View previous topic :: View next topic 
Author 
Message 
Marty R.
Joined: 12 Feb 2006 Posts: 5446 Location: Rochester, NY, USA

Posted: Mon Mar 13, 2006 1:28 am Post subject: Stuck 


This is one I think I should have solved, based on previous puzzles with the same rating from the same source. The first time I solved all but four cells, but they formed a rectangle and all four contained the same pair. I figured the odds were much greater that I had made an error rather than this being a nonunique puzzle. The second time I reached an impasse much earlier. A unique rectangle and XWing were of minimal help.
Code:  
3 2 6 7 19 4 89 5 189 
149 1459 8 3 1569 569 7 16 2 
159 7 19 569 8 2 4 136 1369 

8 69 3 2 569 569 1 7 4 
149 1469 5 8 469 7 3 2 69 
2 469 7 469 3 1 569 8 569 

7 35 4 1 2 56 568 9 3568 
6 1359 2 459 7 8 59 134 135 
159 8 19 4569 4569 3 2 146 7 
 


Back to top 


keith
Joined: 19 Sep 2005 Posts: 3253 Location: near Detroit, Michigan, USA

Posted: Mon Mar 13, 2006 2:15 am Post subject: 


Marty,
From the position you posted, R8C7 must be <5>. Look at B9.
Then, another <5> is pinned, and then there is an Xwing on <5>.
Keith 

Back to top 


Steve R
Joined: 24 Oct 2005 Posts: 289 Location: Birmingham, England

Posted: Mon Mar 13, 2006 4:30 am Post subject: Stuck 


Interesting puzzle, Marty.
Is it possible to post the starting grid?
Steve 

Back to top 


Marty R.
Joined: 12 Feb 2006 Posts: 5446 Location: Rochester, NY, USA

Posted: Mon Mar 13, 2006 5:57 pm Post subject: 


Keith, I'm looking at B9 and can't reason out why r8c7 needs to be a "5".
Steve, the original:
Code:  3675
32
7824
214
53
271
4129
68
8327 


Back to top 


David Bryant
Joined: 29 Jul 2005 Posts: 559 Location: Denver, Colorado

Posted: Thu Mar 16, 2006 2:28 am Post subject: This one is pretty tough 


Hi, Marty!
From the position as you posted it, the reason why r8c7 = 5 should be pretty clear  you have the possibilities listed as {5, 9}, and there's a "9" at r7c8. Unfortunately, that's not enough to make much of a dent in this very tough puzzle.
Code:  
3 2 6 7 19 4 89 5 189 
149 1459 8 3 1569 569 7 16 2 
159 7 19 569* 8 2 4 136 1369 

8 69 3 2 569 569 1 7 4 
149 1469 5 8 469 7 3 2 69 
2 469 7 469* 3 1 69 8 569 

7 35 4 1 2 56 68 9 368 
6 139 2 49 7 8 5 134 13 
159 8 19 4569* 4569 3 2 146 7 
 
You can reason to the conclusion that r7c6 = 5 as follows.
 There are only three places to put a "6" in column 4.
 If r3c4 = 6 then r2c8 = 6, the "6" in box 9 must lie in row 7, and r7c6 <> 6.
 If r6c4 = 6 then r5c9 = 6, the "6" in box 3 must lie in column 8, r7c7 = 6, and r7c6 <> 6.
 If r9c4 = 6 then r7c6 <> 6.
So r7c6 = 5, and that should be enough to get you rolling. dcb 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3253 Location: near Detroit, Michigan, USA

Posted: Thu Mar 16, 2006 4:00 am Post subject: Explanation 


Marty,
R8C7 must be <5> because R7C8 is <9>.
And, R6C9 is the only possible <5> in R6 and / or C9.
And, there is an Xwing in C1 and C4 that says R9C5 is not 5.
This clarifies the position you posted, but by no means solves the puzzle. See David's message for a hint on further progress.
Keith 

Back to top 


Marty R.
Joined: 12 Feb 2006 Posts: 5446 Location: Rochester, NY, USA

Posted: Thu Mar 16, 2006 6:28 pm Post subject: 


As always, thanks to both of you. I don't know how I could stare at that puzzle and not see the "9" in r7c8.
Quote:  And, there is an Xwing in C1 and C4 that says R9C5 is not 5. 
I can't for the life of me find that XWing.
Quote:  You can reason to the conclusion that r7c6 = 5 as follows.
 There are only three places to put a "6" in column 4.
 If r3c4 = 6 then r2c8 = 6, the "6" in box 9 must lie in row 7, and r7c6 <> 6.
 If r6c4 = 6 then r5c9 = 6, the "6" in box 3 must lie in column 8, r7c7 = 6, and r7c6 <> 6.
 If r9c4 = 6 then r7c6 <> 6.
So r7c6 = 5, and that should be enough to get you rolling. dcb 
As I've mentioned before, deductive reasoning is a major weakness. How did you home in on the "6s" in c4 to start that line of reasoning? 

