The New Sudoku Players' Forum

[m_b_metcalf]

The puzzle in 729-char format and the first SE solution steps.

Thanks for another test puzzle. It took my solver just over 8 minutes to find the solution:

Hidden Text: Show

Code: Select all

  1  2  3  4  5  6  7  8  9
  4  5  7  1  8  9  2  3  6
  6  8  9  7  3  2  1  5  4
  2  9  4  6  1  8  5  7  3
  3  7  1  9  4  5  8  6  2
  5  6  8  3  2  7  4  9  1
  7  3  2  8  6  4  9  1  5
  8  1  5  2  9  3  6  4  7
  9  4  6  5  7  1  3  2  8

Regards,

Mike

[dobrichev]

One that starts with ED=12.9 elimination. Minimal with 316 restrictions (high-clue).

Code: Select all

1234.67...2345.....23.567.9...4..7891.3456...1.3..678.12..5678.12...6.89....567.9..345...91...567.....4.678.12..567..12..5..8..234..7.91.34.678912.45..8..23..6.89..3..67...23.5..8.12...6789123..6.89.23..67.9...45.789123456.8...3456.89123456789123...78...345...912....7.912..567.91.34.678....45678912.45.7891.3.5...91.34.6789.23.567891234...8...345.7..123..6.8....45...912.45.7.9.2.4.67....3.5.789.23.5.7...2.4..78912.4..7.9..3.56...1.345678.1.3...789123.56789..3.56.8....4.678.12..5.789.23..67..12.4..78..2..5..89.23..678.1...56..9.234..78.1234.6..9.234567.912.456.8.12.45..8.12.4.6789.234..78..234..78912..5.78.12...6.89.2...6......4.67....3.5.789.2..567891.345678.12.4......23456..912.4..789..34.6..91..4.678.1.3...7.9.2.4.6.89


[tarek]

One that starts with ED=12.9 elimination. Minimal with 316 restrictions (high-clue).

Excellent find … We are getting there!!!

[m_b_metcalf]

One that starts with ED=12.9 elimination. Minimal with 316 restrictions (high-clue).

Hidden Text: Show

Code: Select all

  1  2  3  4  5  6  7  8  9
  4  5  7  1  8  9  3  2  6
  6  8  9  3  2  7  1  5  4
  2  3  1  6  4  5  8  9  7
  7  4  5  8  9  1  6  3  2
  8  9  6  7  3  2  5  4  1
  3  1  8  2  6  4  9  7  5
  5  7  4  9  1  8  2  6  3
  9  6  2  5  7  3  4  1  8

Total time 270.45 seconds

[dobrichev]

Hi Mike,

Are your timings from brute force solver or from one that emulates "human-style" solving?

[Mathimagics]

My P&P style emulates a brute-force solver ...

[m_b_metcalf]

Are your timings from brute force solver or from one that emulates "human-style" solving?

As this starts at 12.9 my program goes straight into brute force, which checks for uniqueness (so doesn't stop when the solution is found).

[dobrichev]

@Mike: If your solver has mode "ok, human techniques don't work, but I am not a human" then there is room for improvement. My solver takes 0.009 seconds to solve this puzzle up to the second solution, and finds first backdoor of size 4 exemplar after 0.206 seconds - after ensuring all possible combinations of 1, 2 and 3 backdoors fail.

@Jim: P&P brute force has been demonstrated to work here. Respect!

[m_b_metcalf]

@Mike: If your solver has mode "ok, human techniques don't work, but I am not a human" then there is room for improvement. My solver takes 0.009 seconds to solve this puzzle up to the second solution, and finds first backdoor of size 4 exemplar after 0.206 seconds - after ensuring all possible combinations of 1, 2 and 3 backdoors fail.

Yes, strange. It takes 0.12s to determine no human method works, and then 0.02s to find the first solution. Why it takes so long to find no other solution I shall investigate.

[dobrichev]

Here is one simpler (starts with ED=12.8 and has backdoor of size 3 in singles + locked) but with even more givens - 320. Previous has BD 4.

Code: Select all

1...5...9123..6..9..345678.1234.6.8.1..45.78..2..56.8.1.3...78912...678912.4.6.89...4.6.89..3.5.7.912...678.1..456..91..45678.1.34..7.9.234....912...6.....3.567...23456..91....6.8.1..4567891.345.7.9.234....9123.5.7891....6..9....5..8912.4567.912345.7..123.567.91234.6.891.34567.912345.7.9.2.45...91....6789.234....91..4.6789.23..678..2.45..8.1..456789.....6.89.2.456..91.3..678.123456..9.23.56789.23.56..9123.56789...456.89..3..67.9.2.4..78.1.3.56...12..5.789...456.8..234..78912..5.78...345678912.4..78.12..56.89123.5.78.123456.89..34..7891....6.89..3.567.....45..8....45.789.23.5678...3456789.2.4.6..91.3..67..1....6.8..23.5..8.1..45678...3.5...91.345.7.9123..678..2.4...89..345..........7.91.3.567..12.45.78.123..67..1.....78.


