happen in a standaard sudoku.
And sisesuso does not test on seeing 3 cells. That's why it is not spotted.
Also a 3rd house (column 6) is used to see the 3 cells. Sisesuso only test on 2 houses.

[Hajime]

The January 2024 version of SiSeSuSo.exe is now available at
https://sites.google.com/site/sisesuso .

Version 22 Jan 2024, 2024-01

• XY,H,M2,M3,W,S,L1,L2,L3-Wing method extended for non-standard Sudoku's (eg. jigsaws)
• AIC method already had a Chopping Axe extended method for subsets, now also for locked sets (pointing/claiming)
• Lots of HTML help texts renewed.

[urhegyi]

I read the description of UR type 5 here:
https://www.sudokuwiki.org/Unique_Rectangles
I have found an example which solves with UR type 5 first and than UR type 1.
Unfortunately Sisesuso needs some AIC to solve. Why doesn't it recognize type 5 as you mention Unique Rectangles (type 1,2,5) are supported?


Edit: Actually sudokuwiki type 5 is Hodoku type 6 or it's accompanied hidden rectangle

[Hajime]

Ok, Hodoku UR 5 is supported. I did not know SudokuWiki has a different numbering. Currently not at home, so I cannot check the working.

[eleven]

I read the description of UR type 5 here:
https://www.sudokuwiki.org/Unique_Rectangles
...
Edit: Actually sudokuwiki type 5 is Hodoku type 6 or it's accompanied hden rectangle

Yes, strange article on SudokuWiki. It says "As of January 2021 we have a new Unique Rectangle elimination!" in the description of a "Type 5" UR.
This type is known as Uniqueness Type 6 - UR meets X-Wing since 2006.

[urhegyi]

Above on 09-01-2024 I already posted a case where SiSeSuSo can't find an xyz-wing in a jigsaw. You answered this can't happen in 9x9 sudoku so it's normal that Sisesuso can't find it.
Here the same seems to happen.

XYZ-wing 256 in R5C6,R6C56 eliminates 2 from R6C7.
How can be explained that Sisesuso misses again?

Code: Select all

#1//B4,JS
3..6.7..47..143..6....3....9.......3..6...3..8.......9....1....5..371..21..2.9..8
333555111383585151388885551338788711437777712449799722499996662494696262444666222


[Hajime]

I try to state: it can not happen in a normal sudoku. the one with rows, columns and boxes. (Indeed somewhat clumsy described)
In jigsaw it is totally different.
Let's investigate Type B XYZ method:
S is a subset of cells in a house X, T is a subset of cells in another house Y. X and Y do intersect. S and T do intersect in subset C.
Currently implemented is Type B :
"If all pairs of cells in S or T outside C have at most one value v in common, then v is not valid in other cells in the intersection X and Y outside C"

Code: Select all

-1 -1 123 | 12 . . | . . .
.  .  13  |
.  .  .   |

I am busy to rebuild this to "If all pairs of cells in S or T outside C have at most one value v in common, then v is not valid in any cell that see all cells with v of S or T."
"Any cell" (in the whole puzzle) is much wider defined then a "cell in the intersection outside C".

In your last example R6C7 is not part of an intersection, but does 'see' all of the 3 cells R5C6,R6C56 with 256.

Same case in puzzle of 10jan2024 where r4c6 is not part of the intersection of row4 and jigsaw2 , but it sees all 3 xyz cells.

Edit: see all cells with v

[Hajime]

I implemented XYZ-method as above. Source-code needs to be cleaned up.
Problem in Urhegyi's puzzle of Wed Jan 10, 2024 12:37 am is fixed.
But still some other problem with his puzzle of Tue Mar 12, 2024 12:31 am
Working on it ....

[urhegyi]

Here another testcase:

[Hajime]

Thanks urhegyi. Tomorrow back from short holiday.
The 2nd jigsaw testcase is tackled. Found the xyz-wing.
Can you supply the digits-code of this 3rd jigsaw puzzle...

[urhegyi]

Code: Select all

#1//B4,JS
.2..8..9....2.1...............816...3.6...8.9...748...............3.2....7..6..1.
655555555665777777466767978446669988444999888443339988433131918332111111322222222

[Hajime]

Basics until:

Code: Select all

+-------+---------------------------------------------------------+   
|   45  |   2     13457   1456    8    457  14567     9     34567 |   
|       +-------+       +-----------------------------------------+   
|  4569   34569 |   45  |  2    3579    1    4567   345678  345678|   
+-------+       +-------+     +------+    +-------+       +-------+   
|   8   | 34569   234579| 4569| 23579| 459| 123459|  3456 | 13457 |   
|       +-------+       +-----+      +----+       +-------+       |   
| 24579    459  | 234579   8      1  |  6   23459 |  3457    3457 |   
|               +-------+------------+    +-------+               |   
|   3      145      6   | 145    57    457|   8       2       9   |   
|               +-------+------------+    +-------+               |   
|  1259    159  | 12359    7      4  |  8   12359 |  356     1356 |   
|       +-------+       +-----+      +----+       +-------+       |   
| 124579| 135689  12359 | 1569| 2359 | 59 | 123459|  568  | 13457 |   
+-------+       +-------+     +------+    +-------+       +-------+   
|  1569   15689 |   59  |  3     579    2   14567   45678   45678 |   
|       +-------+       +-----------------------------------------+   
|   59  |   7       8     459     6     3     45      1       2   |   
+-------+---------------------------------------------------------+   


Then several advanced methods to get the next clue:

Code: Select all

Pointing, Claiming  | (7)r5,J9 => (-7)r3c7 (-7)r4c7 (-7)r7c7
XYZ-Wing Type B on {579} in c5e5,c5e8,J1e2 (-5)c5J1 => (-5)r7c5
Generalized Intersection  | (9)r2,c5,J6 => (-9)r3c5 | (9)J1,r7,c5 => (-9)r7c5 | (1)J1,r7,c7 => (-1)r7c7
Skyscraper (9)r7c6=r3c6-r2c5=r8c5 => (-9)r7c4
Skyscraper (9)r3c6=r7c6-r8c5=r2c5 => (-9)r3c4
Hidden Single in col 4
stte


And yes, the solver finds the XYZ-wing
Busy cleaning up the source-code. May take some days for the next release of SiSeSuSo...

[Hajime]

Earlier than I thought:
The May 2024 version of SiSeSuSo.exe is now available at
https://sites.google.com/site/sisesuso .

Version 04 May 2024, 2024-05

(..W)XYZ method can now solve grouped situations and be used in jigsaw etc...
much help text expanded with images

[urhegyi]

I found 2 more cases in which I think your XYZ-wing routine fails:

Code: Select all

#1//B4,JS
.12...69....8.3...8.9...4.3.........79..3..86.........9.1...2.7...9.4....45...83.
111199999111188999221588889225577788322555778332225577433335677444336666444446666