The New Sudoku Players' Forum

Website · by **denis_berthier** » Thu Dec 09, 2021 7:09 am

mith wrote:Haven't been able to make it up to the office much since my surgery, but I've left the scripts running and have a new 11.8 (and some 11.7s):
Code: Select all
21c ..............1.23...24.5....2..3..1..6....72.8....9....3..7.6..4..8....9...5.... ED=11.7/1.5/1.5 ..............1.23...24.5....3.....1.4.6..2..7.......8..8..7....6.92.....9..5.4.. ED=11.7/1.5/1.5 22c ........1....23.....45...6...5..17...7.8...5.9....5..3.56....7..8.4.....2.......9 ED=11.8/1.2/1.2 ........1....23.....45...6...5..17...7.8...5.9.......3.56....7..8.4.....2...5...9 ED=11.7/1.2/1.2

Hi Mith
congrats for a new 11.8
I checked its BpB class: B6B (or less)

by **mith** » Mon Dec 20, 2021 12:50 am

Nothing higher than 11.6 since last update, but one of them is a new 31c (only the second such minimal):

Code: Select all: ........1.....2.3..45..62......2.3.7.17.8..623..6.7.8....7..81...1.68..3.832..7.6 ED=11.6/1.2/1.2

I need to look into whether this one simplifies with the parity argument ("Thor's Hammer", as we've taken to calling it, due the the shape it made in one of my examples). Boxes 6 and 9 are suspect, but it's not completely obvious how to proceed into boxes 5 and 8 (the most likely boxes). There is probably some version of the argument to make, but I'll have to revisit when I'm less tired.

by **marek stefanik** » Mon Dec 20, 2021 2:21 am

Very nice.

Code: Select all: .-------------------.--------------------.------------------. | 26789 3 2689 | 4589 4579 459 | 4569 4579 1 | | 16789 679 689 | 14589 14579 2 | 4569 3 4589 | | 1789 4 5 | 1389 1379 6 | 2 79 89 | :-------------------+--------------------+------------------: | 45689 569 4689 |*1459 2 %1459 | 3 #459 7 | | 459 1 7 |%3459 8 *3459 |#459 6 2 | | 3 259 249 | 6 #459 7 | 1 8 #459 | :-------------------+--------------------+------------------: | 24569 2569 2469 | 7 *3459 %3459 | 8 1 #459 | | 4579 579 1 |#459 6 8 |#459 2 3 | | 459 8 3 | 2 %1459 *1459 | 7 #459 6 | '-------------------'--------------------'------------------'

13r47c6 and 13r59c6 both create a Thor's Hammer => 13c6 must share a minicolumn, determining the braid pattern in the middle stack
13 in b2 must therefore also share a minicolumn => –1r3c45
Two digits must appear in each of r123c6, r456c5 and r789c4 => one is 2, let's call the other one x

Code: Select all: .-------------------.--------------------.------------------. | 26789 3 2689 | 4589 4579 x459 | 4569 4579 1 | | 6789 679 689 | 14589 14579 2 | 4569 3 4589 | | 1 4 5 | 389 379 6 | 2 79 89 | :-------------------+--------------------+------------------: | 45689 569 4689 | 1459 2 1459 | 3 459 7 | |#459 1 7 | 3459 8 3459 |#459 6 2 | | 3 259 249 | 6 x459 7 | 1 8 459 | :-------------------+--------------------+------------------: | 24569 2569 2469 | 7 3459 3459 | 8 1 459 | | 4579 579 1 |x459 6 8 | 459 2 3 | |#459 8 3 | 2 1459 1459 | 7 #459 6 | '-------------------'--------------------'------------------'

xr59\c18b6 => –xr4c8 => x in b6 must take r5c7
Now r5c1<>x, let's call it y. y in r6 must take r6c9. Let's call the remaining digit among 459 z.

