Overlapping Windoku sensei4

For fans of Killer Sudoku, Samurai Sudoku and other variants

Overlapping Windoku sensei4

Postby urhegyi » Thu Jan 28, 2021 6:55 pm

After 41 minutes calculation time I created this monster:
Code: Select all
2.....1.41..........7.4..6.7..................8...........................4...... 955596669711172227711172227711172227955596669833384448833384448833384448955596669 JSB
.......9.......2...................................41...........69.......3......9 955596669711172227711172227711172227955596669833384448833384448833384448955596669 JSB

Anyone capable to verify?
SiSeSuSo analyze log:
Code: Select all
Eliminated candidates per SER and per Sudoku

Method    \ Sudoku |   SER |     1     2
                   |-------|------------
Not counted elims  |     0 |    78    72
Naked Singles      |   0.1 |    46    23
Hidden Singles     |   0.2 |    69    70
Naked Single  [1]  |   2.5 |    10    19
Naked Pair    [2]  |     3 |    21    35
Naked Triple  [3]  |   3.6 |    26     3
Naked Quad    [4]  |     5 |     4     0
Hidden Quad   [5]  |   5.4 |     1     0
Hidden Triple [6]  |     4 |    18     3
Locked Singles[2]  |   2.8 |    54    36
Locked Singles[3]  |   2.9 |     5    13
Locked Singles[4]  |     3 |     6    10
WXYZ Wing     [4]  |   5.5 |     2     2
VWXYZ Wing    [5]  |   6.3 |     3     1
UVWXYZ Wing   [6]  |   6.6 |     7     2
TUVWXYZ Wing  [7]  |   7.5 |     5     2
>TUVWXYZ Wing [8+] |     8 |     2     1
                   |-------|------------
Eliminated Cand's  |   649 |   357   292
Sum(SER * Cand's)  |1024.3 | 610.7 413.6

Initial Candidates :   650
Maximum SER rating :     8
Labour rating      :1024.3
Total Check ? 650 650

Question to 1to9only: can your sensei4 solver handle JSB?
urhegyi
 
Posts: 361
Joined: 13 April 2020

Re: Overlapping Windoku sensei4

Postby 1to9only » Thu Jan 28, 2021 7:20 pm

urhegyi wrote:Question to 1to9only: can your sensei4 solver handle JSB?

As I mentioned earlier the current elsewhere, JigsawExplainer samurai solver treats all overlapped grids as jigsaws (JS) so will not solve this JSBs and X variants.
1to9only
 
Posts: 2349
Joined: 04 April 2018

Re: Overlapping Windoku sensei4

Postby creint » Thu Jan 28, 2021 7:31 pm

Easy, only 7 layers of locked singles.
Why did it take 41 minutes, you can generate one within 2 seconds.

Here one that took 3 seconds to generate using a solver and 50 seconds to solve using my solver.

Code: Select all
#2//B4,JSB/E10,JSB
2......9.............14.5..7.............4.....16....5...........................
955596669711172227711172227711172227955596669833384448833384448833384448955596669
...........4......6....58..........3.................6.........2.....7.........2.
955596669711172227711172227711172227955596669833384448833384448833384448955596669


Hidden Text: Show
Code: Select all
2 . . . . . . 9 .     
. . . . . . . . .     
. . . 1 4 . 5 . .     
7 . . . . . . . . . . .
. . . . . 4 . . . . . .
. . 1 6 . . . . 5 8 . .
. . . . . . . . . . . 3
. . . . . . . . . . . .
. . . . . . . . . . . 6
      . . . . . . . . .
      2 . . . . . 7 . .
      . . . . . . . 2 .
creint
 
Posts: 248
Joined: 20 January 2018

Re: Overlapping Windoku sensei4

Postby Hajime » Fri Jan 29, 2021 9:07 am

creint wrote:Easy, only 7 layers of locked singles.
I think your famous inter-sub-grid locked singles does the trick, isn't?

Why did it take 41 minutes, you can generate one within 2 seconds.
SiSeSoSo generates a puzzle from a fully solved puzzle. Each round some cells are emptied and will try to solve it with logic methods.
So that takes a while, at the beginning pretty fast but at the end much slower because more methods are needed.

Here one that took 3 seconds to generate using a solver and 50 seconds to solve using my solver.
How do you create a puzzle in 3 secs and than 50 secs to see if it is valid? Can you elaborate on that?

