The New Sudoku Players' Forum

[nedBlake]

Code: Select all

  *-----------*
  |.9.|.8.|6.2|
  |4..|3..|9..|
  |...|.57|...|
  |---+---+---|
  |.8.|...|.9.|
  |...|...|35.|
  |..1|..2|...|
  |---+---+---|
  |.42|...|7..|
  |..8|..3|...|
  |...|.7.|.23|
  *-----------*


[Leren]

Moderate is an understatement. Only singles were required to solve. FWIW here is the solution.

Code: Select all

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

Leren

[nedBlake]

Laren, Don't see how you got that.

I got:

Zero naked singles to start with

then a total of the following that were successful in exposing naked singles:

11 naked pairs
6 naked triples
2 naked quads

Possibly you did better by looking for hidden singles to begin with.
The way I went at it, I solved it without needing to look for any hidden singles, which I find time consuming.

[champagne]

Laren, Don't see how you got that.

I got:

Zero naked singles to start with

then a total of the following that were successful in exposing naked singles:

11 naked pairs
6 naked triples
2 naked quads

Possibly you did better by looking for hidden singles to begin with.
The way I went at it, I solved it without needing to look for any hidden singles, which I find time consuming.

Hi nedBlake,

Discussing what is easy to find and what is not can be an endless discussion.
However, it is generally agreed that, without pencil marks, hidden singles are easier to detect than naked singles.

Using an automatic building of pencil marks gives another view, but most players don't produce pencil marks.

[nedBlake]

Hi Champagne,

I agree that there really is no"best" way to do solve a puzzle. The question is what do most people do. From reading your post, I (or my grader) may be way wrong. The assumption was that most would copy or import the puzzle from this forum into some sort of program or App that at least had the ability to display possible solutions to a cell.

The grader as now designed would assign a grade by first trying to solve the puzzle by looking at block unit singles. This is relatively easy to do by looking at rows/cols crossing blocks. Finding unit singles in rows or cols I thought would be much harder to do - maybe I'm wrong about that.

If a solution is found, then the puzzle is graded as Easy (not that it may be all that "easy" to solve, and that's were the stars come in). It is assumed the user would go ahead and solve the puzzle without needing to display possible solutions.

If the solution was not found (as in the case of this puzzle), the user was given the option to display possibles. Now I imagine there is a lot of difference between people here. I tend to to look at the number of clues, and make decision based on that. The grader at least assumes at this point that they are displayed, in which case we certainly would solve naked singles first, then at least look for the smaller nTuples such as pairs, etc.

Now from what you say, you think that most people will solve puzzles (at least at this level - I can't imagine how they could solve some of the harder puzzles in this forum) without displaying possible solutions. You may be right - I really don't know. I was hoping to post puzzles at the Easy or Moderate level in this forum since I think it could use some variety, but my grader may be way off at least at the Moderate level, and the puzzles generated would be uninteresting or boring because they assumed they would be solved in a totally different way.

[JasonLion]

Out in the world many people use crosshatching to look for hidden singles. Using crosshatching without pencil marks is very very common for newspaper solvers, who either learn it on their own or from nearly all of the books on how to solve Sudoku puzzles. That crowd considers hidden singles in blocks the simplest thing to look for, with hidden singles in rows/cols not far behind. From that point of view, naked singles are difficult as they require either pencil marks or enumerating all possibilities for each cell till you find a cell with only one possibility, both of which are comparatively tedious.

On the forum we give hardly any notice to singles (naked or hidden, block or row/col).

For the puzzles that are posted on the forum, the default assumption is that you are looking for a solution using first Simple Sudoku basics, then one non-basic move, and finally singles to the end, 'stte', or singles and locked candidates to the end, 'lcstte'.

[Leren]

Yes, the solution I found included hidden singles (of which there were many). I doubt if many people would regard hidden singles to be harder to spot than naked tuples etc.

Just to check I put this puzzle into Hodoku and it solved it the same way as I did, naked and hidden singles, only the order being slightly different.

... And just to double check I tried it in Andrew Stuart's solver, which you can find here and it never got past naked and hidden singles either.

... And just to triple check I found a solver here and followed it's hints, which for the first few cells were, would you believe, yes, that's right, hidden singles. I gave up after a while.

Leren

[nedBlake]

Certainly solving this puzzle was very easy just looking at unit singles. After solving block units, there were so many clues that solving row/col unit singles was relatively easy.

I am changing my Grader accordingly (to solve all unit singles first, instead of just block singles), which would then put this puzzle into the Easy category. It does however make it harder creating the tougher Moderate puzzles . With the modified Grader, I was not able to generate any puzzles having the old Moderate *** requirements of number and difficulty of Moderate strategies needed to solve the puzzle, and so will have to weaken them to generate puzzles with this rating.

A Grader to be any good must solve a puzzle like most people would solve it, and so it is much harder to code than a Solver, which uses some computer-friendly algorithm. It is very important in the create process, since it not only grades a puzzle, but also provides information that lets you judge the quality of the created puzzle.

I am still not sure I know how "most people" solve a puzzle. From the posts on this puzzle, I was given the idea that people still first apply strategies that people used in solving a puzzle in the daily paper using a pencil. But is this true, particularly in this forum, if they can load the puzzle into an app that lets you display possible solutions in empty cells along with other visual aids.

[JasonLion]

I am still not sure I know how "most people" solve a puzzle.

There certainly isn't a single answer to this question. It breaks down into groups who use similar approaches within the group, but across groups there are some dramatic variations. There is also a fair sized collection of individuals who each have their own semi-unique, approach.

Usage of apps that figure pencil marks for you varies widely. Some here totally rely on automatic pencil marks, while others (including a fair number of very advanced solvers) avoid using automatic pencil marks. You will find a wide range of attitudes on computer assisted solving in general here, with a whole crowd that is into challenging their code writing abilities, while others take pride in solving puzzles in their head.

Historically there has been very little agreement about difficulty grading systems. When we simply have to use a grader we generally agree to use Sudoku Explainer, despite it's problems. But don't take that to mean that Sudoku Explainer is an especially good grader. It is more that it doesn't make the usual big obvious mistakes, as opposed to it getting everything right.

[eleven]

As Jason said, most people learned to know sudoku from the newspaper. The majority of those look for hidden singles (also hidden pairs and triples) first, because this doesn't need any pencil marks. A common way to continue is to write pencil marks with 2, then 3 candidates.

This is very different to the game played here.
Many of the puzzles here (the easier ones) can be solved on paper with 2 or 3 "advanced" moves like x|y|z-wing, skyscraper or unique rectangles. And this is, how newspaper solvers would do it. But here people try to find a solution with only one advanced move. So at least you should have a solver, which can tell you, if a move actually solves a puzzle with singles (if not, you would look for another move - but on paper i would never erase a correct move).

[nedBlake]

I have to agree with what all of you are saying in one way or another. Grading a puzzle based on "how most people would solve it" is first of all probably impossible because we just don't know what that is, and secondly it's just the wrong way to look at. We shouldn't base a grade on personal whims.

This puzzle should have been graded Easy, because it can be solved by the easiest strategy - looking for unit singles. The rating would at least be informative.
Because it would take 8 passes looking at all unit types, it would have 3 stars, and I would probably import it into an app and solve it as a Moderate as that sounds pretty tedious.