Code: Select all: .-------------------.--------------------.------------------. | 26789 3 2689 | 4589 4579 x459 | 4569 4579 1 | | 6789 679 689 | 14589 14579 2 | 4569 3 4589 | | 1 4 5 | 389 379 6 | 2 79 89 | :-------------------+--------------------+------------------: | 45689 569 4689 | 1459 2 1459 | 3 459 7 | |y459 1 7 | 3459 8 3459 |x459 6 2 | | 3 259 249 | 6 x459 7 | 1 8 y459 | :-------------------+--------------------+------------------: | 24569 2569 2469 | 7 3459 3459 | 8 1 459 | |a4579 a579 1 |x459 6 8 |b459 2 3 | |a459 8 3 | 2 1459 1459 | 7 c459 6 | '-------------------'--------------------'------------------'

(x=yz7)b9p457 – yr8c7 = yr9c8 => –xr9c8 => x in c8 must take r3c8, x=9, the puzzle is essentially solved

Marek

by **mith** » Mon Dec 20, 2021 2:27 am

Nice solution, marek

I was just about to write up the minicolumn conclusion, along a more complicated argument why it can't be r79c6 in b8 (two cases result in your # cells in c1 ending up the same color, but the third case is nasty and I ended up coloring the whole grid before getting 459 in r1c468 which breaks the 4579 in r1c57).

Just the minicolumn thing (placing the 1 in r3) reduces skfr to 10.9, and placing the 13 in c5b8 reduces it further to 10.0, but your argument is much more effective.

by **mith** » Tue Dec 21, 2021 10:47 pm

Code: Select all: 29c ........1.......2.....13.45...3.56....5.78....3716.....6.73...8.785.1...3.1.86.5. ED=11.8/1.2/1.2 ........1.......2.....13.45...3.56...371.....6.5.78....6.73...8.785.1...3.1.86.5. ED=11.8/1.2/1.2 30c ........1.......2.....13.45...3.56....5.78....3716.5...785.1...3.1.86...56.73...8 ED=11.8/1.2/1.2 ........1.......2.....13.45...3.56...371..5..6.5.78....785.1...3.1.86...56.73...8 ED=11.8/1.2/1.2

(all minimals of the same 32c)

Website · by **denis_berthier** » Sun Dec 26, 2021 8:05 am

mith wrote:
Code: Select all
29c ........1.......2.....13.45...3.56....5.78....3716.....6.73...8.785.1...3.1.86.5. ED=11.8/1.2/1.2 ........1.......2.....13.45...3.56...371.....6.5.78....6.73...8.785.1...3.1.86.5. ED=11.8/1.2/1.2 30c ........1.......2.....13.45...3.56....5.78....3716.5...785.1...3.1.86...56.73...8 ED=11.8/1.2/1.2 ........1.......2.....13.45...3.56...371..5..6.5.78....785.1...3.1.86...56.73...8 ED=11.8/1.2/1.2

(all minimals of the same 32c)

I didn't have much time, but I finally checked them: the four are in B6B (or lower).

Reminder: as of now, there are only 3 puzzles known to be in B7B (and none higher).

by **coloin** » Sun Dec 26, 2021 7:21 pm

denis_berthier wrote:Reminder: as of now, there are only 3 puzzles known to be in B7B (and none higher).

denis_berthier wrote:
..X....X..X....X.XX...........X.X..X.X.XX....X....X....X.X....X..X....X.....X.X.X ; eleven#3 ; B7B ; 11.9
.X...XX..X...X......X.......X.....X.X.X....X..X.X....X....X..X....X.X..X...X..X.X ; eleven#22 ; B7B ; 11.9
X.......X.X.X...X...X...X...X.X.........X.......X.X.X...X...X...X...X.X.X.......X ; Metcalf ; B7B ; 11.8

and on p 60 of this thread

denis_berthier wrote:I chose these 3 because they are the only (known) puzzles with BpB rating = 7.

Code: Select all: ..3....8..5....2.17...........5.8..6.9.12....8....3....6.9....5..4....7.....1.6.2 1 ~ elev;1;5;22; .2...67..4...8......9.......3.....7.5.8....4..1.3....2....9..5....6.1..3...2..6.7 2 ? 5.......9.2.1...7...8...3...4.6.........5.......2.7.1...3...8...6...4.2.9.......5 3 = m_b_metcalf

which are these in the file

Code: Select all: 12.3.....4.5...6...7.....2.6..1..3....453.........8..9...45.1.........8......2..7;11.90;11.90;2.60;elev;1;5;22; #1 ........1.....2.3...4.5.6...3.....72.4....8..6..8..9...7...1...4.56.....8..59....;11.80;1.20;1.20;dob;12_12_03;247964;22 ;#2 5.......9.2.1...7...8...3...4.6.........5.......2.7.1...3...8...6...4.2.9.......5;11.80;11.80;3.40;tax;m_b_metcalf;22;20; #3

