View previous topic :: View next topic 
Author 
Message 
keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Sun Jul 27, 2008 11:59 am Post subject: The Finned XYWing 


Last Sunday I proposed the idea of a "Finned XYWing" here.
If the XYwing is: XZXYYZ.
The "fin" is an extra value on any one of the three cells:
A) VXZXYYZ,
B) XZVXYYZ, or
C) XZXYVYZ.
Either the XYwing is true, or "V" is true. Cases A and C are the same thing.
For example, the Brain Basher's "Super Hard" of July 19:
Code:  Puzzle: BB071908sh
++++
 4 . 3  8 . 2  6 . 9 
 . . .  1 . 6  . . . 
 1 . .  . 9 .  . . 7 
++++
 6 3 .  . . .  . 7 8 
 . . 7  . 1 .  3 . . 
 2 4 .  . . .  . 9 5 
++++
 7 . .  . 8 .  . . 3 
 . . .  7 . 1  . . . 
 8 . 9  3 . 5  7 . 1 
++++ 
Basics get us to here: Code:  ++++
 4 57 3  8 57 2  6 1 9 
 59 25789 25  1 457 6  2458 3 24e 
 1 2568 256  45a 9 3  2458 245d 7 
++++
 6 3 1  2459 245c 49  24 7 8 
 59 59 7  24b 1 8  3 246 246 
 2 4 8  6 3 7  1 9 5 
++++
 7 1 246  249 8 49  245 2456 3 
 3 256 2456  7 246 1  9 8 246 
 8 26 9  3 246 5  7 246 1 
++++ 
Note the cells a, b, c. Either c is <4>, or the XYwing 45, 24, 25 is true.
Asellus made the rather brilliant observation that if the XYwing is true, c is solved as <5> (look at C5). Either way, c cannot be <2>. This solves the puzzle.
In the grid above, there is also an XYZwing, a, d, e. Note that you can regard an XYZwing as a finned XYwing. In this case, either the pivot cell, d, is <4>, or the XYwing 45, 25, 24 is true. Either way, R3C7 is not <4>. This XYZwing does not solve the puzzle.
Keith
Last edited by keith on Sun Jul 27, 2008 12:04 pm; edited 1 time in total 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Sun Jul 27, 2008 12:00 pm Post subject: Another Finned XYWing Example 


The "Very Hard" puzzle from this site for July 24:
Code:  Puzzle: DS072408vh
++++
 . . .  5 . .  . 7 9 
 . . 3  8 . .  6 . 2 
 6 . .  . . 2  . . . 
++++
 . 3 .  . 2 .  . 1 5 
 . . 1  . . .  2 . . 
 2 4 .  . 9 .  . 6 . 
++++
 . . .  1 . .  . . 3 
 1 . 2  . . 9  5 . . 
 3 7 .  . . 6  . . . 
++++ 
After basics:
Code:  ++++
 4 2 8  5 6 3  1 7 9 
 9 1 3  8 47 47  6 5 2 
 6 5 7  9 1 2 348a 38 48b 
++++
 78 3 69  46 2 78  49 1 5 
 78 69 1  46 378 5  2 39 47 
 2 4 5  37 9 1  38c 6 78d 
++++
 5 69 69  1 48 48  7 2 3 
 1 8 2  37 37 9  5 4 6 
 3 7 4  2 5 6 89 89 1 
++++ 
Look at a, b, and c. Either a is <8>, or the XYwing bac is true: a is <34>. The XYwing eliminates <8> in d, solving c as <8>. (Look at R6.)
Either way, the <8> in C7 cannot be in R9C7, and the puzzle is solved.
Last edited by keith on Sun Jul 27, 2008 12:57 pm; edited 1 time in total 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Sun Jul 27, 2008 12:00 pm Post subject: Examples and comments 


Here are some comments and links to other examples.
The reasons I thought that this idea might be interesting are twofold:
1. The pattern is easy to see, particularly if you are scanning for XYwings or XYZwings.
2. It seemed to me it might reveal connections or implications that other methods do not.
You might read the original thread.
This thread has some finned XYwings and other methods for a fairly difficult puzzle.
Another example by ravel, and comments by Asellus are here.
Keith 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Sun Jul 27, 2008 6:03 pm Post subject: Finned XYWings: Quick Eliminations 