Back to top 


David Bryant
Joined: 29 Jul 2005 Posts: 559 Location: Denver, Colorado

Posted: Thu Mar 16, 2006 7:35 pm Post subject: Spotting "template" patterns 


Marty R wrote:  How did you home in on the "6s" in c4 to start that line of reasoning? 
Well, it was a combination of luck and experience.
Just to be clear about it, here's what my copy of the puzzle looked like when I had to start hunting real hard.
Code:  
3 2 6 7 19 4 89 5 189 
1459 1459 8 3 1569 569 7 16 2 
159 7 19 56 8 2 4 136 1369 

8 69 3 2 569 569 1 7 4 
149 1469 5 8 469 7 3 2 69 
2 469 7 469 3 1 69 8 5 

7 35 4 1 2 56 68 9 368 
6 139 2 49 7 8 5 134 13 
159 8 19 4569 4569 3 2 146 7 
 
At this point there are 36 unresolved cells. Not seeing anything obvious, I started concentrating on all the cells that had just two possible candidates. I was hoping to find a "doubleimplication chain", which often emerges when all else fails. In this case, though, there are only 14 "pairs" among the unresolved cells, and I couldn't find one of those patterns.
Anyway, I was staring at the possible locations for the digit 6 when I noticed that the two cells r4c4 and r6c7 (both of which appear as {5, 6} in the grid) are linked together in such a way that if r4c4 were "6" then r6c7 couldn't be "6". That prompted me to investigate the other possible locations for a "6" in column 4, and voila!
Now that I'm staring at this thing again I notice that a limited form of "Nishio" is possible in this situation. Here's the grid again.
Code:  
3 2 6 7 19 4 89 5 189 
1459 1459 8 3 1569 569 7 16 2 
159 7 19 56 8 2 4 136 1369*

8 69 3 2 569 569 1 7 4 
149 1469 5 8 469 7 3 2 69 
2 469 7 469 3 1 69* 8 5 

7 35 4 1 2 56* 68 9 368 
6 139 2 49 7 8 5 134 13 
159 8 19 4569 4569 3 2 146* 7 
 
Start by supposing that r7c6 = 6. We reach a contradiction as follows.
 r7c6 = 6 ==> r9c8 = 6 (only spot left in bottom right 3x3 box)
 r7c6 = 6 ==> r7c7 = 8 ==> r1c7 = 9 ==> r6c7 = 6.
 With "6" at r6c7 and r9c8, the only spot left in column 9 is r3c9.
I've marked these cells with asterisks in the grid above. As you can see, it's now impossible to fit a "6" in column 4. dcb 

Back to top 


dejsmith
Joined: 23 Oct 2005 Posts: 42

Posted: Thu Mar 16, 2006 8:57 pm Post subject: Slight VAriations 


Guys, I am getting pretty good at arriving at contradictions; but my "start hunting real hard" position had a few slight variations that got in the way of following both Keith & David. I had R2C1=1459 & R3C4=569. Notice my R2C1 is the same as David's & blocks Keith's X Wing attempt. However my R3C4 is the same as Marty's & kept me from noticing David's 56 pair combination with R7C6. Perhaps they are not necessary & I am nit picking; but I am trying to understand these small differences.
Thanks...Dave 

Back to top 


David Bryant
Joined: 29 Jul 2005 Posts: 559 Location: Denver, Colorado

Posted: Thu Mar 16, 2006 11:55 pm Post subject: "Coloring" eliminates "9" at r3c4 


Dave Smith wrote:  ... my R3C4 is the same as Marty's & kept me from noticing David's 56 pair combination with R7C6. 
Hi, Dave! That's a good question.
Code:  
3 2 6 7 19 4 89 5 189 
1459 1459 8 3 1569 569 7 16 2 
159 7 19= 569 8 2 4 136 1369 

8 69 3 2 569 569 1 7 4 
149 1469 5 8 469 7 3 2 69 
2 469 7 469 3 1 69 8 5 

7 35 4 1 2 56 68 9 368 
6 139+ 2 49 7 8 5 134 13 
159 8 19~ 4569 4569 3 2 146 7 
 
Here's how you can eliminate the "9" at r3c4.
 If r3c3 = 9 then r3c4 <> 9.
 If r3c3 <> 9 then r9c3 = 9 (only two spots for a "9" in column 3).
 But r9c3 = 9 ==> r8c4 = 9 ==> r3c4 <> 9.
In other words, the "connected pairs" in column 3 and in row 8 combine to rule out the possibility of a "9" at r3c4. dcb 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3253 Location: near Detroit, Michigan, USA

Posted: Fri Mar 17, 2006 1:47 am Post subject: Starting over ... 