Website · by **denis_berthier** » Mon Dec 27, 2021 3:25 am

Coloin, thanks for recalling this list. The original list of 3 appears in [PBCS1, 2011, tables 11.3 and 11.5]:

Code: Select all: ..3....8..5....2.17...........5.8..6.9.12....8....3....6.9....5..4....7.....1.6.2; 11.9 eleven #3 .2...67..4...8......9.......3.....7.5.8....4..1.3....2....9..5....6.1..3...2..6.7; 11.8 eleven #22 5.......9.2.1...7...8...3...4.6.........5.......2.7.1...3...8...6...4.2.9.......5; 11.8 Metcalf

(I dealt separately with eleven's list, because there was a clear description of how it was produced.)

The problem with champagne's collection of hardest is, it was a total mess from one version to the next: puzzles disappearing, being re-numbered (in a different file), renamed, morphed. That's why the version you give here may be morphs of those in PBCS.
It seems eleven#3 was a morph of dobxxx (but dobxxx was not in champagne's collection when I dealt with it for PBCS1). I have no idea of who found it first.

Anyway, the main result is, in spite of all the work done since that time (2011) for finding "hardest" puzzles, no new one in B7B has been found.

by **mith** » Mon Dec 27, 2021 7:24 pm

Code: Select all: 32c ..............1.23.24..5....3...674.64..273.88......62.78....3626..7.8.44.36...7. ED=11.5/1.2/1.2 (DCFC+DFC) ..............1.23.24..5....3...674.64..2.3.88.7....62.78....3626..7.8.44.36...7. ED=11.5/1.2/1.2 (DCFC+DFC)

Current max SE for minimals by clue count:

Code: Select all: 19c - 10.5 20c - 11.8 21c-23c - 11.9 24c - 11.8 25c - 11.7 26c-27c - 11.8 28c - 11.7 29c-30c - 11.8 31c - 11.6 32c - 11.5 33c - 11.3 34c - 11.1 35c-36c - 10.6

by **hendrik_monard** » Tue Dec 28, 2021 5:52 pm

I found some new 11.7''s and 11.6's. Most of them resulted from vicinity search on mith's list of November 3th.
These are the 11.7's:
98.7.....7..68......5..48...9..7..3...2.....5......4...3..6..8......7.6......15.2;23;11.7;11.7;2.6
98.76.5..7.5.498...645.....6...87....97..4.8....95.....4..............53.......21;27;11.7;1.2;1.2
987......65.7...........6....43..72....1...46....2.3.1...41..73....762.4.....31..;27;11.7;1.2;1.2
987......6..8.......5......5..48.32....3.5..4....26.58.6..34..2...5..48....2...63;28;11.7;1.2;1.2
987......6..8.......5.......6..432...5.2..46....5...83...4.53.....38..24....2685.;28;11.7;1.2;1.2
987......65..........76......4.732.....6.41.....12..3....43..76...2..41.....16..2;27;11.7;1.2;1.2
987......6..8.......5......5..48.32....3.5..4....26..8.6..345.2...5..48....2...63;28;11.7;1.2;1.2
987......6..8.......5.......6..4352....5..4.8...2...36...4.83.2...35..4.....62.85;28;11.7;1.2;1.2
98.76.5..7.5..9....64.5....8.....4.....8...5.......32.6...47...5..98.....79..6..4;27;11.7;1.2;1.2
98.76.5..7.4..8....56..9...8...96....9.8...5..6754........74.........4.3.......72;27;11.7;1.2;1.2
9876.....6..8.......5.......6..4.3...5.3.246....5...28...4.52.....28..43....3685.;29;11.7;1.2;1.2
98.76.5..7.54.9....46.8....6..8.7....97.46.....859......4....8.....7...5.......32;28;11.7;1.2;1.2
98.76.5..7.54.9.....6.8....6..8.7...4.859.....97.46.....4....8.....7...5.......32;28;11.7;1.2;1.2
98.76.5..7.58.4....46.9....6.........7958......46.9.7.5......32...4............51;27;11.7;1.2;1.2
9876.....6..8.......5......5..48.32....3.5..4....26..8.6..3..52...2.46.3......48.;28;11.7;1.2;1.2
9876.....6..8.......5.......6..4.53..5.3.2.46......8.2...4.5.2....28.3.4....36.85;29;11.7;1.2;1.2
98.76.5..7.5.......64.58...8............9..5.......32.6.9.47...5..98.....7...6..4;26;11.7;1.2;1.2
98.76.5..7.5.48....4.......87...4...6.9.5.....5498..7..9......3.....6.........9.2;26;11.7;1.2;1.2
98.76.5..7.5..8....64..9...6.84.7....7..56.4..5.9......9..........8....5......3.2;26;11.7;1.2;1.2
98.76.5..7.5.......64..9...6.84.7....7..56.4..5.98.....9..........8....5......3.2;26;11.7;1.2;1.2
98.76.5..7.5..9....648.....4..95.....796.4.8.....87.....6.....8...4............32;26;11.7;1.2;1.2
98.76.5..7.5.49....648.....4..95.....796...8.....87.....6.....8...4............32;26;11.7;1.2;1.2