The finned XYwing is just a way to start a bifurcation: Either "A" is true, or "B" is true. Presumably, you could just follow A and B until you find a candidate that is either true, or one that is false, for both A and B.
But, the hope is that this is not just another way to start a trialanderror elimination. As Asellus might say, "Are there compact patterns we can look for?"
I think so. Suppose the finned wing is:
XZXYVYZ
Either the XYwing (XZXYYZ) is true, or the fin (V) is true.
1. Suppose the XYwing solves the finned cell as Z. Then you can eliminate Y as a candidate in the finned cell. That is Asellus' elimination in the first post of this thread. (You can interchange Y and Z here. It does not matter which wing candidate is solved in the finned cell.)
2. Suppose the XYwing solves some other cell as V. Then, that solved cell and the finned cell are a strong link. V can be eliminated in all cells that see them both. I think ravel and I have both used this, see the second message of the thread, and the links in the third.
3. Note that 2. above is a true strong link. In the solution, only one of the cells is V. So, the link can be put in any chain in place of a "regular" strong link. I think it was Asellus who noted that the stong links revealed by finned XYwings can otherwise be very difficult to recognize.
4. It may be that both the XYwing and the fin lead to some other candidate, say "U", being true in two different cells. Again, those two cells are pincers, eliminating U in any cell that sees them both.
But, I think this last one is outside the spirit of trying to find "compact" eliminations for those of use who use pencil and paper.
Keith
(I am now done for the time being. No more posts from me until others post responses on this thread.)
Last edited by keith on Mon Jul 28, 2008 12:04 am; edited 2 times in total 

Back to top 


wapati
Joined: 10 Jun 2008 Posts: 472 Location: Brampton, Ontario, Canada.

Posted: Sun Jul 27, 2008 6:21 pm Post subject: 


This isn't EXACTLY the topic but...
I combine patterns frequently. I just used a UR (which gave me "ab", say) to be the end of an xywing, ab ac bc, but in 4 cells.
Did that make it a sorta xychain or was it a URxywing?
It is pattern recognition in my books, as is this finned xywing you are examining. 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Sun Jul 27, 2008 11:54 pm Post subject: 


wapati wrote:  This isn't EXACTLY the topic but...
I combine patterns frequently. I just used a UR (which gave me "ab", say) to be the end of an xywing, ab ac bc, but in 4 cells.
Did that make it a sorta xychain or was it a URxywing?
It is pattern recognition in my books, as is this finned xywing you are examining. 
wapati, I think this is EXACTLY the point. How do you recognize and combine patterns that otherwise would be guesswork?
Keith 

Back to top 


Asellus
Joined: 05 Jun 2007 Posts: 865 Location: Sonoma County, CA, USA

Posted: Mon Jul 28, 2008 10:42 am Post subject: 


keith wrote:  3. Note that 2. above is a true strong link. In the solution, only one of the cells is V. So, the link can be put in any chain in place of a "regular" strong link. I think it was Asellus who noted that the stong links revealed by finned XYwings can otherwise be very difficult to recognize. 
Yes, I think it probably was.
I would hasten to add that anything determined to be true IF the potential XY Wing is true is thus strongly linked with the fin digit(s). [Note that the fin doesn't necessarily have to be a single digit. However, using a multidigit fin can be tricky and there are no examples of such a thing posted as yet.] It is important to note, however, that these strong links are strong inference links and not necessarily conjugate (strong) links. So, be sure you understand this distinction if you are going to employ these links in chains.
On a related note, in my original posts on this subject, I cautiously stated that the fintowing link was a "unidirectional" strong link. I have now decided that this was not correct and have posted an explanatory edit to the original post. So... you can forget about that "unidirectional" stuff! 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Mon Jul 28, 2008 10:37 pm Post subject: 


Asellus,
I am not sure I understand the distinction between "Inferential" and "Conjugate", but that is OK. I am not a programmer (any more), and I check my chains in both directions to make sure they are correct.
But, I do have a question. Suppose the fin is V, and the XYwing solves some other cell as U. Can I do Medusa coloring on U and the fin, V, as if it was a twocandidate cell?
Suppose my Medusa colors are Red and Green.
If the fin is Red, can I color U in its cell as Green? (I think so.)
If U in its cell is Green, can I color the fin Red? (I am less sure.)
Maybe I am questioning your directional assertion. It is certainly possible that the XYwing is true and the fin is true both lead to some cell where T is true. See my original post. If T is true, can I make any statement about the the finned XYwing?
Keith
[The heck with Holiday Inn Express. The tag line should be: "... but I did have a product of Sonoma County last night."] 