I went back to the starting position. You can get to this position with basic methods:
Code: 
++++
 3 2 6  7 19 4  89 5 189 
 1459 1459 8  3 1569 569  7 16 2 
 159 7 19  569 8 2  4 136 1369 
++++
 8 69 3  2 569 569  1 7 4 
 149 1469 5  8 469 7  3 2 69 
 2 469 7  469 3 1  69 8 5 
++++
 7 35 4  1 2 56  68 9 368 
 6 139 2  49 7 8  5 134 13 
 159 8 19  4569 4569 3  2 146 7 
++++

There is a unique rectangle  look at the <19>'s in C3. For uniqueness, either R4C1 or R9C1 is <5>, so R3C1 is not <5>.
Now there is an Xwing on <5> in C1 and C4, so R9C5 cannot be <5>.
Keith 

Back to top 


dejsmith
Joined: 23 Oct 2005 Posts: 42

Posted: Fri Mar 17, 2006 4:13 am Post subject: Great Lessons! 


Guys, what great lessons! You both have shown me another level of awareness that I need to consider more often. Thanks for your patience.
Dave 

Back to top 


Marty R.
Joined: 12 Feb 2006 Posts: 5446 Location: Rochester, NY, USA

Posted: Fri Mar 17, 2006 6:18 pm Post subject: Re: Starting over ... 


David, thank you for the explanation. I'm familiar with the term "Nishio", but I wonder if it's really a separate technique. If I understand it correctly, it seems to really be part of a forcing chain. Many times I've done a forcing chain and instead of forcing values in cells, one of the starting points has led to an invalid solution. I think that's what a Nishio is, even though I didn't start out thinking about one.
By the way, is your "doubleimplication chain" the same as what is often referred to as a "forcing chain"?
keith wrote:  I went back to the starting position. You can get to this position with basic methods:
Code: 
++++
 3 2 6  7 19 4  89 5 189 
 1459 1459 8  3 1569 569  7 16 2 
 159 7 19  569 8 2  4 136 1369 
++++
 8 69 3  2 569 569  1 7 4 
 149 1469 5  8 469 7  3 2 69 
 2 469 7  469 3 1  69 8 5 
++++
 7 35 4  1 2 56  68 9 368 
 6 139 2  49 7 8  5 134 13 
 159 8 19  4569 4569 3  2 146 7 
++++

There is a unique rectangle  look at the <19>'s in C3. For uniqueness, either R4C1 or R9C1 is <5>, so R3C1 is not <5>.
Now there is an Xwing on <5> in C1 and C4, so R9C5 cannot be <5>.
Keith 
Keith, are you off a column? Do you mean R3C1 as part of the rectangle and the "5" precluded from R2C1?
As to the XWing in columns 1 and 4: maybe I have a misunderstanding of what an XWing is, but I thought a requirement was that the rows involved have no more of the number in question. I.e., each row in the XWing can contain only two of the number. So I keep seeing that "5" in R9C5 and thinking that that precludes an XWing because of three "5s" in R9.
As usual, I'm in a state of.....CONFUSION. 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3253 Location: near Detroit, Michigan, USA

Posted: Fri Mar 17, 2006 9:50 pm Post subject: 


Marty,
Quote: 
Keith, are you off a column? Do you mean R3C1 as part of the rectangle and the "5" precluded from R2C1?

Yes. Actually, I am off a row.
As I said, the Xwing is in the columns. The possibility <5> occurs only twice in each of C1 and C4, and these possibilities line up in R3 and R9 to form a rectangle. One of the two squares in each row must be <5>, so we can exclude <5> as a possibility in any other squares in these rows.
Marty, the same argument is true if you interchange the words "row" and "column". Or, just rotate the puzzle ninety degrees.
Keith 