Code: Select all
#2//B4,JSB/E10,JSB
2......9.............14.5..7.............4.....16....5...........................
955596669711172227711172227711172227955596669833384448833384448833384448955596669
...........4......6....58..........3.................6.........2.....7.........2.
955596669711172227711172227711172227955596669833384448833384448833384448955596669
My solver with all its methods checked ON can only find 1 cell to be solved: g1r2c4=2 !! :shock: :o but that cell can also be found using "singles only" :D
Last edited by Hajime on Fri Jan 29, 2021 9:36 am, edited 1 time in total.
User avatar
Hajime
 
Posts: 326
Joined: 20 April 2018
Location: Netherlands

Re: Overlapping Windoku sensei4

Postby Hajime » Fri Jan 29, 2021 9:20 am

urhegyi wrote:
Code: Select all
WXYZ Wing     [4]  |   5.5 |     2     2
VWXYZ Wing    [5]  |   6.3 |     3     1
UVWXYZ Wing   [6]  |   6.6 |     7     2
TUVWXYZ Wing  [7]  |   7.5 |     5     2
>TUVWXYZ Wing [8+] |     8 |     2     1

Those lots of (STUVW)XYZ wings worries me a bit. I will look into that....
Puzzle solved, confirmed by creint, but still a little weird.
User avatar
Hajime
 
Posts: 326
Joined: 20 April 2018
Location: Netherlands

Re: Overlapping Windoku sensei4

Postby creint » Fri Jan 29, 2021 10:35 am

Generating with a logic solver like miniSat or with DLX is fast (it already check if its valid), but you don't know the rating unless you solve it.
Solving with logic solver like DLX is fast < 0.1 sec, but using methods like singles/chains it does take <50 seconds in my solver.
creint
 
Posts: 248
Joined: 20 January 2018

Re: Overlapping Windoku sensei4

Postby Hajime » Fri Jan 29, 2021 12:21 pm

creint wrote:Generating with a logic solver like miniSat or with DLX is fast (it already check if its valid), but you don't know the rating unless you solve it.
Solving with logic solver like DLX is fast < 0.1 sec, but using methods like singles/chains it does take <50 seconds in my solver.

SiSeSuSo solves the puzzle of urhegyi between 0.085 and 0.400 seconds. Because there is a random inside the BruteForce method time varies.
I think BruteForce or DLX is not a logic method, BTW. Solving this puzzle with logic methods takes 7.0 secs.

Solving your puzzle with BruteForce varies between 0.2 en 2.0 secs.
User avatar
Hajime
 
Posts: 326
Joined: 20 April 2018
Location: Netherlands

Re: Overlapping Windoku sensei4

Postby urhegyi » Fri Jan 29, 2021 3:01 pm

Later today I will try to solve them with two windoku solvers in parallel. Start with 1 grid till it can't continue. Do the same for the other grid. Update the common area pencilmarks by hand in both processes. Continue with the other grid. And so on till it's solved, hopefully?
urhegyi
 
Posts: 361
Joined: 13 April 2020

Re: Overlapping Windoku sensei4

Postby creint » Fri Jan 29, 2021 7:34 pm

Updating pencilmarks and solve them partly does not work.
How to solve after this when overlap has same pencilmarks:
Code: Select all
2        134568   34568    3578     35678    35678    13678    9        134678
135689   3456789  3456789  2        35689    36789    134678   13678    1368
3689     36789    36789    1        4        36789    5        23678    2368
7        34569    34569    3589     123589   12389    1234689  123468   13689
3569     23569    23569    5789     135789   4        12368    1278     13679
3489     23489    1        6        2379     2379     2379     247      5
13489    2345789  2345789  45789    12689    125689   12689    124578   124679
1489     245789   245789   34589    236789   12356789 12346789 1234568  1346789
145689   1456789  2456789  345789   13789    123578   12378    13578    13489

Code: Select all
3589     123589   12389    1234689  123468   13689    34569    345679   24579
5789     135789   4        12368    1278     13679    2369     369      259
6        2379     2379     2379     247      5        8        1349     1249
45789    12689    125689   12689    124578   124679   1249     1479     3
34589    236789   12356789 12346789 1234568  1346789  1459     145789   1245789
345789   13789    123578   12378    13578    13489    12459    14589    6
135789   3456789  356789   12345678 134789   123468   134569   1345689  14589
2        345689   35689    134568   13489    13468    7        1345689  14589
13589    345678   356789   1345678  134689   134678   134569   2        14589
creint
 
Posts: 248
Joined: 20 January 2018

Re: Overlapping Windoku sensei4

Postby Hajime » Fri Jan 29, 2021 8:15 pm

creint wrote:Updating pencilmarks and solve them partly does not work

So what is the secret method (not being brute force)?
Your famous locked singles between 2 sub-grids? Busy implementing this !
Or yet another method?
User avatar
Hajime
 
Posts: 326
Joined: 20 April 2018
Location: Netherlands

Re: Overlapping Windoku sensei4

Postby urhegyi » Fri Jan 29, 2021 8:36 pm

SiSeSuSo_20210129_213115.png
SiSeSuSo_20210129_213115.png (33.95 KiB) Viewed 76 times
urhegyi
 
Posts: 361
Joined: 13 April 2020

Re: Overlapping Windoku sensei4

Postby creint » Fri Jan 29, 2021 10:28 pm

Hajime wrote:
creint wrote:Updating pencilmarks and solve them partly does not work

So what is the secret method (not being brute force)?
Your famous locked singles between 2 sub-grids? Busy implementing this !
Or yet another method?