Back to top 


Asellus
Joined: 05 Jun 2007 Posts: 865 Location: Sonoma County, CA, USA

Posted: Tue Jul 29, 2008 10:34 am Post subject: 


keith wrote:  But, I do have a question. Suppose the fin is V, and the XYwing solves some other cell as U. Can I do Medusa coloring on U and the fin, V, as if it was a twocandidate cell? 
In general, no. This gets to that strong inference versus conjugate distinction (and for me, it is all about the logic of sudoku solving and nothing whatsoever about computer programming!). A strong inference exists between two things when they cannot both be false. That is the case here. U and V cannot both be false so they possess a strong (inference) link. But, they are not necessarily conjugate: is it possible that both of them might be true.
(While people commonly call conjugate links "strong" links, it is actually the case that they are both strong... "cannot both be false"... and weak ... "cannot both be true". That is why a conjugate link can "substitute" for a weak link in an AIC. It isn't really substituting for a weak link because it IS a weak link, as well as being a strong link. But, our U and V link is only strong as far as we know.)
However, there is a way to exploit this U=V link with Medusa. But, you must use multicoloring. Let's say we color U "A". Then, we use this to color an "Aa" cluster. Next, we color V "B" and construct a "Bb" cluster from that. Now, all "A"s are strongly linked with all "B"s and so can perform trapping. However, the "a"s and the "b"s are weakly linked and so cannot trap anything.
Usually, folks "bridge" two color clusters by exploiting a weak link between them, in which case the strong "trapping" colors are the ones opposite those that are involved in the weak link bridge. But, a "strong only" inference link can also bridge two clusters, as in this case. The difference is that, now, the strong "trapping" colors are the ones involved in the bridge.
Quote:  Maybe I am questioning your directional assertion. 
The bidirectional nature of the strong inference is a separate issue. Perhaps you are uncertain because you are trying to see this as a conjugate link, which it is not. As a strong inference, it is bidirectional: Vfalse implies Utrue and Ufalse implies Vtrue. But, Vtrue doesn't imply anything; nor does Utrue.
Quote:  It is certainly possible that the XYwing is true and the fin is true both lead to some cell where T is true. See my original post. If T is true, can I make any statement about the the finned XYwing? 
There is no "if" about it: T must be true! Other than saying that the Finned XY Wing makes T true, what would you like to say about it? This is a different situation from what we were considering above. Here, you have a discontinous AIC loop were T is connected by strong links on both ends. We could represent this using a generic chain notation, to which I will add the truefalse implication sequences written below it:
Code:  T = a  V = Wing  b = T
L to R > F T F T F T
T F T F T F < R to L 
(AICs of any length may be inserted in the "a" and "b" positions. As before, V is the fin.) No matter which way we read the chain, an assumption that T is false leads to a contradiction (that T must be true). So, T must be true.
Now... in the case that all of the links in this loop other than "V=Wing" are conjugate links, then both the fin and the wing must be true! This might strike you as impossible. But, it isn't. That's because the statement "the XY Wing is true" for these implication purposes means only that at least one of its pincer digits is true (since that is all that is required to render the wing victims false). And, that can happen even if the "ZXXYYZ" pattern itself is destroyed, as it is if the fin is true.
I hope this has answered your questions... assuming you're still with me. 

Back to top 


keith
Joined: 19 Sep 2005 Posts: 3173 Location: near Detroit, Michigan, USA

Posted: Tue Jul 29, 2008 10:24 pm Post subject: 


Asellus,
Quote:  I hope this has answered your questions... assuming you're still with me. 
Asellus, yes, and yes. I think my issue is to reconcile my (sometimes unstated) definitions and terminology with yours. And, I do not intend to confuse this thread with that discussion.
Quote:  ... means only that at least one of its pincer digits is true 
I would have said, "... means only that the pincer digit is true in at least one of the pincer cells."
Thank you,
Keith
[Did the earth move, for you, today?] 

Back to top 


Johan
Joined: 25 Jun 2007 Posts: 206 Location: Bornem Belgium

Posted: Wed Aug 06, 2008 12:53 am Post subject: A Finned XYZwing ?? 


