The New Sudoku Players' Forum

[ArkieTech]

Code: Select all

 *-----------*
 |...|...|..3|
 |.1.|.2.|98.|
 |6..|.8.|1..|
 |---+---+---|
 |3..|..1|76.|
 |..9|.4.|2..|
 |.81|7..|..9|
 |---+---+---|
 |..7|.1.|..5|
 |.54|.3.|.7.|
 |2..|...|...|
 *-----------*


Play/Print this puzzle online

[SpAce]

Code: Select all

.--------------------.-----------------.------------------.
| 9   a247     8     | 1     57  4567  | 456  d45(2)  3   |
| 45   1       35    | 3456  2   34567 | 9     8      67  |
| 6   a47(2)  a35(2) | 3459  8   34579 | 1    d45(2)  7-2 |
:--------------------+-----------------+------------------:
| 3   b24      25    | 58    9   1     | 7     6     b48  |
| 7    6       9     | 358   4   358   | 2    c15    c18  |
| 45   8       1     | 7     6   2     | 345   345    9   |
:--------------------+-----------------+------------------:
| 8    39      7     | 246   1   46    | 346   2349   5   |
| 1    5       4     | 2689  3   689   | 68    7      26  |
| 2    39      6     | 458   57  4578  | 348   1349   14  |
'--------------------'-----------------'------------------'

(2)b1p[89=2] - (2=48)r4c29 - (8=15)r5c89 - (5=42)r13c8 => -2 r3c9; stte

aka:

Alien 7-Fish (Rank 1):

{2B1 13N8 4N29 5N89} \ {1r5 2r3c2b3 4r4b3 5c8 8c9} => -2 r3c9; stte

[eleven]

Code: Select all

 *---------------------------------------------------------------*
 |  9   b247   8     |  1     a57   4567    |  46-5 d245    3    |
 |  45   1     35    |  3456   2    34567   |  9     8      67   |
 |  6   c247   235   |  3459   8    34579   |  1    d245   d27   |
 |-------------------+----------------------+--------------------|
 |  3    24    25    |  58     9    1       |  7     6      48   |
 |  7    6     9     |  358    4    358     |  2     15     18   |
 |  45   8     1     |  7      6    2       |  345   345    9    |
 |-------------------+----------------------+--------------------|
 |  8    39    7     |  246    1    46      |  346   2349   5    |
 |  1    5     4     |  2689   3    689     |  68    7      26   |
 |  2    39    6     |  458    57   4578    |  348   1349   14   |
 *---------------------------------------------------------------*

(5=7)r1c5 - r1c2 = r3c2 - (7=5)r3c9,r13c8 => -5r1c7, stte

Just a note:
Solutions like that i don't find with contradictions, but forcing chains, here from the strong link 7r13c2.
Another one is using the skyscraper 5c17:
5r1c7 = r6c7 - r6c1 = r2c1 - (5=2)r23c3 - (2=6)r23c9 => -6r1c7

[SpAce]

Solutions like that i don't find with contradictions, but forcing chains, here from the strong link 7r13c2.

I found mine with a contradiction this time. Started quite randomly coloring with the Kite cluster of 6s in r1,b3,c9, found some useless "trap" eliminations along the way, and then a contradiction in r13c2 proving one parity false. Interestingly there weren't any 6s in my solution, but one nice thing about GEM is that any par-candidate of the same parity works equivalently. So when you find a contradiction they're all equally doomed and you can just pick the most convenient first victim.

[SteveG48]

Code: Select all

 *--------------------------------------------------------------------*
 | 9     b247    8      | 1     a57     4567   | 46-5  d245    3      |
 | 45     1      35     | 3456   2      34567  | 9      8      67     |
 | 6     c247    235    | 3459   8      34579  | 1     d245   d27     |
 *----------------------+----------------------+----------------------|
 | 3      24     25     | 58     9      1      | 7      6      48     |
 | 7      6      9      | 358    4      358    | 2      15     18     |
 | 45     8      1      | 7      6      2      | 345    345    9      |
 *----------------------+----------------------+----------------------|
 | 8      39     7      | 246    1      46     | 346    2349   5      |
 | 1      5      4      | 2689   3      689    | 68     7      26     |
 | 2      39     6      | 458    57     4578   | 348    1349   14     |
 *--------------------------------------------------------------------*


(5=7)r1c5 - r1c2 = r3c2 - (7=245)b3p289 => -5 r1c7 ; stte

Dang! Exactly the same as Eleven.

However, it let's me add to the contradiction discussion. I found this by contradiction after deciding that eliminating 5 at r1c7 would solve the puzzle. So I asked myself what would happen if I assigned a 5 instead of removing it. That leads to:

5r1c7 - (5=7)r1c5 - r1c2 = r3c2 - (7=245)b3p289; contradiction on 5s in b3. In my early days here, that's how I would have presented it. I quickly learned that this was frowned on. Fortunately, I soon realized that just dropping the first term made my chain shorter and turned it into an acceptable "pincer" solution. Problem solved.

[SpAce]

However, it let's me add to the contradiction discussion. I found this by contradiction after deciding that eliminating 5 at r1c7 would solve the puzzle. So I asked myself what would happen if I assigned a 5 instead of removing it.

Thanks for sharing, Steve! That's exactly what I also do if I fail to find an stte-solution otherwise in a reasonable time. In that case I pick a promising backdoor candidate (either one that I've recognized as such, or one reported by Hodoku) and find a contradiction (if none found easily, I try another one). Then it's just a matter of turning it into a verity (AIC or Kraken) for reporting. However, I prefer to find a solution without resorting to this method, because starting with a known stte-elimination feels a bit like cheating.

