The New Sudoku Players' Forum

[koushanejad74]

A more challenging Blocku

Blocku_10_7_5_Problem.pdf

(162.22 KiB) Downloaded 141 times

Rules and explanations available at the following URL:
http://forum.enjoysudoku.com/blocku-part-ii-t36323.html

[tarek]

Hi koushanejad74,

As you are posting regularly now (and I hope this continues), may I ask you to try to post your future puzzles using characters to make it easy to copy/paste (For instance like Leren demonstrated in a different thread using X to show the black cell, 0 or . to show an empty white cell).

If there is an image, it would be also better to hotlink it using a service like imgur.com which you may find easier when you post the same image across several message boards.

this ultimately would achieve saving as much space as possible on the server/service that Jason has kindly used to host this forum

thanks,

tarek

[koushanejad74]

Hi koushanejad74,

As you are posting regularly now (and I hope this continues), may I ask you to try to post your future puzzles using characters to make it easy to copy/paste (For instance like Leren demonstrated in a different thread using X to show the black cell, 0 or . to show an empty white cell).

If there is an image, it would be also better to hotlink it using a service like imgur.com which you may find easier when you post the same image across several message boards.

this ultimately would achieve saving as much space as possible on the server/service that Jason has kindly used to host this forum

thanks,

tarek

Hi Tarek,

Thanks a ton, will do as you said, just a quick question: every url has a text and hyperlink, using the "URL" button how do I achieve that?

[tarek]

Each message board may have different rules but most will allow BBCode

I keep it simple keeping the text and hyperlink the same (make sure it has the http:// at the start

Code: Select all

Example:
[url]http://forum.enjoysudoku.com[/url]

If you need to do a text that hyperlinks then do the following

Code: Select all

[url=Hyperlink]TEXT[/url]

For instance this text links to the new players forum

tarek

[Edit: After consulting the BBCode manual I've corrected the advice]

[Leren]

I'm cobbling together a bit of a Blocku Solver based on my Str8ts solver. It managed to completely solve this latest puzzle using only the compartment interaction rules I outlined in the previous thread.

Leren

[Wecoc]

I tried your puzzle.

Code: Select all

68..7.X...
.......X..
.3.2..6..X
....X.....
..X....67.
..9X2.....
.6...X2...
......5.X.
X.7.......
.X.......8

Solution: Show

Code: Select all

684579X132
9534867X21
531297684X
3421X58967
21X3489675
879X241356
76895X2413
42683157X9
X976123584
1X57634298

It was still quite easy in my opinion.
As I said in the other topic, I think not giving all the X from the start would make more interesting puzzles.

You can try this one:

Code: Select all

.1....
5.....
..3..5
.5....
.....1
2...5.

[koushanejad74]

I'm cobbling together a bit of a Blocku Solver based on my Str8ts solver. It managed to completely solve this latest puzzle using only the compartment interaction rules I outlined in the previous thread.

Leren

Leren,

Then I need to make it more difficult, working on it

[koushanejad74]

I tried your puzzle.

Code: Select all

68..7.X...
.......X..
.3.2..6..X
....X.....
..X....67.
..9X2.....
.6...X2...
......5.X.
X.7.......
.X.......8

Solution: Show

Code: Select all

684579X132
9534867X21
531297684X
3421X58967
21X3489675
879X241356
76895X2413
42683157X9
X976123584
1X57634298

It was still quite easy in my opinion.
As I said in the other topic, I think not giving all the X from the start would make more interesting puzzles.

You can try this one:

Code: Select all

.1....
5.....
..3..5
.5....
.....1
2...5.

Wecoc,

Thanks for your reply,

1- your suggestion is great, I'm working on it. based on you suggested, all the blocks are not given and the user has to find numbers as well as blocks,
2- I didn't quite get what you meant by the sample you sent me, could you please elaborate? did I get it right?

[SpAce]

I solved this to the end without candidates. In other words it was either easier than the previous ones, or my routine has improved, or both. I did use one X-Wing, though, but I would guess it was not required at all (but possibly made for faster progress by solving a key cell easily).

Anyway, I think you need a better way to evaluate the difficulty of your puzzles. If I understood correctly, your solver doesn't really use logical solving methods but guessing, and then you rate a puzzle based on how many guesses it needed. That doesn't sound like even close to an accurate method to determine the perceived difficulty by a human player.

A more useful rating is based on the most difficult logical technique required (like what Sudoku Explainer does). Another related measure is the summed value of all of the techniques required to solve the whole puzzle, which tells more about the total tediousness than the maximum difficulty (Hodoku and SudokuWiki use that). A third measure would be the narrowness of the solve path, but I don't know if anyone really rates that.

I guess you don't currently have any idea of those kinds of ratings for your puzzles, if you don't have a solver that uses logical techniques only. That's why your announced ratings don't seem to match reality.

[Leren]

Code: Select all

68....X...
.......X..
.3.2.76..X
....X.....
..X....67.
..9X2.....
.6...X2...
......5.X.
X.7.......
.X........

Here is my first attempt at a non-trivial Blocku puzzle. Leren

[SpAce]

As I said in the other topic, I think not giving all the X from the start would make more interesting puzzles.

You can try this one:

Code: Select all

.1....
5.....
..3..5
.5....
.....1
2...5.

I tried. It wasn't easy, but it was doable. Or so I thought until I got stuck. Are you sure it has a single solution? If so, then I've made a mistake or fail to see some constraint. Either way, I get two solutions.