This puzzle is from "Die Presse" Saturday 2nd of August and was posted by nataraj in the other puzzles forum, the puzzle can be solved with the advanced VH techniques.
Reading the latest posts by Keith, Asellus and others, for finned xywings i tried to spot one of these in the grid, but found another(I'm not sure about this) finned creature, a finned xyzwing, which eliminates both <5>'s in R12C7.
The question is, can the logic of a finned xywing (either/or the xywing, the fin is true) be used for a finned xyzwing or was this a lucky shot??
Code:  ++++
 . 3 .  . . .  . 2 . 
 . . .  . . .  . . . 
 1 . 2  . . .  9 . 7 
++++
 . 1 .  . . .  . 9 . 
 9 4 .  2 . 1  . 3 6 
 . . .  . 8 .  . . . 
++++
 . . 1  . . .  7 . . 
 . 2 .  6 . 5  . 4 . 
 . . 5  7 . 4  1 . . 
++++ 
After basics:
Code:  ++++
 568 3 49  1 5679 6789  4568 2 458 
 568 7 49  589 2569 2689  34568 1 3458 
 1 68 2  3458 3456 368  9 58 7 
++++
 2378 1 3678  345 34567 367  2458 9 458 
 9 4 78  2 57 1  58 3 6 
 23 5 36  349 8 369  24 7 1 
++++
 4 69 1  389 239 2389  7 56 235 
 378 2 378  6 1 5  38 4 9 
 368 689 5  7 239 4  1 68 238 
++++

The finned {683683(7)8} xyzwing:
If the {368} xyzwing is true, it eliminates <8> in R9C2 => R3C2=8 => R3C8=5(a)
If the Fin=7 in R8C3 is true => R5C3=8 => R5C7=5(b)
=> R12C7<>5
Code:  ++++
  5 
  5 
 =8   =5 a 
++++
   
 =8  =5 b 
   
++++
   
 3(7)8   
 368 8   68 
++++ 


Back to top 


nataraj
Joined: 03 Aug 2007 Posts: 1029 Location: Vienna, Austria

Posted: Wed Aug 06, 2008 6:24 am Post subject: 


Johan, the reasoning is perfectly valid.
The "Either the wing ... OR the fin ..." constitutes a bifurcation like any other, (e.g. "either r1c3=4 or r1c3=9"), in my opinion no better or worse. If both assumptions lead to the same result, that result must be true.
When using such a "proof by exhaustion", you just got to be careful that no possible alternatives are forgotten. There must not be "a middle ground" (tertium no datur). Either case 1 or case 2 (or possibly both, but I prefer to make sure the cases don't overlap), but no third possibility.
My take on these finned creatures: Are finned xy(z) wings easier to spot than simple bivalue cells? I don't think so. Do they turn up a usable elimination more often than bivalue cells? Don't know but in general I doubt it. Personal conclusion: I dont go looking for them but sometimes use them when I step on one.
Finned xwing are a slightly different matter (but still, this is strictly personal taste!) in that I usually have a sketch in front of me and both the xwing pattern and the fin tend to stand out very clearly in those diagrams.
As for using the finned guys in order to obtain strong links not easily found otherwise, that is an idea with great potential value. With multicoloring, wwing and mwing making extensive use of the simple strong links (those with only two "6" cells in a row for example), finding a strong link that jumps houses should be very valuable when solving the more difficult puzzles... 

Back to top 


Asellus
Joined: 05 Jun 2007 Posts: 865 Location: Sonoma County, CA, USA

Posted: Wed Aug 06, 2008 8:54 am Post subject: 


To comment a bit on what nataraj said...
These "finned wings" function by creating a strong inference between the "fin" and the grouped pincer digits of the wing. It is not an "either/or" (conjugate) situation. It is possible that both are true. [In this finned XYZ Wing case, if the <7> fin is true, it is still possible that one of the two remaining <8> pincer digits is true; we don't know.] What is not possible is that both things are false.
It is obvious that if the "fin" is false, one or more of the pincer digits must be true since we then have a straightforward XYZ Wing. It is slightly less obvious that if all of the pincer digits are false then the fin must be true; however, that is the case. Once we accept this as given, then we only have to consider the strong inference that the "finned wing" contains. That means that we don't need to spend time thinking about "if this is true" or "if that is false" etc. We just construct an AIC that exploits the strong inference.
In Johan's example, the grouped pincer <8>s are: (8)r8c3r9c18
The strong inference of the "finned wing" is: (7)r8c3=(8)r8c3r9c18
To make it clear to others that you are working with XYZ Wing pincers, you could write:
(7)r8c3=XYZ Wing[(8)r8c3r9c18]
It is then natural to add the weak link to one or more of the wing victims:
(7)r8c3=XYZ Wing[(8)r8c3r9c18]  (8)r9c2
Once you have this core structure of the finned wing, you can append alternate links as you are comfortable with. Johan exploited the strong link on <8> in c2 plus bivalue sequences (as in XY chaining):
(5=8)r5c7  (8=7)r5c3  (7)r8c3=XYZ Wing[(8)r8c3r9c18]  (8)r9c2=(8)r3c2  (8=5)r3c8; r12c7<>5
The <5>s on the ends of the chain are strongly linked so we know they can function as pincers for eliminations. If you like, you could also see this as a 58 WWing with the following external strong link on <8> that "activates" it:
(8=7)r5c3  (7)r8c3=XYZ Wing[(8)r8c3r9c18]  (8)r9c2=(8)r3c2
So, rather than thinking "true, false, true, false..." as I construct these sorts of things, I think "strong, weak, strong, weak..." as I trace out the links. It's like thinking "red, green, red, green..." when constructing a color chain (which I do without any marking in basic cases).
Since XY and XYZWings are just the simplest cases of the more general "paired ALS" eliminations, any such paired ALS "wing" can also be finned. And, WWings can be finned as well.
By the way, the "fin" doesn't need to be a single digit nor limited to a single cell of the wing! In principle, grouped digits can also serve as a "fin." However, you have to be certain that you can form a valid weak link to the grouped "fin" digits in order for this to work. 