It has the same solution - Platinum trellis.

[Mathimagics]

@Mladen: Thanks! I like to think that fsss2 and I have special relationship, that we are sympatico ...

But it wasn't difficult only because I happened to be very lucky, I was just settling down to try in turn an estimated 350 trials, and hit the answer on trial 37, so could have spent ten times as long solving it.

I will definitely have to check that puzzle out!

There's a standard "How to Solve Sudoku" page at the front of the "Sudoku Master" mags that I sometimes buy. It includes this:

Most of the people who have bought this book are probably "logic" solvers of Sudoku. That is, you don't write a number in a square until you are absolutely certain it's the correct number.
Logic solvers work with pens, trial-and-error solvers need pencils and erasers.

Ok, them's the rules. We divide solvers into two groups, the clever sophisticated ones, and the bogans. The puzzle is, who belongs to the first group?

It's unclear whether the first group are permitted pencilmarks at all. One can record them with a pen, although surely that has to get a bit messy, what if the pen leaks?

Oh dear, hang on a sec, that was a dead end that I picked (darn!). I missed an obvious elimination - when you record pencilmarks you are literally writing numbers in the square. Ok? So we can pencil (you know what I mean) in that "pencilmarks permitted?" is NO. Got it!

Now let's see, hmm ... there is still ambiguity. How do we (in the first group) deal with numbers that are OUTSIDE the puzzle area. Tricky ... gosh, I'm really struggling now, these specs are pretty vague! Does it have multiple resolutions?

Oh, now I really do give up, because it gets worse - what about lines, squiggles, the circling of squares, doodles, etc ... if you can't use them in the puzzle area then that makes pencilmarks outside the area more challenging. But "we" are clever people, so maybe that's possible ...

Maybe you have to be in "clevers" group to understand the rules of the group?

[tarek]

There is the backdoor size and then there is the number of backdoorSize sets.

If your BD = 2 puzzle has 100 pairs of cells that could crack the puzzle with singles if guessed correctly. That could be tougher for somebody who guesses for a living than a BD = 2 puzzle with only 1 pair.

But it wasn't difficult only because I happened to be very lucky, I was just settling down to try in turn an estimated 350 trials, and hit the answer on trial 37, so could have spent ten times as long solving it.

Still is amazing no matter how many times I see this.

[dobrichev]

This is more a religion than a science.

The only definition close to "logic" that makes sense to me is

It's very difficult to formulate a precise definition of a pattern, but one essential is that they should be capable of being identified without having to track any logic streams but just by counting and checking the defined elements exist in a qualifying distribution.

Even this definition can be (weakly) applied to the referred above P&P method, classifying it as "logical"

You author forgot to mention

Most of the people who have bought this book are probably "logic" solvers of Sudoku. That is, you don't write a number in a square until you are absolutely certain it's the correct number.
Logic solvers work with pens, trial-and-error solvers need pencils and erasers as well as tools like electronic devices and other technology achievements.

[dobrichev]

There is the backdoor size and then there is the number of backdoorSize sets...

Actually mollwollfumble is systematically examining a whole house / digit template configuration at once which is not as trivial as recursive backdoor hitting.

[dobrichev]

One puzzle making SE lost even before first elimination. Backdoor 3, 326 restrictions.

Code: Select all

1234..7..123.5.7...23.567.912.4...891.3456....234.6.8..2..5.78.12...6.8912.4567.9..345...9....567..1..4..78.1...567..1.34...8..23...7.91.34.6..912..5.7891.3..6789..3..67...23....8.1.3.5.789123..6789.2...67....345.78912345..8...3456.891..4.6.89123..678...345...912...67.9.2.4567..1.345678.1..456.89..3.5..891.3.56..9..34.6789.2345678912345.78...345.7..123..6.8.12..5...912.45.7.912.4.678...3.5.789.23...7...2....78912345...912.456.8.1.34..7...23..678912345678..23.56.8.1..4.67..12..5.....23..67..1..4..78..2.....89.23.5678.12..56..91234..7891234....9.2.45.78912.456.8..2.45..8.12...67891.34...8..234..78912.4567.912.....89.2...678.1..4.678...3.5.7.912..56.89...4567..12...67...23.5...912.4567....34.6.891..4.678.1.3.........4...89