How do you guys track the chains? In your head or what? As you know, I use coloring unless the chains are very short and easy to see. I also need it much less if I solve on p&p and use my strong-link markup, because then I can both see chains more easily and am prepared to spend more time looking for them, but these daily puzzles I solve on Hodoku (much faster, obviously).

5r1c7 - (5=7)r1c5 - r1c2 = r3c2 - (7=245)b3p289; contradiction on 5s in b3. In my early days here, that's how I would have presented it. I quickly learned that this was frowned on.

Those kinds of solutions are still seen here from time to time, and yes, I do frown upon them For me it's the most difficult chain type to decipher, or at least it was the first time I saw it. It's not the contradiction that bothers me but the asymmetric links at the ends. It's really hard for me to see immediately what the chain actually proves. As a contradiction chain I'd write (and in my early days used to write) it like this:

5r1c7 - (5=7)r1c5 - r1c2 = r3c2 - (7=245)b3p289 - 5r1c7 => -5 r1c7

It's even longer, but then it's a valid Discontinuous Nice Loop, and the conclusion is clear as a day. I'm pretty sure that would be acceptable everywhere, even though it would be considered old-fashioned. I still think it's the most intuitive chain type for a beginning chain builder, or at least it was for me. I remember that it took me a while to start thinking in AIC terms and drop those redundant (and restricting) start and end nodes. Now it's a second nature, but for a beginner I might still recommend DNLs as an intro to chaining.

Fortunately, I soon realized that just dropping the first term made my chain shorter and turned it into an acceptable "pincer" solution. Problem solved.

Yes, a DNL (or almost DNL like your example) is of course trivial to turn into an AIC. As you surely know, that's not always the case with more complex contradiction scenarios, and I think eleven had those in mind. In those cases it might be not only easier to present the contradiction solution but also more representative of the logic used to find it. I don't really mind if someone wants to use that option, especially in eleven's case, because I think it might produce very instructive insights into his thinking and solving methods.

Personally I don't miss the option because I actually like the conversion process and also because my solving style rarely produces very innovative contradictions. (I'm hoping that might change if I see more examples of it, however. I'm pretty sure I'm still missing out on some interesting types of solving logic.)

[SteveG48]

How do you guys track the chains? In your head or what? As you know, I use coloring unless the chains are very short and easy to see.

I use coloring as well. I initially adopted Hodoku because it colors either cells or individual candidates.

5r1c7 - (5=7)r1c5 - r1c2 = r3c2 - (7=245)b3p289; contradiction on 5s in b3. In my early days here, that's how I would have presented it. I quickly learned that this was frowned on.

Those kinds of solutions are still seen here from time to time, and yes, I do frown upon them For me it's the most difficult chain type to decipher, or at least it was the first time I saw it. It's not the contradiction that bothers me but the asymmetric links at the ends. It's really hard for me to see immediately what the chain actually proves.

Personally, I've never had any problem with understanding the significance of a chain that begins with a weak link on a particular candidate and ends with a strong link on the same candidate, where the starting and ending cells are in the same house. Clearly, the candidate can be eliminated in the starting cell. When I used that form in the early days I didn't bother to say "contradiction". I assumed that it was obvious. Nevertheless, I've come to love the AIC that begins and ends on a strong link.

My real point, of course, is that the difference between the chains beginning on a weak link and those beginning on a strong link is so small in many cases that it's hard to see why people would find one formulation acceptable and the other not.

[SpAce]

I use coloring as well.

Great! That makes at least two of us then. I think this kind of information is very valuable for improving players. It shows that pretty decent solutions can be found manually with relatively simple tricks, and it doesn't require genius-level pattern-recognition skills -- or even knowing every pattern in the books. That should be encouraging information. I think too much of the available teaching material concentrates on illustrating various patterns, but fails to tell how to actually spot them, or better yet, use their contained logic without even recognizing them.

I initially adopted Hodoku because it colors either cells or individual candidates.

And even both at the same time, which is sometimes handy. I just think some of its default colors are awful, but fortunately they can be tweaked. I'd like to have even more colors and options but for most practical purposes it's quite decent. Then again, it's the only sudoku software (besides SudokuWiki online solver) I've used, so I don't know about the capabilities of others.

Personally, I've never had any problem with understanding the significance of a chain that begins with a weak link on a particular candidate and ends with a strong link on the same candidate, where the starting and ending cells are in the same house. Clearly, the candidate can be eliminated in the starting cell.

Good for you! I can't help that I find it very unintuitive. Any asymmetry annoys me too, so it's also a matter of aesthetics.

[SteveG48]

And even both at the same time, which is sometimes handy. I just think some of its default colors are awful, but fortunately they can be tweaked.

Speaking of which, if I can wander off topic for a moment, I periodically have Hodoku revert to its defaults. Then I have to reset all my personal preferences. Do you have that problem? And a solution?

[SpAce]

Speaking of which, if I can wander off topic for a moment, I periodically have Hodoku revert to its defaults. Then I have to reset all my personal preferences. Do you have that problem? And a solution?

Yes, it sometimes does that for no apparent reason. I've just saved the preferences to a file so they can be easily reloaded. You can find that option in "File:Save Configuration As..."

Another weird and slightly annoying problem I've noticed is that the color chooser sometimes misses mouse clicks. Do you have that too?

[SteveG48]

Speaking of which, if I can wander off topic for a moment, I periodically have Hodoku revert to its defaults. Then I have to reset all my personal preferences. Do you have that problem? And a solution?

Yes, it sometimes does that for no apparent reason. I've just saved the preferences to a file so they can be easily reloaded. You can find that option in "File:Save Configuration As..."

Another weird and slightly annoying problem I've noticed is that the color chooser sometimes misses mouse clicks. Do you have that too?

Yes. I don't know why, but it may be that background calculations are going on that upset things.

Thanks for the configuration tip. I'll try that.