With the generic locked singles you cannot solve my puzzle.
Forcing nets will solve it so < SE 10
creint
 
Posts: 248
Joined: 20 January 2018

Re: Overlapping Windoku sensei4

Postby urhegyi » Sat Jan 30, 2021 5:09 pm

Hajime wrote:
urhegyi wrote:
Code: Select all
WXYZ Wing     [4]  |   5.5 |     2     2
VWXYZ Wing    [5]  |   6.3 |     3     1
UVWXYZ Wing   [6]  |   6.6 |     7     2
TUVWXYZ Wing  [7]  |   7.5 |     5     2
>TUVWXYZ Wing [8+] |     8 |     2     1

Those lots of (STUVW)XYZ wings worries me a bit. I will look into that....
Puzzle solved, confirmed by creint, but still a little weird.

To reproduce this situation with a lot of wings consider the normal sudoku:
Code: Select all
...4.8....85...12..1.3.5.9....1.7.....2.4.5...6.....4...7...9...31...78.8..7.3..6

Solving this with singles, hidden/naked subsets, pointing/claiming and STUVWXYZ-wing:
Code: Select all
Eliminated candidates per SER and per Sudoku

Method    \ Sudoku |   SER |     1
                   |-------|------
Not counted elims  |     0 |    92
Naked Singles      |   0.1 |    39
Hidden Singles     |   0.2 |    43
Naked Pair    [2]  |     3 |     2
Naked Quad    [4]  |     5 |     6
Locked Singles[2]  |   2.8 |     3
VWXYZ Wing    [5]  |   6.3 |     1
UVWXYZ Wing   [6]  |   6.6 |     2
TUVWXYZ Wing  [7]  |   7.5 |     3
>TUVWXYZ Wing [8+] |     8 |     1
                   |-------|------
Eliminated Cand's  |   192 |   192
Sum(SER * Cand's)  | 106.9 | 106.9

Initial Candidates :   192
Maximum SER rating :     8
Labour rating      : 106.9

Solving without STUVWXYZ-wing but x-wing/swordfish enabled:
Code: Select all
Eliminated candidates per SER and per Sudoku

Method    \ Sudoku |   SER |     1
                   |-------|------
Not counted elims  |     0 |   101
Naked Singles      |   0.1 |    47
Hidden Singles     |   0.2 |    23
Naked Pair    [2]  |     3 |     2
Naked Quad    [4]  |     5 |     6
Locked Singles[2]  |   2.8 |     3
X-Wing        [2]  |   3.2 |     4
Swordfish     [3]  |   3.8 |     6
                   |-------|------
Eliminated Cand's  |   192 |   192
Sum(SER * Cand's)  |  89.3 |  89.3

Initial Candidates :   192
Maximum SER rating :     5
Labour rating      :  89.3
urhegyi
 
Posts: 361
Joined: 13 April 2020

Re: Overlapping Windoku sensei4

Postby Hajime » Sat Jan 30, 2021 5:35 pm

Thank you urhegyi. I am busy to limit the dept of some methods, that also will limit the length/size/count of the found chains/cells. So the user can test if some method is somehow redundant, so eg no Jellyfish or no STUVWXYZ-wing.
User avatar
Hajime
 
Posts: 326
Joined: 20 April 2018
Location: Netherlands

Re: Overlapping Windoku sensei4

Postby urhegyi » Sat Jan 30, 2021 6:41 pm

Hajime wrote:Thank you urhegyi. I am busy to limit the dept of some methods, that also will limit the length/size/count of the found chains/cells. So the user can test if some method is somehow redundant, so eg no Jellyfish or no STUVWXYZ-wing.

Yellyfish (length 4) causes no problems, you can keep them, but it are the longer wings and x(y)-chains with a length > 5 that take more time and can be often replaced by another simpler variant of solving method. The sets of length 4 and 5 are essential for the working of the solver, the higher ones not. So the only thing to consider with fishes is when you find a yellyfish in a row, there's always a shorter one in a column and vice versa. Same counts for a swordfish which can often be replaced by an x-wing by switching the dimensions of your search (rows<-->columns).
urhegyi
 
Posts: 361
Joined: 13 April 2020


Return to Sudoku variants