Sudoku16x16Explainer - SE for 16x16 sudokus

Programs which generate, solve, and analyze Sudoku puzzles

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby 1to9only » Thu Nov 18, 2021 12:22 pm

1to9only wrote:SE16 did not rate any of blue's ultimate 16x16s puzzles posted here.
The 1st puzzle is at least ED=12.8, as from these solving steps:
Code: Select all
2.6, Pointing: Cells R1C1,R3C1: 15 in block and column
2.6, Pointing: Cells R6C6,R8C6: 1 in block and column
2.6, Pointing: Cells R15C1,R15C3: 3 in block and row
11.3, Region Forcing Chains: 3 in row ==> R2C16.3 off
12.0, Contradiction Forcing Chain: R5C4.10 on ==> R3C3.16 both on & off
12.4, Contradiction Forcing Chain: R2C7.7 on ==> R6C16.9 both on & off
12.6, Contradiction Forcing Chain: R9C2.10 on ==> R6C12.13 both on & off
12.8, Contradiction Forcing Chain: R2C7.9 on ==> R15C15.10 both on & off


About a year ago I fixed a bug where a couple of 9's were not changed to 16's in the original conversion!! ooops!

[Edit: 25Nov2021] These puzzles take a very very long time to serate. The 1st puzzle is rated ED=12.9/12.9/2.6.
[Edit: 29Nov2021] The 2nd of blue's puzzle is likely rated ED=13.1/12.9/2.6 (after 15 solving steps!!).
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby urhegyi » Wed Dec 08, 2021 7:16 pm

Like to analyze this sudoku and eventually dump the solvepath from the command prompt.
What's the correct input format(1..16 OR 1..G OR A..P)?
And the correct syntacs for this:
No more logfile, so have to redirect the output?

Usage - hints
java.exe -Xrs -Xmx500m -cp Sudoku16Explainer.jar diuf.sudoku.test.hints --input=puzzle(s).txt

Usage - analyze
java.exe -Xrs -Xmx500m -cp Sudoku16Explainer.jar diuf.sudoku.test.Tester --input=puzzle(s).txt

1..7..8...G.....
....2......C5891
8.63.CA4........
........458A..EC
.4BG.1.3........
.......87BD..51.
C.3.6G9B....7...
.7.E....61.3.G.9
.B.1....FD.8.7.2
9.4.G7EF....D...
.......DG69..CB.
.2E5.8.6........
........3AEB..57
2.14.93A........
....F......6813A
B..8..5...C.....
urhegyi
 
Posts: 739
Joined: 13 April 2020

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby 1to9only » Wed Dec 08, 2021 8:46 pm

urhegyi wrote:Like to analyze this sudoku and eventually dump the solvepath from the command prompt.

From the command-line, only 1 input format is supported: 1 line per puzzle, 256 chars, A-P format.
A puzzle converter to the A-P format used by SE16 is here.

Usage - serate
Code: Select all
java.exe -Xrs -Xmx1g -cp Sudoku16Explainer.jar diuf.sudoku.test.serate --input=puzzle(s).txt --output=output.txt

Usage - hints, pencilmarks
Code: Select all
java.exe -Xrs -Xmx1g -cp Sudoku16Explainer.jar diuf.sudoku.test.hints --input=puzzle(s).txt
java.exe -Xrs -Xmx1g -cp Sudoku16Explainer.jar diuf.sudoku.test.hints --input=puzzle(s).txt 1>output.txt
java.exe -Xrs -Xmx1g -cp Sudoku16Explainer.jar diuf.sudoku.test.hints --input=puzzle(s).txt 1>output.txt 2>nul

STDOUT output (long hints) should (always) be redirected to a file.
STDERR output (short hints) is for debug, it shows progress on screen, if not needed redirect this to a file or to nul.

