SE rating for overlapping Sudoku puzzles - suggestion

Programs which generate, solve, and analyze Sudoku puzzles

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby 1to9only » Tue Oct 29, 2019 8:20 pm

Hajime wrote:Let's try an easy samurai that can be solved with singles only. But neither subgrids can be solved individually.

George-NJ3H's samurai posted here:
Code: Select all
..1..9......286.....4...3.95.9..........1...4....6..2..6.5........1.....97....... ED=1.5/1.2/1.2
...7.8.9.1.....47..4.........2.9...7....21...3.8.....4.2.............8.5...2.4... ED=1.5/1.2/1.2
.......2....9........37..........3.66.8..........2...7...6..........5........1... ED=2.3/1.2/1.2
..5..9.....7..........6.......8.....1.....4.35..9.6....1.7...4.......162.2..8.... ED=1.5/1.2/1.2
......3.8...5.4......2..4..8.5.6..4.....2..7.9......8..63.....7.....9....7...6..1 ED=2.0/1.2/1.2
ED=2.3/1.2/1.2
781349256395286741624751389519472638236815974847963125163528497452197863978634512
235748196186935472947612358612493587754821963398576214523187649471369825869254731
497168523863952471512374869249517386678493215135826947321689754986745132754231698
865479321247135986391268754972843615186527493534916278613752849758394162429681537
754691328132584796698273415825967143346128579917345682463812957581739264279456831
Solved.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby Hajime » Wed Oct 30, 2019 8:00 am

So the overall SE is the maximum of the subgrid's SE.
I think that is not fair...
User avatar
Hajime
 
Posts: 1385
Joined: 20 April 2018
Location: Fryslân

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby 1to9only » Wed Oct 30, 2019 9:05 am

The samurai rating is ED=HighestER/LowestEP/LowestED, as in another of nj3h's samurai posted here:
Code: Select all
983.........6....4.4.8.5..........52....28.....5...7...2...1.......9....8.7.5.... ED=1.5/1.2/1.2
...6..9..3.6..1....5....463...2.4..6....9.1..58...6......5...2........59....2..3. ED=4.4/1.5/1.5
............4.9......8.7....62.3..........172.51.7.......7.4......2.5............ ED=1.5/1.2/1.2
5.9.7........9.....7...3.....2...3......41..........27.5.6.8......3....6826...... ED=2.0/1.2/1.2
....2..1........92...8...6.58...1.......6.5.....7.5..6.7....4533.4..7......1..9.. ED=1.5/1.2/1.2
ED=4.4/1.2/1.2
983214675152679834746835291398167452471528963265943718629381547534792186817456329
247635918396481572158972463971254386632897145584316297819563724723148659465729831
547623819186459723329817465762138594938546172451972386213764958874295631695381247
589476213361592874274183695492867351735241968618935427953618742147329586826754139
958326714631574892247819365586431279713962548429785136172698453394257681865143927
Solved.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby Hajime » Wed Oct 30, 2019 9:25 pm

1to9only wrote:The samurai rating is ED=HighestER/LowestEP/LowestED

Who decided this, can that be changed or discussed at least?
A Samurai is more difficult that its toughest sub grid.
And a Sumo is more difficult than a Samurai
User avatar
Hajime
 
Posts: 1385
Joined: 20 April 2018
Location: Fryslân

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby m_b_metcalf » Wed Oct 30, 2019 10:43 pm

Hajime wrote:
1to9only wrote:The samurai rating is ED=HighestER/LowestEP/LowestED

Who decided this, can that be changed or discussed at least?
A Samurai is more difficult that its toughest sub grid.

Surely that depends on whether the Samurai can be solved by iterating over the five subgrids or whether, as in ruud's hardest, a global 21x21 approach is required.

Regards,

Mike Metcalf
User avatar
m_b_metcalf
2017 Supporter
 
Posts: 13637
Joined: 15 May 2006
Location: Berlin

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby 1to9only » Wed Oct 30, 2019 11:32 pm

Hajime wrote:Who decided this, can that be changed or discussed at least?

Any rating system is subjective... I've given a few examples of how my own SamuraiExplainer is progressing, I've currently hit a few snags rating the harder samurais, so it's likely to be mothballed for a few months before I have another hack at it.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby 1to9only » Thu Oct 31, 2019 12:30 pm

SE is unsuitable for solving (rating) the harder samurai puzzles, this is mainly because SE expects solving 9x9 sudokus having single solutions, whereas in samurais the subgrids will often have multiple solutions, and m_b_metcalf's suggestion of a 21x21 grid is the better approach.

Here is another of nj3h's samurai posted here:
Code: Select all
.72.....5.9.1.5.......8.2..6...92...1......8.......6329.1...........8.....3..4...
...83......7...8....2.5..6..2.4......4....5.773...6..9......97....641......9.....
.....2................1.....42.....9...1..8..8...6.........3........6......7.....
...5.....5.4..2...6.3...........4.9.156.........3....8.3..4.65..61.7........2.84.
...6....4...38.......7..1..65...7.8......26.........5.8.1.....6.......92..2.56...