Here's where I got stuck, and the two solutions I see:

Hidden Text: Show

Code: Select all

x 1 . . 3 4
5 x . . 2 3
1 2 3 x 4 5
4 5 x 3 1 2
3 4 . . x 1
2 3 . . 5 x

The two solutions:

Code: Select all

x 1 2 5 3 4
5 x 1 4 2 3
1 2 3 x 4 5
4 5 x 3 1 2
3 4 5 2 x 1
2 3 4 1 5 x


x 1 5 2 3 4
5 x 4 1 2 3
1 2 3 x 4 5
4 5 x 3 1 2
3 4 2 5 x 1
2 3 1 4 5 x


So, is it me or is it you?

[SpAce]

Here is my first attempt at a non-trivial Blocku puzzle. Leren

That was more fun. I used an S-Wing for the coup de grâce. I also used a couple of X-Wings and a UR earlier, but I doubt they mattered at all, so it probably would have worked as an OTP move. Can't know for sure, though. The solution seemed kind of familiar

[Wecoc]

Are you sure it has a single solution? If so, then I've made a mistake or fail to see some constraint. Either way, I get two solutions.

This forum needs a big fat facepalm emoji for people like me
You are right, there's a disambiguity between 25s and 14s in c3 and c4, I have no idea how did I miss that.
The good thing is if you set the 4 in r2c4 as a given, for example, all the "blocks placement" part is pretty much the same, which is the part I was focusing on. It just makes more trivial the "numbers placement" part.

I didn't quite get what you meant by the sample you sent me, could you please elaborate? did I get it right?

It was just an example on a 6x6 grid (numbers 1-5 and block in each row/col) that didn't include any block from the start, but I messed up. Anyways it was just a fast example on how would I try to make some challenging puzzles.

Another way to approch this would be considering the block is like a "9" in a default sudoku grid (9x9).
That way you can also include the 3x3 blocks as an extra constrain. That being said, I'm not sure a sudoku with those restrictions is even possible.

I tried to adapt my solver to check that, and as far as I tried I didn't find a valid solution for a Sudoku with those constrains.

In a set of X continuous white cells A-B >= X is forbidden

That's correct. I'll post here some pseudo-code (the comments from my code) so you can see the steps I used to implement this.

Comments: Show

Code: Select all

Check validity of the current grid

- Find each placed X (here represented as a 9)
- Find each group of cells (each side of the 9: down, left, right, up)
- Continue if the group is not empty
- If there's only one cell in the group (and it's solved) it should be either 1 or 8
- If there's more than one cell, get the distance between the higher & lower solved numbers in the group
   [ If it's higher or equal than the number of cells, the puzzle is broken ]

Remove candidates based on solved cells

- Find each placed X (here represented as a 9)
- Find each group of cells
- Continue if the group is not empty
- Get all possible candidates in that group based on its size
   Example: 123 & 678
- Check the solved cells of the group
- If a solved cell invalidates one of the possibilities,
  then it will be also removed in the unsolved cells
   Example: 123 & 678, a solved cell is 6 therefore the others are 78
- Remove in each cell of the group all non-valid candidates
   (In the example, all except 678)

Remove candidates based on unsolved cells

- Find each placed X (here represented as a 9)
- Find each group of cells
- Continue if the group is not empty
- Get all possible candidates in that group based on its size
   Example: 123 & 678
- Check the unsolved cells of the group
- If an unsolved cell invalidates one of the possibilities,
  then it will be removed in all unsolved cells including itself
   Example: 123 & 678, a cell has candidates 4678 therefore it's 678
- Remove in each cell of the group all non-valid candidates
   (In the example, all except 678)

[SpAce]

This forum needs a big fat facepalm emoji for people like me
You are right, there's a disambiguity between 25s and 14s in c3 and c4, I have no idea how did I miss that.

Don't sweat it. I'm just happy I didn't make a mistake right after bragging about my almost-zero error rate Fortunately the problem appeared so late that it was possible to see what it was and solve it anyway.

However, I did almost lose my work because I tried to use a UR a bit earlier, which resulted in a no solution state. That's when I knew something was off -- either in my solving or in the puzzle. I'm glad it was the puzzle. Yet I would have been pissed if I'd really lost my work because of a bug, so please try to avoid in the future. There are no savepoints or unlimited undos in p&p solving. Luckily I saw the problem quickly enough so I could still salvage the grid.

The good thing is if you set the 4 in r2c4 as a given, for example, all the "blocks placement" part is pretty much the same, which is the part I was focusing on.

And I think you did a good job in that. The blocks placement was the most interesting part in the puzzle. The idea is definitely worth pursuing.

It just makes more trivial the "numbers placement" part.

More trivial? That would imply that the original was trivial to begin with. Did you try to solve it manually yourself? I wouldn't call any of it exactly trivial, but you're right that the numbers placement was the easier part -- after the blocks were solved. Too bad I didn't write all of my steps down, so I can't remember the exact solving order, but I know I didn't solve many numbers before I had all of the blocks placed.

Anyway, all in all, this puzzle was definitely the hardest so far despite being the smallest, and I don't mean the buggy ending. (I haven't tried the fixed version so I don't know how much easier it got after that.) Leren's puzzle was a walk in the park compared to it.

Added 1. I now solved the fixed version as well. It felt much easier, though I don't know how much of it was due to the added clue. No chains needed like the first time, just pretty basic deductions using the compartment rules. I guess I should solve the buggy version again to see if I just missed some easy deductions with that, or if it actually was significantly harder.

Added 2. I solved the buggy version again. It was definitely harder than the fixed one, but not nearly as hard as it felt the first time. As I suspected, I'd missed some relatively easy compartment deductions. No chains needed after all.

[Leren]

Code: Select all

....X2..5.
7.....X...
..X7......
.......X..
.X5.4.....
...X...5.6
4....7...X
2.3.6...X.
X.....2...
.....X....

Another non-trival Blocku. Leren