The New Sudoku Players' Forum

by **ArkieTech** » Wed Jan 30, 2019 11:54 am

Code: Select all: *-----------* |...|..9|7..| |.1.|.2.|.8.| |6.3|5..|4..| |---+---+---| |4..|...|2..| |.3.|...|.6.| |..1|...|..7| |---+---+---| |..5|..2|6.9| |.2.|.1.|.5.| |..6|3..|...| *-----------*

Play/Print this puzzle online

by **Cenoman** » Wed Jan 30, 2019 2:07 pm

Code: Select all: +----------------------+---------------------+-----------------+ | 258 d58 248 | 1 b46 9 | 7 3 c56 | | 57 1 47 | 46 2 3 | 9 8 56 | | 6 9 3 | 5 78 78 | 4 2 1 | +----------------------+---------------------+-----------------+ | 4 e5678 78 | 678 a678-5 1 | 2 9 3 | | 2579 3 279 | 29 57 4 | 1 6 8 | | 29 68 1 | 29 3 68 | 5 4 7 | +----------------------+---------------------+-----------------+ | 3 78 5 | 478 478 2 | 6 1 9 | | 789 2 789 | 678 1 678 | 3 5 4 | | 1 4 6 | 3 9 5 | 8 7 2 | +----------------------+---------------------+-----------------+

(6)r4c5 = r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2 => -5 r4c5; ste

by **SteveG48** » Wed Jan 30, 2019 2:17 pm

Code: Select all: *-----------------------------------------------------------* |d258 8-5 248 | 1 b46 9 | 7 3 a56 | |d57 1 47 | 46 2 3 | 9 8 56 | | 6 9 3 | 5 b78 78 | 4 2 1 | *-------------------+-------------------+-------------------| | 4 5678 78 | 678 5678 1 | 2 9 3 | |c2579 3 279 | 29 b57 4 | 1 6 8 | | 29 68 1 | 29 3 68 | 5 4 7 | *-------------------+-------------------+-------------------| | 3 78 5 | 478 b478 2 | 6 1 9 | | 789 2 789 | 678 1 678 | 3 5 4 | | 1 4 6 | 3 9 5 | 8 7 2 | *-----------------------------------------------------------*

(5=6)r1c9 - (6=4578)r1357c5 - 5r5c1 = 5r12c1 => -5 r1c2 ; stte

by **SpAce** » Wed Jan 30, 2019 3:19 pm

Code: Select all: .------------------------.------------------.-----------. | 258 8-5 248 | 1 c46 9 | 7 3 c56 | | d(5)7 1 47 | 46 2 3 | 9 8 d56 | | 6 9 3 | 5 78 78 | 4 2 1 | :------------------------+------------------+-----------: | 4 a(5)678 78 | 678 b5678 1 | 2 9 3 | | 279-5 3 279 | 29 57 4 | 1 6 8 | | 29 68 1 | 29 3 68 | 5 4 7 | :------------------------+------------------+-----------: | 3 78 5 | 478 478 2 | 6 1 9 | | 789 2 789 | 678 1 678 | 3 5 4 | | 1 4 6 | 3 9 5 | 8 7 2 | '------------------------'------------------'-----------'

(5)r4c2 = (5-6)r4c5 = (65)r1c59 - 5r2c(9=1) => -5 r1c2,r5c1; stte

or more symmetrically written:

5r4c(2=5) - 6r(4=1)c5 - 6r(1=2)c9 - 5r2c(9=1) => -5 r1c2,r5c1; stte

by **SpAce** » Wed Jan 30, 2019 3:50 pm

Cenoman wrote:(6)r4c5 = 6r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2 => -5 r4c5; ste

Nice symmetry. I think that pattern could be called S-Wing. You could also eliminate -6 r4c2.

Btw, these types of chains with several bilocation hops can be considerably compacted with the 3D notation (only matters if they're much longer anyway), but I think it also makes the symmetry and the wing-nature stand out better:

I (just now) wrote:6r(4=1)c5 - (6=5)r1c9 - 5r(1=4)c2 => -6 r4c2, -5 r4c5; stte

But yeah, I know you all hate it so enough of that

by **Ngisa** » Wed Jan 30, 2019 7:29 pm

Code: Select all: +----------------------+---------------------+--------------+ | 258 8-5 248 | 1 c46 9 | 7 3 56 | | e57 1 e47 | d46 2 3 | 9 8 56 | | 6 9 3 | 5 78 78 | 4 2 1 | +----------------------+---------------------+--------------+ | 4 a5678 78 | 678 b5678 1 | 2 9 3 | | 279-5 3 279 | 29 57 4 | 1 6 8 | | 29 68 1 | 29 3 68 | 5 4 7 | +----------------------+---------------------+--------------+ | 3 78 5 | 478 478 2 | 6 1 9 | | 789 2 789 | 678 1 678 | 3 5 4 | | 1 4 6 | 3 9 5 | 8 7 2 | +----------------------+---------------------+--------------+

Almost like the others
(5)r4c2 = (5-6)r4c5 = r1c5 - (6=4)r2c4 - (4=75)r2c13 => - 5r1c2,r5c1; stte

Clement

by **Cenoman** » Wed Jan 30, 2019 11:20 pm

SpAce wrote:Nice symmetry. I think that pattern could be called S-Wing. You could also eliminate -6 r4c2.

