The New Sudoku Players' Forum

[FLOCKMAN]

Thank you for your swift reply

I am looking for the 10 17-puzzles with clue frequencies of 222222221.

Previous older posts said there were only 5. Do you think that there may be any more?

[coloin]

Code: Select all

........1.......23..4.56......7....8.5...4....1.3.......7........82.....9.....6..
..............1..2..3.....4.......5.....6..7.21...8......3......4....6.98..5.7...
..............1..2.34....5.......36...7....4.8....9......67....2........19......8
..............1..2.34....5........4...6...37.8....9......67....2........19......8
................12..3..4................35..678.....9......64.5.7.......91.8.....
.................1..2..3.4........5.....1.....6..7...8.....6..7..5..8...2.43.9...
........1........2..3..4........1.......25...67.....8.....7.93...5....4..8.6.....
........1........2..3..4........253..6.....4.7...8........6..97...8.....1.5......
..............1.23..4.56.........4.7...8....958..........2.......7...6...3.9.....
........1.....2.....3.....4.......5....67..3..8..4.........8....9..51...6.....29.

I very much doubt there are any more .......
Not easy to confirm perhaps .......
C

[FLOCKMAN]

Thank you for your comprehensive reply. I will now work on them to see similarities and differences.

[mith]

The subject of 17 clue puzzles came up today elsewhere; I was able to find an archive snapshot of Gordon's list, with 49151 puzzles: https://web.archive.org/web/20121231092 ... n/sudoku17

I remember seeing somewhere that there are a few more beyond this list - are they catalogued anywhere?

[JPF]

Here is a list of 49157 puzzles on dobrichev's site.

JPF

[Serg]

Hi, mith!
This is addendum to dobrichev's list: 17-clue puzzle #49158 discovered by Blue in 2019.

Serg

[JPF]

Oh! I missed this one!

JPF

[mith]

Thanks to you both!

[champagne]

Hi, mith!
This is addendum to dobrichev's list: 17-clue puzzle #49158 discovered by Blue in 2019.

Serg

49158 is the current list of known. Blue estimate is that 3/4 more could be discovered. The full scan of the solution grids started 2 years ago.
It has been verified that all 17 clues puzzles having at least one band with more than 6 clues are known.
The scan for the distribution 665 is ongoing, about half way. No new 17 showed up so far.

[Mathimagics]

I did send an email to Gordon Royle last year, with the 7 puzzles that were not in his list, and he replied "thanks".

But it seems he has not updated his list ...

[dobrichev]

Hi Mathimagics,

Yo sent me message to update my list last year, and today I shamefully updated it to all 49158 puzzles

[Mathimagics]

You were next on my list!

[MartinCHarvey]

Incidentally, I had a moment to put the 49 thousand known 17 clue sudokus through my grid checker, which determined that they had 33 thousand essentially different grids.

This would tentatively suggest that the 49 thousand known minimal sudokus are but the tip of the iceberg.

[JPF]

To the captain of the Titanic : we got your message here too.
You will find some feedback.

More seriously, you are pointing out an interesting point.
The search for 17s was in its principle copied from the search for 16s: by examining UAs in the entire catalog of grids.
Another approach could consist in the study of the 17-clue patterns likely to contain a valid puzzle.
One can legitimately wonder why some 17-clues patterns contain 36,34,...valid puzzles while the vast majority contain none.
The positive point is that once the pattern is known, it is easy to find the valid 17s it contains.
The difficulty, unresolved to date, is to find the right patterns out of 55,113,078,988...

JPF

[MartinCHarvey]

To the captain of the Titanic : we got your message here too.
You will find some feedback.

More seriously, you are pointing out an interesting point.
The search for 17s was in its principle copied from the search for 16s: by examining UAs in the entire catalog of grids.
Another approach could consist in the study of the 17-clue patterns likely to contain a valid puzzle.
One can legitimately wonder why some 17-clues patterns contain 36,34,...valid puzzles while the vast majority contain none.
The positive point is that once the pattern is known, it is easy to find the valid 17s it contains.
The difficulty, unresolved to date, is to find the right patterns out of 55,113,078,988...

JPF

I have never captained a ship, and it's not sinking!

Yes it is a good point. My suspicion is that the clues have to be evenly balanced across the constraint space, and that, following the UA approach, you could show with an initial set that even if all subsequent choices added a maximal amount of information/ ruled out as many degrees of freedom as possible satisfying subsequent constraints, that there were still degrees of freedom left.

It probably *is* possible to do it this way round, but I haven't had the time. Maybe in the winter when I'm playing less golf.