The New Sudoku Players' Forum

[jeanboucher]

Code: Select all

+-------------+-------------+-------------+
| 68  1   9   | 3   578 58  | 4   2   67  |
| 2   568 58  | 4   789 89  | 3   1   67  |
| 3   4   7   | 6   1   2   | 9   8   5   |
+-------------+-------------+-------------+
| 1   7   6   | 5   2   3   | 8   4   9   |
| 4   589 58  | 1   89  7   | 6   3   2   |
| 89  3   2   | 89  4   6   | 7   5   1   |
+-------------+-------------+-------------+
| 689 689 14  | 2   389 189 | 5   7   34  |
| 5   2   3   | 7   6   4   | 1   9   8   |
| 7   89  14  | 89  35  15  | 2   6   34  |
+-------------+-------------+-------------+


My solver give a hint at coloring, the problem is...
I don't understand coloring!!!
Any good Explanation or Tutorials?

Thanks

[Sped]

Code: Select all

 
 *--------------------------------------------------*
 | 68   1    9    | 3    578  58   | 4    2    67   |
 | 2    568  58   | 4    789  89   | 3    1    67   |
 | 3    4    7    | 6    1    2    | 9    8    5    |
 |----------------+----------------+----------------|
 | 1    7    6    | 5    2    3    | 8    4    9    |
 | 4  5(8)9  58   | 1    89a  7    | 6    3    2    |
 | 89   3    2    | 89A  4    6    | 7    5    1    |
 |----------------+----------------+----------------|
 | 689  689  14   | 2    389  189  | 5    7    34   |
 | 5    2    3    | 7    6    4    | 1    9    8    |
 | 7    89A  14   | 89a  35   15   | 2    6    34   |
 *--------------------------------------------------*


There are two 8s in row 9. One must be true for 8, the other false. Mark r9c2 "A" and r9c4 "a".

There are two 8s in column 4. One must be true for 8 and the other false. r9c4 is already marked "a" so mark r6c4 "A". Either all the "A" cells are 8 or all the "a" cells are 8.

There are two 8s in box 5. r6c4 is already marked "A" so mark r5c5 "a".

Since either the "a" or the "A" cells are guaranteed to be 8, cells that see both an "A" and an "a" cannot possibly be 8.

r5c2 sees the "A" in r9c2 and the "a" in r5c5. Since one of those cells must be an 8, r5c2 can lose its 8.

[jeanboucher]

Thank you Sped, I see it better now...

But, What make you choose these 8's ???
Is it because they have strong link?

[keith]

Maybe you need a new solver

Code: Select all

+-------------+-------------+-------------+
| 68b 1   9   | 3   578 58  | 4   2   67  |
| 2   568 58  | 4   789 89  | 3   1   67  |
| 3   4   7   | 6   1   2   | 9   8   5   |
+-------------+-------------+-------------+
| 1   7   6   | 5   2   3   | 8   4   9   |
| 4   589 58  | 1   89A 7   | 6   3   2   |
| 89A 3   2   | 89a 4   6   | 7   5   1   |
+-------------+-------------+-------------+
| 689 689 14  | 2   389 189 | 5   7   34  |
| 5   2   3   | 7   6   4   | 1   9   8   |
| 7   89a 14  | 89A 35  15  | 2   6   34  |
+-------------+-------------+-------------+

There is a remote naked pair <89> in the chain aA which eliminates <89> in R5C2.

Coloring on <8> in the same chain eliminates <8> as a candidate in R7C1.

Then, the coloring on <8> can be extended to R1C1, "b", which takes out <8> in R1C5.

There is a Unique Rectangle <58> in R25C23, and you can eliminate <8> from R25C2.

Keith

[jeanboucher]

I was reading another thread about Strong link "skyscraper"...

Can I do that?



Can I eliminate candidates r7-c1, r7-c2 & r5-c2 ?

[jeanboucher]

Wow man..., So many solutions!!!

I am using Sadman Sudoku ( I bought it mainly to learn)

Do you recommend any other application?

[keith]

This is a short coloring chain on <8>. One of R7C1 or R9C3 must be <8>. But, you cannot eliminate the <8> in R7C2. The other two are OK.

The true "fork" or skyscraper does not require the strong link in C4. Only the strong links in R7 and R9 are enough to say one or both of R7C1 and R9C3 are <8>.

I use Sudoku Susser. It does not do coloring. The manual is excellent. I think it does Unique Rectangles and fishy cycles as well as anyone. It does not do fins or sashimi fish. (Fish are X-wings, Swordfish, etc.) It does not do Turbot Fish, which includes your skyscrapers.

And, by the way, after you make the above eliminations, see if you can extend the coloring to eliminate an <8> in R1.

Keith

[ravel]

Do you recommend any other application?

I can recommend the Sudoku Explainer. In your example it shows a unique rectangle for 58 in r25c23 as the next hint, which allows you to eliminate 8 in r25c2 (and i think this is the easiest elimination to spot).
With "Get next hint" it shows turbot fishes (2 strong links) for:
- r5c2<>9
- r5c5=9
- r6c1=9
- r6c4<>9
- r5c2<>8 (without "filter hints with similar outcome", because eliminated by the UR too)
- r7c1<>8
- r7c1<>9
- r7c5<>9
- r7c6<>9
- r9c2<>9
- r9c4=9
Most eliminations lead to the same result.

More advanced colorings (3 or more strong links, multiple or grouped coloring) are resolved with "X-chains".

(One drawback: you cannot "step back")

[udosuk]

Personally I think the easiest move is the "remote naked pair" of {89} in r6c14+r9c24 which forces r5c2=5 and solves the puzzle immediately...

[ravel]

Agree, thats what i would have spotted first, when solving it manually. This shows how blind we can become for obvious things when following a special solver's strategy.

[ronk]

I see "remote naked pairs" as a special case of colouring. One is coloring two digits simultaneously instead of just one. They're easier to spot, but you still have to follow the chain to assign true/false parity to [ed: the] cells involved.

[Len]

A blue collar approach would be
r6c1=8=>r6c4=9=>r9c4=8=>r9c2=9=>r7c1=6
else
r6c1=9=>r6c4=8=>r9c4=9=>r9c2=8=>r7c1=6
This value of r7c1 solves it directly.

[jeanboucher]

Thank you so much to all of you guys!!!

udosuk, you have spot the easiest answer...

I wish I had some kind of a hierarchy list of steps to solve sudoku.

I may have stared at this grid for an hour at least.
At the end I've quit & ask for help.

[udosuk]

udosuk, you have spot the easiest answer...

keith was the one who spotted that move... Just that he included 3 other moves afterwards, which are all unnecessary...

Personally I think ONE good move to reduce a puzzle to singles is quite enough... Life is too short for everyone to find every available move from every possible position of every puzzle...

BTW I used to always dig the "white collar" approach (no forcing chains, everything pattern-based) but that proves too powerless against the monsters (such as the "Ocean's 12" or however many there are now )... I think the "blue collar" approach is the only way to go these days...