The longer list of 11.6's are in the hidden list hereafter.

In order to be able to check novelty compared to mith's list and to the existing database ph_2010, I had to make a version of mith's list in which the minimal form that he uses is replaced by the canonical form (isomorph with highest numerical value after replacing dots by zero's).
It becomes quite complicated to have to work with two different systems. Personally, I prefer to work with the canonical form, but if there is a convincing reason to change format, I can adapt. In that case I shall have to develop a VBA script for calculating the other form. About a year ago, I already wrote a VBA script for calculating the canonical form.
Am I right to suppose that the minimal form is the isomorph with the lowest numerical value? If not so, is there a guide of some kind explaining the rules for the minimal form?

Hendrik

Hidden Text: Show

by **eleven** » Tue Dec 28, 2021 10:12 pm

Hi Hendrik,

congrats to the new puzzles, i know, how hard they are to find.

There are many ways to canonicalize puzzles and solution grids, but so far i can see it, only 2 survived. Champagne choose to use the highest numerical value, the others i know of the lowest. But i see no big problem to switch from one to the other.

Website · by **denis_berthier** » Wed Dec 29, 2021 5:29 am

.
I think it would be easier for everyone if a public software producing the "canonical form" used in champagne/(now mih ?) meta-collection was made available in a language that can run on any OS. There are so many canonicalisation programs around that one may already exist, but then it would be useful to have an explicit reference to it.

by **mith** » Wed Dec 29, 2021 6:53 pm

Denis/Hendrik, I use the minlex from gsf: http://gsf.cococlyde.org/download/sudoku

IIRC it may have some bugs and not always find the actual minlex grid but at least is consistent? There has been an effort recently on discord to canonicalize sudoku x grids, maybe a bug-free version for classic sudoku will come out of that at some point.

I’ll check and add Hendrik’s latest to my local database later today. In an ideal world I’d get a database up online with some sort of input mechanism so people can add their new puzzles and automatically check whether they are already in the database and insert them if not. Maybe a project for the new year.

by **mith** » Wed Dec 29, 2021 7:15 pm

Here's the output of my search on Hendrik's puzzles:

Hidden Text: Show

First set of ratings is SE, second set is skfr; at the moment I'm mostly running SE on 11.3+ skfr (the number of 11+ skfr puzzles I haven't rated yet is absurd). It's interesting how many hadn't been found at all; I'm guessing the first filter (gsf's q2) is skipping them (edit: checked the first two, and yes, they are unusually low q2 for such high SE puzzles).

That gap from 11.0 skfr to 11.7 SE is as big as I've seen, I think.

by **JPF** » Wed Dec 29, 2021 7:54 pm

hi Mith,

IIRC it may have some bugs and not always find the actual minlex grid but at least is consistent?

Yes that point was discussed here
Both gsf'program and dobrichev's gridchecker have the same bug: for some puzzles (not full grids) they don't give the minlex value.
But ... there are good reasons to believe that they give for any puzzles a canonalized form c ; i.e: c(P1)=c(P2) if and only if P1 and P2 are equivalent.

Note that gridchecker is much faster than gsf's.

btw, I saw 78 duplicates in your november 3rd list.

JPF

The New Sudoku Players' Forum

The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)

Re: The hardest sudokus (new thread)