This is correctly solved up to this point:
Code: Select all
.72.....5.9.125....1..8.2..63..92...12.....8...9..16329.12........9.8.2.2.3.14... ED=2.6/1.2/1.2
...83.7....716.8....275..6..2.4......4....5.773...6..9......97.97.641...2..9..... ED=2.6/1.2/1.2
.....2....2....97.....1.2...42.....9...1..8..8...6.....64..3........6.2.285741396 ED=3.6/1.2/1.2
...538.64584.62...6.349.2853.8..4.96156.8.4...4.3.6..8832.4.657461875...795623841 1.0, Hidden Single: R9C2: 9 in block
...6....4.2.38..65...7.51.865...728....5.26..2...6..5.8.12..5.6.......929.2.56... ED=1.7/1.2/1.2

SE is solving subgrid 4, and r9c2=9 is the last solving step.
Code: Select all
+-------------+-------------+-------------+
| 29  12  79  | 5   3   8   | 17  6   4   |
| 5   8   4   | 17  6   2   | 39  17  39  |
| 6   17  3   | 4   9   17  | 2   8   5   |
+-------------+-------------+-------------+
| 3   27  8   | 27  15  4   | 15  9   6   |
| 1   5   6   | 279 8   79  | 4   237 23  |
| 29  4   79  | 3   15  6   | 157 12  8   |
+-------------+-------------+-------------+
| 8   3   2   | 19  4   19  | 6   5   7   |
| 4   6   1   | 8   7   5   | 39  23  239 |
| 7   9   5   | 6   2   3   | 8   4   1   |
+-------------+-------------+-------------+

Then SE finds this:
Code: Select all
4.5, Unique Rectangle type 1: Cells R2C7,R2C9,R8C9,R8C7 on 3, 9

SE then removes candidates 3,9 from r8c9, and it later solves the hidden single 2 in r8c9 - this leads to an unsolvable samurai.
.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby m_b_metcalf » Thu Oct 31, 2019 3:34 pm

I'm away from home at the moment and cannot access my samurai solver, but I can investigate sub-puzzle 4, and note that, if one looks for all its solutions and retains the common clues, then one gets
Code: Select all
 . . . 5 3 8 . 6 4
 5 8 4 . 6 2 . . .
 6 . 3 4 9 . 2 8 5
 3 . 8 . . 4 . 9 6
 1 5 6 . 8 . 4 . .
 . 4 . 3 . 6 . . 8
 8 3 2 . 4 . 6 5 7
 4 6 1 8 7 5 . . .
 7 9 5 6 2 3 8 4 1

No. of givens = 53

brute found 3 solution(s)

Common clues (74):
 9 2 7 5 3 8 1 6 4
 5 8 4 1 6 2 . 7 .
 6 1 3 4 9 7 2 8 5
 3 7 8 2 1 4 5 9 6
 1 5 6 7 8 9 4 . .
 2 4 9 3 5 6 7 1 8
 8 3 2 9 4 1 6 5 7
 4 6 1 8 7 5 . . .
 7 9 5 6 2 3 8 4 1

thereby providing two more values in an overlap region and thus, presumably, advancing the puzzle towards a final solution. However, none of that fits into an SE rating framework

Regards,

Mike
User avatar
m_b_metcalf
2017 Supporter
 
Posts: 13637
Joined: 15 May 2006
Location: Berlin

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby 1to9only » Thu Oct 31, 2019 9:44 pm

I notice that SudokuSolver solves this samurai by taking a different solving path:
Code: Select all
Switched to grid: 4
133. Grid:04 Unique Corner on {15} at r4c5, r4c7, r6c5, r6c7
133a. Cell r6c7 cannot contain {15}
134. Grid:04 Naked Single 7 found at r6c7
135. Grid:04 Naked Single 1 found at r1c7

By grid4 r1c7=1, switching to grid3 solves grid4 r2c7=9 (this is grid3 r8c1=9) and the grid4 UR goes away.
.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby 1to9only » Fri Nov 01, 2019 12:28 pm

1to9only wrote:Here is another of nj3h's samurai posted here:
Code: Select all
.72.....5.9.1.5.......8.2..6...92...1......8.......6329.1...........8.....3..4...
...83......7...8....2.5..6..2.4......4....5.773...6..9......97....641......9.....
.....2................1.....42.....9...1..8..8...6.........3........6......7.....
...5.....5.4..2...6.3...........4.9.156.........3....8.3..4.65..61.7........2.84.
...6....4...38.......7..1..65...7.8......26.........5.8.1.....6.......92..2.56...

This has now been solved by tweaking SE to find the 1,5 UR first, and disabling BUG (as it screws up grid3 later):
Code: Select all
.72.....5.9.1.5.......8.2..6...92...1......8.......6329.1...........8.....3..4... ED=3.0/1.2/1.2
...83......7...8....2.5..6..2.4......4....5.773...6..9......97....641......9..... ED=4.0/1.2/1.2
.....2................1.....42.....9...1..8..8...6.........3........6......7..... ED=3.6/1.2/1.2
...5.....5.4..2...6.3...........4.9.156.........3....8.3..4.65..61.7........2.84. ED=4.5/1.2/1.2
...6....4...38.......7..1..65...7.8......26.........5.8.1.....6.......92..2.56... ED=4.2/1.2/1.2
ED=4.5/1.2/1.2
872369145496125873315487269638792514124653987759841632941276358567938421283514796
614832795597164832382759164829475316146398527735216489461523978978641253253987641
358972461421635978796418253542387619637129845819564732164293587973856124285741396
927538164584162973613497285378214596156789432249356718832941657461875329795623841
587621934124389765396745128659417283748532619213968457831294576465173892972856341
Solved.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: SE rating for overlapping Sudoku puzzles - suggestion

Postby Hajime » Sat Nov 02, 2019 8:43 am

After each candidate elimination (and cell solving) in overlapping areas you need to start all over again with all the methods in ascending SE order for involved subgrids. See my code on 16 sept of this thread.
Pretty slow algorithm that probably can be optimised.
User avatar
Hajime
 
Posts: 1385
Joined: 20 April 2018
Location: Fryslân

Previous

Return to Software