The New Sudoku Players' Forum

[ab]

I seem to remember a statement from Nikoli cliaming that computer generated puzzles were deficient because you could fill in all the 5s then all the 7s etc. Here's the Guardian's 'hard' puzzle yesterday:

Code: Select all

 7 3 . | . 2 8 | . . .
 . . 9 | 1 . . | 4 . .
 . . . | . . . | 6 . .
 ------+-------+------
 . . 5 | 4 . . | . 9 .
 . 2 . | . . . | . 3 .
 . 9 . | . . 5 | 2 . .
 ------+-------+------
 . . 6 | . . . | . . .
 . . 8 | . . 1 | 5 . .
 . . . | 3 7 . | . 2 6



i started by filling in all the 3s and all the 2s! Also how can they give that puzzle a 'hard' rating??

The previous 2 days puzzles were rated medium and they were more of a challenge. Having tried them I was beginning to believe those who claim that handmade puzzles are different. I'm beginning to think it's just hype, maybe so they can sell their puzzles at a premium. A well programmed bit of software can generate puzzles that are just as good, if not better.

[lunababy_moonchild]

I think Nikoli's hand constructed puzzles 'feel' different - they do to me - but I would never say that they are better, or harder. Then again I've only ever done Nikoli's in Carol Vorderman's Massive book.

I think that puzzle generation is a matter of taste.

I know I certainly like to change the puzzle setter, now and again, for a bit of variety but I wouldn't say that I like either CG or HC better than the other. I do both. Then again I solve all of my puzzles on paper so perhaps that makes a difference.

Luna

[tso]

I'm in the process of putting together a post of puzzles, half of which will be computer generated, half hand-made, but with the same clue placements and similar ball-park difficulty. We can have a non-scientific test to see to what degree solvers can differenciate them.

I'm biased, but I do think there are differences that are apparent to long-time solvers, not only between machine versus human, but possibly between human and another. Not that bridging this gap isn't possible, only that it hasn't been bridged yet. Like all Turing tests -- with good programming, a computer *can* make music or write poetry that could be indistinguishable from man-made -- I'm sure the bar is much lower with Sudoku.

If there *are* objective differences, that doens't imply better or worse, only different.