Usage - Tester
Code: Select all
java.exe -Xrs -Xmx1g -cp Sudoku16Explainer.jar diuf.sudoku.test.Tester puzzles.txt puzzles.log

The other SEs work similarly.
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby urhegyi » Mon Jan 03, 2022 4:28 pm

1to9only wrote:
1to9only wrote:SE16 did not rate any of blue's ultimate 16x16s puzzles posted here.
The 1st puzzle is at least ED=12.8, as from these solving steps:
Code: Select all
2.6, Pointing: Cells R1C1,R3C1: 15 in block and column
2.6, Pointing: Cells R6C6,R8C6: 1 in block and column
2.6, Pointing: Cells R15C1,R15C3: 3 in block and row
11.3, Region Forcing Chains: 3 in row ==> R2C16.3 off
12.0, Contradiction Forcing Chain: R5C4.10 on ==> R3C3.16 both on & off
12.4, Contradiction Forcing Chain: R2C7.7 on ==> R6C16.9 both on & off
12.6, Contradiction Forcing Chain: R9C2.10 on ==> R6C12.13 both on & off
12.8, Contradiction Forcing Chain: R2C7.9 on ==> R15C15.10 both on & off


About a year ago I fixed a bug where a couple of 9's were not changed to 16's in the original conversion!! ooops!

[Edit: 25Nov2021] These puzzles take a very very long time to serate. The 1st puzzle is rated ED=12.9/12.9/2.6.
[Edit: 29Nov2021] The 2nd of blue's puzzle is likely rated ED=13.1/12.9/2.6 (after 15 solving steps!!).

I tried the 2nd blue's puzzle too, but it stopped here?
Can we conclude it needs more advanced steps than a 13.1?
logfile:
Code: Select all
2.6, Pointing: Cells R10C11,R12C11: 13 in block and column: r2c11<>13, r4c11<>13, r14c11<>13, r15c11<>13, r16c11<>13
2.6, Pointing: Cells R3C14,R3C16: 5 in block and row: r3c10<>5, r3c12<>5
2.6, Pointing: Cells R5C10,R5C12: 6 in block and row: r5c2<>6, r5c4<>6, r5c15<>6
2.6, Pointing: Cells R9C1,R9C3: 14 in block and row: r9c14<>14
2.6, Pointing: Cells R12C9,R12C11: 15 in block and row: r12c14<>15
2.6, Pointing: Cells R13C14,R13C16: 5 in block and row: r13c10<>5, r13c12<>5
12.8, Contradiction Forcing Chain: R11C16.16 on ==> R9C2.8 both on & off: r11c16<>16
12.8, Contradiction Forcing Chain: R1C16.15 on ==> R3C12.14 both on & off: r1c16<>15
12.8, Contradiction Forcing Chain: R11C16.4 on ==> R10C5.16 both on & off: r11c16<>4
12.9, Contradiction Forcing Chain: R10C11.5 on ==> R16C13.3 both on & off: r10c11<>5
12.9, Contradiction Forcing Chain: R11C14.15 on ==> R5C2.8 both on & off: r11c14<>15
1.2, Hidden Single: R11C16: 15 in block: r11c16=15
12.9, Contradiction Forcing Chain: R7C1.10 on ==> R4C15.13 both on & off: r7c1<>10
13.0, Contradiction Forcing Chain: R13C16.3 on ==> R4C14.9 both on & off: r13c16<>3
13.1, Contradiction Forcing Chain: R11C14.7 on ==> R10C5.3 both on & off: r11c14<>7
13.1, Contradiction Forcing Chain: R1C16.1 on ==> R15C12.14 both on & off: r1c16<>1
12.8, Contradiction Forcing Chain: R6C15.10 on ==> R9C14.12 both on & off: r6c15<>10
12.8, Contradiction Forcing Chain: R6C13.9 on ==> R7C3.1 both on & off: r6c13<>9
12.8, Contradiction Forcing Chain: R8C13.9 on ==> R16C9.14 both on & off: r8c13<>9
13.0, Contradiction Forcing Chain: R6C15.13 on ==> R14C1.13 both on & off: r6c15<>13
13.0, Contradiction Forcing Chain: R15C16.3 on ==> R8C13.2 both on & off: r15c16<>3
12.8, Contradiction Forcing Chain: R8C15.10 on ==> R9C10.12 both on & off: r8c15<>10
13.1, Contradiction Forcing Chain: R10C11.4 on ==> R14C7.16 both on & off: r10c11<>4
13.1, Contradiction Forcing Chain: R13C16.13 on ==> R1C10.12 both on & off: r13c16<>13
13.0, Contradiction Forcing Chain: R9C16.16 on ==> R15C12.8 both on & off: r9c16<>16
13.1, Contradiction Forcing Chain: R10C1.16 on ==> R2C1.13 both on & off: r10c1<>16
13.1, Contradiction Forcing Chain: R11C15.2 on ==> R8C15.6 both on & off: r11c15<>2
ED=20.0/12.9/2.6
urhegyi
 
Posts: 739
Joined: 13 April 2020

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby 1to9only » Sat Jan 08, 2022 12:09 pm

urhegyi wrote:Can we conclude it needs more advanced steps than a 13.1?