Back to top 


nataraj
Joined: 03 Aug 2007 Posts: 1029 Location: Vienna, Austria

Posted: Wed Aug 06, 2008 8:41 pm Post subject: 


Asellus, thanks for the clarification! In my post, I was too quick to call the finned wing a "strong link" when in reality the strong link is only between the fin and the (grouped) pincers, as you pointed out.
A complete building block for eliminations would then have to include not only any (possibly grouped) victims
Quote: 
It is then natural to add the weak link to one or more of the wing victims:
(7)r8c3=XYZ Wing[(8)r8c3r9c18]  (8)r9c2
Once you have this core structure of the finned wing,

but also
Quote: 
you can append alternate links as you are comfortable with

I would even go as far as saying that one more link is mandatory in all those cases where the fin does not "see" the victim directly. This addidional strong link provides that "other end" that the final victim will "see" (and get eliminated by seeing)
Simple case: fin "sees" victim of the wing (like in a usual finned xwing)
<victim> (fin)=[wing pincers] <victim>
That simple case could be rephrased as "victim sees both ends of the (fin)=[wing pincers] strong link"
General case:
a "core" consisting of
(fin)=[wing pincers][wing victims]=("other end")
and any number of transports (strong/weaksequences) tacked on to either end of the core.
In Keith's initial example, (4)r4c5 would be the fin, [(5)r3c4r4c5] the wing pincers, (5)r12c5 the wing victims and (5)r4c5 the "other end"
The whole finned xywing elimination looks like this, then
(4)r4c5=XYWing[(5)r3c4r4c5](5)r12c5=(5)r4c5;r4c5<>2
I hope I got it mostly right now, any corrections much appreciated! 

Back to top 


Asellus
Joined: 05 Jun 2007 Posts: 865 Location: Sonoma County, CA, USA

Posted: Wed Aug 06, 2008 9:40 pm Post subject: 


nataraj,
That looks fine, though I'd point out two things.
First, unless I'm missing something, I don't believe that the simple case is possible with finned XY or XYZWings. The fin would need to be the same digit as the victims/pincers. In the XYZ Wing case, all three cells already contain that digit (as pincers). There's no place for such a fin! In the XY Wing, the only place for such a fin is the pivot. But, that would just make it an XYZ Wing. Perhaps such an arrangement is possible in a 2ALS "wing" that involves more cells, though I'm not certain (nor am I expecting to see an example any time soon!).
Second, I just want to make clear that one does not need to "weak link" to all of the (grouped) wing victims, though that is an option. The weak link can be to any single victim or to any (sub)set of the victims considered as a group... whatever works to provide the ensuing useful strong link. Often, it is only a "single victim" link that is useful. 

Back to top 


nataraj
Joined: 03 Aug 2007 Posts: 1029 Location: Vienna, Austria

Posted: Wed Aug 06, 2008 9:51 pm Post subject: 


Thanks, Asellus!
I had not really thought that far about the simple case ... you are right, there is no place for a candidate that sees both pincers AND the fin in an xy(z)wing.
And on your second point I agree completely. Almost put it in myself but did not want to add too much text and distract from the basic element:
[transports...]fin=wing pincerswing victim(s)=other end[transports...]
which in itself (with or without the transports) is another strong link to be used as is convenient... 

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