Btw, these types of chains with several bilocation hops can be considerably compacted with the 3D notation (only matters if they're much longer anyway), but I think it also makes the symmetry and the wing-nature stand out better:

I (just now) a écrit:
6r(4=1)c5 - (6=5)r1c9 - 5r(1=4)c2 => -6 r4c2, -5 r4c5; stte

But yeah, I know you all hate it so enough of that

BTW there was a minor typo in my chain. Should read "(6)r4c5 = r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2 => -5 r4c5, -6r4c2; ste"
Thanks, Space for the name and the extra elimination.

You seem to hold Eureka notation responsible for the poor compacity (here and in other posts). This is not the main reason. I hold the cell reference rncn for responsible.
I post also on other sudoku forums, especially http://sudoku.com.au and www.sudokuwiki.org
Their cell reference systems are in the hidden diagrams:

Hidden Text: Show

The above chain in rncn: (6)r4c5 = r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2 => 35 characters (blanks uncounted)
The same chain in "sudokuwiki" (6)D5 = A5 - (6=5)A9 - A2 = (5)D2 => 25 characters (blanks uncounted)
The same chain in "chessboard" (6)e6 = e9 - (6=5)i9 - b9 = (5)b6 => 25 characters (blanks uncounted)
The same chain in "3D notation" 6r(4=1)c5 - (6=5)r1c9 - 5r(1=4)c2 => 29 characters (blanks uncounted)
My computer is able to write any of the first three (easy translation from one to any other). No doubt it could also learn the fourth...

You can see that the Eureka notation of AICs is compatible with other cell references than rncn. The lack of compacity is specific to this site, due to the original very old and very bad decision to use rncn reference. I do not suggest to reverse the decision taken so many years ago. We just have to bear the heaviness of rncn reference...
You believe that "3D notation" is close to Eureka notation. To me, it is not. The rules for using parenthesis are such that, when you see parenthesis, you know that that inside, you find digits and only digits; and this is true even for single digits. For people reading Eureka AICs for years, "3D notation" is confusing.

I will not enter endless discussions on this topic. I state my position for once. Feel free to write chains in 3D notation. On my side, I feel free not to read them.

by **SpAce** » Thu Jan 31, 2019 3:10 am

Cenoman wrote:You seem to hold Eureka notation responsible for the poor compacity (here and in other posts). This is not the main reason. I hold the cell reference rncn for responsible.

I've also mentioned that reason before. However, compactness is not the only reason why I like the 3D notation. You do realize that it's not in any way tied to rncn? It could be used just as well with the J9 or chessboard notations to make them even more compact.

Btw, how do you deal with boxes in those notations? The one nice thing about rncn is that it's compatible with bnpn (which also works well with 3D). I've also mentioned before that I could easily support switching to J9 or chessboard if I knew how the box notation was resolved. Then again, I don't think either of them is perfect. What about David's suggestion to use (obviously different sets of) letters for both rows and columns? Then your precious digits would have an even more special status, probably making 3D less confusing as well.

I post also on other sudoku forums, especially http://sudoku.com.au

Then you should know where I partly picked the idea for 3D. On that site you can find examples of Steve K using an even more compact form of it. I liked the idea but not his implementation, so I modified it a bit. Another influence was Denis' ideas, of course (though I liked his notation even less). The most credit I'd probably give to Allan Barker's cube thinking, which I think is reflected pretty well in the 3D notation.

The above chain in rncn: (6)r4c5 = r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2 => 35 characters (blanks uncounted)
The same chain in "sudokuwiki" (6)D5 = A5 - (6=5)A9 - A2 = (5)D2 => 25 characters (blanks uncounted)
The same chain in "chessboard" (6)e6 = e9 - (6=5)i9 - b9 = (5)b6 => 25 characters (blanks uncounted)
The same chain in "3D notation" 6r(4=1)c5 - (6=5)r1c9 - 5r(1=4)c2 => 29 characters (blanks uncounted)

And 23 characters (blanks uncounted) if you use 3D with either J9 or chessboard. If blanks are counted (as they should because they affect the total length), then it's 27 (3D with J9/CB) vs 33 (normal J9/CB). So, you're comparing apples and oranges. I think this represents the truth better:

Code: Select all: (6)r4c5 = r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2 (Eureka rncn) : 43 (6)D5 = A5 - (6=5)A9 - A2 = (5)D2 (Eureka J9) : 33 (6)e6 = e9 - (6=5)i9 - b9 = (5)b6 (Eureka CB) : 33 6r(4=1)c5 - (6=5)r1c9 - 5r(1=4)c2 (3D rncn) : 33 6(D=A)5 - (6=5)A9 - 5(A=D)2 (3D J9) : 27 6e(6=9) - (6=5)i9 - 5b(9=6) (3D CB) : 27

So, even with rncn it's just as short as normal J9/CB (in this case), and when using either of those, it's clearly shortest (of course).