Back to top 


David Bryant
Joined: 29 Jul 2005 Posts: 559 Location: Denver, Colorado

Posted: Fri Mar 17, 2006 11:57 pm Post subject: "Nishio" is special 


Marty R wrote:  David, thank you for the explanation. I'm familiar with the term "Nishio", but I wonder if it's really a separate technique. If I understand it correctly, it seems to really be part of a forcing chain. Many times I've done a forcing chain and instead of forcing values in cells, one of the starting points has led to an invalid solution. I think that's what a Nishio is, even though I didn't start out thinking about one. 
You're welcome, Marty.
"Nishio" is a variety of forcing chain. But it's special. In particular it has a special name because it was invented in Japan, before the sudoku puzzles became wildly popular in the U.K. and in the U.S.A.
Nishio works by showing that a single digit cannot be placed in a particular cell because if that digit is placed in that cell, it will not be possible to place all 9 occurrences of that digit in the puzzle. So Nishio works by concentrating on a single digit.
Forcing chains that end in contradictions may involve more than one digit. Or the contradiction that is reached may involve forcing the same digit to appear twice in the same column, or row, or 3x3 box. These kinds of forcing chains are useful, but they're not "Nishio."
Marty R wrote:  By the way, is your "doubleimplication chain" the same as what is often referred to as a "forcing chain"? 
Well, the term "forcing chain" is generally used to refer to any chain of reasoning that begins by supposing that a certain cell has a particular value.
A "doubleimplication chain" is a form of forcing chain, but it's also a little bit special. DIC's come in two flavors.
 The first one  my favorite flavor  starts from a cell in which only two digits are possible. Two different chains lead away from this cell, each based on a different assumption about the contents of the starting cell. And the two chains intersect somewhere, demonstrating that a particular value cannot appear in the cell where the two chains intersect.
The simplest form of this flavor DIC is the familiar XYWing. Only four cells are involved, and each "chain" has two links in it. A common form of this flavor of DIC is just a little more complicated, involving a third link on one side of the chain.
Code:  . 35 . . 37 . 57 . .
. 56 . . 78 . . . 67 
This is a "5star constellation."
 If r1c2 = 3 then r1c5 = 7, forcing r2c5 = 8.
 If r1c2 = 5 then r2c2 = 6, forcing r2c9 = 7 and therefore r2c5 = 8.
You can find quite a bit of discussion about this kind of a "DIC" by searching for the word "constellation" in this forum.
 The second flavor of DIC just tests one value that might occur in the starting cell. It follows two different chains of inference leading away from this cell; those chains intersect somewhere else in the puzzle and contradict each other, thus demonstrating that the assumed starting value is incorrect.
I hope that all makes sense, Marty. Oh  I'm curious. Where did you find the puzzle that started this whole thread? It's a very tough one! dcb 

Back to top 


Steve R
Joined: 24 Oct 2005 Posts: 289 Location: Birmingham, England

Posted: Sat Mar 18, 2006 1:14 am Post subject: Stuck 


Yes, please tell us more. This is the best puzzle I have seen by a country mile. I should not have realised it without the help of DCB’s keen eye: the grid invites all sorts of complexities but succumbs to an elementary solution.
Quality puzzle and most elegant resolution.
Steve. 

Back to top 


Marty R.
Joined: 12 Feb 2006 Posts: 5446 Location: Rochester, NY, USA

Posted: Sat Mar 18, 2006 5:51 pm Post subject: 


keith wrote:  Marty,
Quote: 
Keith, are you off a column? Do you mean R3C1 as part of the rectangle and the "5" precluded from R2C1?

Yes. Actually, I am off a row.
As I said, the Xwing is in the columns. The possibility <5> occurs only twice in each of C1 and C4, and these possibilities line up in R3 and R9 to form a rectangle. One of the two squares in each row must be <5>, so we can exclude <5> as a possibility in any other squares in these rows.
Marty, the same argument is true if you interchange the words "row" and "column". Or, just rotate the puzzle ninety degrees.
Keith 
Keith, I guess I was asleep at the switch, thinking that if it was in the columns, I had to be looking vertically instead of horizontally. I see it quite clearly now.
Quote:  Oh  I'm curious. Where did you find the puzzle that started this whole thread? It's a very tough one! 
Once again, thank you David for your explanations.
I borrowed "Sudoku for Dummies" from the library. I was actually somewhat disappointed with it because I thought it would be a book of strategies; instead, it was a book of puzzles with just a few pages covering the most basic techniques. But I copied onto blank grids all the puzzles graded "diabolical" and this was one of them. Most of them aren't as difficult and interesting as this one.
I might mention that the puzzles graded "tough" from http://sudoku.com.au are troublesome for me. A number of sources that offer puzzles have four degrees of difficulty, but I have more trouble with this fourth level than others' fourth. 

Back to top 




You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum

Powered by phpBB © 2001, 2005 phpBB Group
