Fish-finding

Post the puzzle or solving technique that's causing you trouble and someone will help

Fish-finding

Postby Yogi » Sat Apr 09, 2022 9:24 pm

Does anyone here use the ‘Fishing With Lines’ technique described in Sudocue, or is there a thread on it in this forum?
Like others, it seems interesting to me but I have not been able to really get a handle on it from the description in that website.
It seems to be related to the simple way of identifying Empty Rectangles with one vertical and one horizontal line.
Maybe this puzzle would be a useful example to work with:

64.3....2..9..2..6253.69.1..35..6..8..6.7.5..4.....16..62594...5.46..2..39...7645

Code: Select all
+---+---+---+
|64.|3..|..2|
|..9|..2|..6|
|253|.69|.1.|
+---+---+---+
|.35|..6|..8|
|..6|.7.|5..|
|4..|...|16.|
+---+---+---+
|.62|594|...|
|5.4|6..|2..|
|39.|..7|645|
+---+---+---+
User avatar
Yogi
2017 Supporter
 
Posts: 349
Joined: 05 December 2015
Location: New Zealand

Re: Fish-finding

Postby StrmCkr » Sun Apr 10, 2022 2:34 am

Some do, some teach, the rest look it up.
stormdoku
User avatar
StrmCkr
 
Posts: 1431
Joined: 05 September 2006

Fish-finding

Postby Yogi » Sun May 08, 2022 5:02 am

I finally found an example that can be worked through with this method. Whether it’s worthwhile for Swordfish remains to be seen. Maybe it would be more useful for bigger fish. The important thing in fish-hunting is not to be looking for fish! I’m told that only about 3% of puzzles have full 3x3 Swordfish, leaving the vast majority to include at least one row or column in the Base Set as being only a Conjugate Pair. This means it is much more useful to check and see if any CP you find may be part of a fish.

1.85..2345..3.2178...8..5698..6.5793..59..4813....865298.2.631.......8.....78.9..

Code: Select all
+---+---+---+
|1.8|5..|234|
|5..|3.2|178|
|...|8..|569|
+---+---+---+
|8..|6.5|793|
|..5|9..|481|
|3..|..8|652|
+---+---+---+
|98.|2.6|31.|
|...|...|8..|
|...|78.|9..|
+---+---+---+


Here we have a Swordfish in candidate 4 in rows 247. There are currently two instances of 4 already solved so, according to Ruud’s description, there should be 7 lines to be drawn, which must not include the rows or columns of the solved 4s, nor have two lines intersecting any unsolved candidate 4. Now think about it. Where could these lines be? Once you realise that, since this case is a Swordfish in rows and we won’t be trying to eliminate candidates in the Base Sets, then the Swordfish itself supplies three of these lines in those rows. That leaves only four other lines to draw for this exercise, and they can be spotted in columns 1468. Then indeed it does follow that you can eliminate all candidate 4s which have not had a line drawn through their cell.

However, is this approach better or easier than working up from CPs?
User avatar
Yogi
2017 Supporter
 
Posts: 349
Joined: 05 December 2015
Location: New Zealand

Re: Fish-finding

Postby StrmCkr » Sun May 15, 2022 1:01 am

http://forum.enjoysudoku.com/the-ultimate-fish-guide-t4993.html

is another guide people use for constructing fish:

However practically from a manual solver perspective can be easily lost in translation

this presentation is the easiest i can describe for your consumption
something akin to this
post

consider the normal Mark up of a grid in spacial definitions Refereed to as

RC markup which stands for the Row,Col Position listing N where N is the solved Value of the "Grid"
Some Might list the Potential placements {pencil marks}

Pencil marks aren't derived how you think they are

Pencil Marks are derived From the spaces:

Row Number Listing Col
Col Number Listing Row
Box Number : listing Square

which are the positions left "ON" by placements of N in R,C as each placement turns off 8 locations in box,row and col

when Rn,Cn,Bn, space is combined with RC we Now see the original Grid + Potential markups