You can see that the Eureka notation of AICs is compatible with other cell references than rncn.

I knew that, of course. You should also see that so is 3D.

You believe that "3D notation" is close to Eureka notation. To me, it is not. The rules for using parenthesis are such that, when you see parenthesis, you know that that inside, you find digits and only digits; and this is true even for single digits. For people reading Eureka AICs for years, "3D notation" is confusing.

I understand that very well. One can still argue that it's intrinsically better because it has less redundancy and more consistency along every axis, in addition to being more compact. The only downside is readability for those used to vanilla Eureka, but even that could be improved, for example by using different kinds of brackets for the different axes: eg. (1=2) for digits, [1=2] for rows, {1=2} for columns. Even now there's no ambiguity because the nrc-ordering is fixed, but that would provide an additional visual cue. The chain in question would then be:

6r[4=1]c5 - (6=5)r1c9 - 5r[1=4]c2

I will not enter endless discussions on this topic. I state my position for once. Feel free to write chains in 3D notation. On my side, I feel free not to read them.

No one's forcing you to do anything. I'm just surprised you chose to use a misleading comparison table to make your point.

PS. I'd actually be more interested in your take on the multi-BUG issue. You're very comfortable with various kinds of DPs, and I bet you could rather easily modify your program to run some tests to provide a better understanding of the situation. Finding a single counter-example would disprove my hypothesis (on which I have very little confidence). Or maybe you could see a way to disprove (or prove!) it without even running any tests.

by **SpAce** » Thu Jan 31, 2019 6:08 pm

I'm beginning to think my favorite notation would be 3D-aK, where only actual digits (i.e. n-axis) are digits. Then rows are lower-case {a..k} and columns upper-case {A..K} (minus 'i/I' and 'j/J' for readability). That's actually very readable, because when the order is NrC the lower-case letters in the middle provide perfect contrast, and there's no confusion about which axis is being linked even if the same kind of brackets are used. That way your chain would translate:

(6)r4c5 = r1c5 - (6=5)r1c9 - r1c2 = (5)r4c2

<=>

6(d=a)E - (6=5)aK - 5(a=d)B

That's about perfect, I think. Too bad the 'B' is still easy to mix with '8' and '3', but I could live with that (I could live with 'j/J' too). The only question remaining is how to deal with boxes.

by **Cenoman** » Thu Jan 31, 2019 11:26 pm

SpAce wrote:I'd actually be more interested in your take on the multi-BUG issue. You're very comfortable with various kinds of DPs, and I bet you could rather easily modify your program to run some tests to provide a better understanding of the situation. Finding a single counter-example would disprove my hypothesis (on which I have very little confidence). Or maybe you could see a way to disprove (or prove!) it without even running any tests.

Believe it or not, my DPs are manually spotted, including their guardians. So I can't offer any run for tests.
Anyhow I have read blue's contribution. I'm far from having his skills. I can't even make a difference between apples and oranges...

by **SpAce** » Fri Feb 01, 2019 1:40 am

Cenoman wrote:Believe it or not, my DPs are manually spotted, including their guardians.

Ok, great! In that case my suggestion was even more of a compliment, as your manual skills in that department are obviously outstanding. I've actually thought (without any judgment) that you'd use your program to help with those, because you're so good at not only spotting DPs but also finding the optimum guardian sets (internal/external/mixed) for the simplest eliminations. Most of what I know about that stuff I've learned from you, and I still have a ways to go.

I have great respect for your skills, which is why it saddens me that there's obviously been some tension between us lately. I don't honestly know where and why it started on your part, because I thought we were on friendly terms until your totally surprising reactions to the SDC discussion a while back. That really caught me off-guard because I hadn't seen it coming. The only explanation I could see back then was that you'd been building up resentment until you could no longer keep the lid on it. I just don't know what caused that resentment in the first place, as I can't remember any heated arguments between the two of us. All I know is that things haven't been the same since then, on my part either, which seems to result in sharper than necessary commenting by both parties.

I can't even make a difference between apples and oranges...

Referring to the previous paragraph, I refuse to take the bait this time.

by **SpAce** » Fri Feb 01, 2019 2:07 am

One more thing about those notations. I finally found the post where David voiced his wishes:

David P Bird wrote:When I gain world dominance rows will be labelled a to j, omitting i and columns p to x, allowing the digits to stay as they are so we can lose n, r, & c.

So, that's yet another possibility. I'd prefer my 9aK (pronounced "gack", lol) suggestion, though, because it's simpler with the same benefits. (If need be, the P/p to X/x range could be used to replace bnpn, though.) I do agree with David that it would be great if digits were only used for the n-axis (whether 3D is used or not).

An interesting variant might also be one with the n-digit sitting in the middle, between two lower-case letters: rNc.

The New Sudoku Players' Forum

January 30, 2019

January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019

Re: January 30, 2019