Some of the 13.1 steps are Contradiction Forcing Chain (w/ 24577 - 32768 nodes), at least 2 are 30k+ nodes.
The higher 13.2 rating has 32769 - 49152 nodes.
So SE could have run out of memory! A Java error would have appeared on screen.
Or something else...
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby urhegyi » Wed Jan 19, 2022 6:14 pm

I retried running blue-02 on different computers to exclude memory issues.
They all found line 27 in above log.
After this they all found a next step (after about 19h)but couldn't rate it(ER 20.0) nor display it in the log. No error message appeared.
See that there is now a new version that adds node counts, but it took up to 238h to arrive at the point the program halts with an earlier version.
Interested if it is possible to import the first 27 eliminations as a sukaku and then test only the last step in the new version(12 jan 2022).
urhegyi
 
Posts: 739
Joined: 13 April 2020

Postby 1to9only » Sat Jan 22, 2022 1:01 pm

Sudoku16Explainer has these techniques disabled because the 9x9 SE code did not convert well to the bigger 16x16 size:
UniqueLoop, AlignedPairExclusion, and AlignedTripletExclusion.

These will be enabled in a future release, as some improvements have been made.
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby urhegyi » Thu Jan 27, 2022 10:15 pm

Any chance this ED=7.1/1.2/1.2 rated hexadoku would be significantly lower rated when w-wing routine was enabled?
Code: Select all
.E2...D.C....7.4
.9..F....8.B...2
.C..2A8......G.E
..8.E.6.A..1.B..
2..A.5.619F.....
...5.91G..3.F.E.
...4....8.C.2.1.
.....2A.....3.48
73.D.....6G.....
.2.C.4.F....1...
.B.9.G..7F4.8...
.....7CD3.E.4..9
..7.9..1.G.3.4..
9.1......CA8..5.
G...D.5....7..C.
A.E....3.4...8G.

In 256 bytes on one line in A-P format:
Code: Select all
.NB...M.L....G.D.I..O....H.K...B.L..BJH......P.N..H.N.F.J..A.K..B..J.E.FAIO........E.IAP..C.O.N....D....H.L.B.A......BJ.....C.DHGC.M.....FP......B.L.D.O....A....K.I.P..GOD.H........GLMC.N.D..I..G.I..A.P.C.D..I.A......LJH..E.P...M.E....G..L.J.N....C.D...HP.
urhegyi
 
Posts: 739
Joined: 13 April 2020

Postby 1to9only » Fri Jan 28, 2022 8:18 am

urhegyi wrote:Any chance this ED=7.1/1.2/1.2 rated hexadoku would be significantly lower rated when w-wing routine was enabled?

There is no SE 16x16 solver with W-Wing technique available.
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Postby 1to9only » Mon Jan 31, 2022 9:29 am

A problem with convert16x16.exe was reported here.
This is now fixed, the updated program can be downloaded here.
The updated program is 64-bit only.
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby urhegyi » Wed Feb 02, 2022 8:24 am

update blue 16x16 ultimate:
nr1: solved in 43-44h:ED=12.9/12.9/2.6
https://github.com/Urhegyi/Sudoku-Uploads/blob/main/sol-blue-mist-01.txt
nr2 :couldn't continue after 237-238h: only 27 lines solved: ED=20.0
nr7: solved in 30h:ED=12.8/12.8/2.6
https://github.com/Urhegyi/Sudoku-Uploads/blob/main/sol-blue-mist-07.txt
urhegyi
 