Rn,Cn, space is where we find basic Fish and yes most of the time very few are full constructs.

how does this work?
here is the PM Grid view
Code: Select all
.-----------------------.-----------------.------------.
| 1     679      8      | 5   679    79   | 2  3   4   |
| 5     469      469    | 3   469    2    | 1  7   8   |
| 247   2347     2347   | 8   147    147  | 5  6   9   |
:-----------------------+-----------------+------------:
| 8     124      124    | 6   124    5    | 7  9   3   |
| 267   267      5      | 9   237    37   | 4  8   1   |
| 3     1479     1479   | 14  147    8    | 6  5   2   |
:-----------------------+-----------------+------------:
| 9     8        47     | 2   45     6    | 3  1   57  |
| 2467  1234567  123467 | 14  13459  1349 | 8  24  567 |
| 246   123456   12346  | 7   8      134  | 9  24  56  |
'-----------------------'-----------------'------------'


lets convert this grid to an Rn,Cn,Bn view

my linked picture lets focus on the Row - # x col

https://www.mediafire.com/view/4a5lha1da5xdohr/Capture4.JPG#

top down is the Row {0 - 8}
left to Right represents the "Digit" 1 - 9
stored Cells is the Col the digit is found in

how to use these spaces
if you like "drawing lines" or slicing this is what you are doing with lines each line you draw marks a "hit" on cells thats how the Digit you are evaluating
in a Row sword fish are operating you are drawing lines down the Columns to see how many hit with at most 3 per col

focus on Digit "4"
in my chart
Code: Select all
R       Digit {4}
0 |
1 |    1 2   4
2 |  0 1 2   4 5
3 |    1 2   4
4 |
5 |    1 2 3 4
6 |    1 2
7 |  0 1 2 3 4 5  7
8 |  0 1 2     5   7


we are looking for 3 rows with exactly the same "set" or sub set with max of 3 Cols listed
which happens in Row 1,3,6 as they all contain exactly the set of {1,2,4}

which produces: {in hodoku}
Swordfish: 4 r247 c235 => r3c235,r6c235,r8c235,r9c23<>4

in the view of Rn space the Rows that haves [124] are locked for digit "4" and all other Rows cannot use [1,24]
which updates to this:
Code: Select all
R       Digit {4}
0 |
1 |    1 2   4
2 |  0         5
3 |    1 2   4
4 |
5 |        3 
6 |    1 2
7 |  0     3   5  7
8 |  0         5   7


next example is :
using the Cn space Digit "4" you should be able to spot the

Swordfish: 4 c168 r389 => r3c235,r8c2345,r9c23<>4 {in hodoku}

very easily.
hint: Show
: 278 Cols 0,5,7


this method for fish finding is also applicable to finding "subsets" and was pointed out way back in 06
check out Col (3)- #s 1, 4 both list "5,7" Row we have a hidden pair of 1,4 @ R57C3
{ has no eliminations for us but its an example}

Bn space isn't the easiest one to utilize effectively which is why non basic fish aren't as easy to translate

It seems to be related to the simple way of identifying Empty Rectangles with one vertical and one horizontal line.

easy sure

look at box 2 digit 4
Draw a line down Col 5
Draw a line across Row 3
if all the digits of Box 2 for "4" land on the lines only we have a starting box

next
for each of the Cols that intersect row 3 draw a col line if any of these cols hold 2 "4"s with 1 in the row position
{hint there isn't any}

next check the Col line for any row that has 2 "4"s with at least 1 in the col position
we find one @ row 7
mark it and we have constructed the ER
Empty Rectangle: 4 in b2 (r7c35) => r3c3<>4

for me id suggest building it backwards
first find a row or column with a Bi local for digit x
then check the boxes as described above.
User avatar
StrmCkr
 
Posts: 1431
Joined: 05 September 2006


Return to Help with puzzles and solving techniques