Posts: 739
Joined: 13 April 2020

Re: Sudoku16x16Explainer - SE for 16x16 sudokus

Postby urhegyi » Mon Feb 14, 2022 8:01 am

Rechecked blue ultimate 16x16 nr 2 with the last available version of 02/02/2022. No progress after 27 solving lines. Stopped at exactely the same place as all versions I tested before. Memory use without using Xrs and Xmx parameters is not problematical between 3.5 and 4.5 gigabytes. What will be the exact rating for the last step it determines as SE 20.0? Is it a kind of chaining level not enabled because it will take to much time to rate? Interested to know why it stops exactely on this line. Hope there comes more clarity in the future.
Code: Select all
I.J.M.E.....N.D..K.B.P...I...C.JL.C.I...G.P...K..N.E...C.H.D...BG.B...C.K.N.L....E...B.L.O.A.P......H.F.M.B.O.E....M.K.A.C.L.H.N....O.A.J.F.M.I..O...F.N.A.B.K..A.M...L.C.H.E....J.C...E.N.G...FB.D.A...F.L...P..G.N.E...K...D.HF.E.K.D.....J.O..P.A.G.I...O.F.L
urhegyi
 
Posts: 739
Joined: 13 April 2020

Postby 1to9only » Mon Feb 14, 2022 10:01 am

urhegyi wrote:Is it a kind of chaining level not enabled because it will take to much time to rate?

The ED=20.0 means that no solving step was found after trying all the solving techniques in the program.

The solving techniques in SE16 are the same as those in SE9:
Code: Select all
Base level
 SER | nestinglimit (0=nonesting)
 |   | |
 8.5 0 0 DynamicForcingChain
 9.0 1 0 DynamicForcingChainPlus +
 9.5 2 0 NestedForcingChain + Forcing Chains
10.0 3 0 NestedForcingChain + Multiple Forcing Chains (8.0)
10.5 4 0 NestedForcingChain + DynamicForcingChain (8.5)
10.5 4 1 NestedForcingChain + DynamicForcingChainPlus (9.0)
10.5 4 2 NestedForcingChain + NestedForcingChain (9.5)

Edit: Updated the above - from looking up the source code, and I may still be wrong!!

The last technique 4-2 (level-nesting) may be the one that returns ED=13.1 for blue's #2 puzzle.
I think (not too sure about this) that it is only used in some of the very high rated 9x9 sudokus.

4-3 (level-nesting) could be added, but the program will run for even longer...
Code: Select all
10.5 4 3 NestedForcingChain + NestedForcingChain (10.0)

I might try this later, but it may not happen as it ties up the computer for so long...
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

Re:

Postby 999_Springs » Sat Feb 26, 2022 5:53 pm

1to9only wrote:The last technique 4-2 (level-nesting) may be the one that returns ED=13.1 for blue's #2 puzzle.
I think (not too sure about this) that it is only used in some of the very high rated 9x9 sudokus.

i'd be very surprised if SE has a chain level beyond nested dynamic fc (4 0 in your notation)

all 9x9 puzzles solve with nested dynamic fc. if you mean pencilmark 9x9 puzzles instead, i haven't had a detailed look but my guess is that they stop at nested dynamic fc too - this is because if you click "get all hints" in the GUI in regular SE on a 11.9 step and wait 15 minutes for all of them to load, you get a huge list of nested dynamic fc hints that top out at 12.7, which is close to the ratings of the top 9x9 pencilmark sudokus

does anyone have an example of a 9x9 pencilmark sudoku that sukakuexplainer (or similar) can solve and specifically uses a nested dynamic-plus fc in the solve path?
999_Springs
 
Posts: 591
Joined: 27 January 2007
Location: In the toilet, flushing down springs, one by one.

Postby 1to9only » Sat Feb 26, 2022 7:23 pm

All the 11.8's/11.9's from the Patterns Game and champagne's ph_2010 list are Nested Forcing Chains (+ Dynamic Forcing Chains), in my above ist:
10.5 4 0 NestedForcingChain + DynamicForcingChain (8.5)

For blue' s Ultimate 16x16s, I have a version of SE16 that outputs some screen info, and I think the 12.9-13.1 steps are probably:
10.5 4 1 NestedForcingChain + DynamicForcingChainPlus (9.0)

This is unconfirmed. I might (in future) output some additional info for each step solved.
User avatar
1to9only
 
Posts: 4175
Joined: 04 April 2018

PreviousNext

Return to Software