The New Sudoku Players' Forum

[gsf]

28 clue, pi symmetry, symmetric minimal, 1 swordfish, the rest singles

Code: Select all

1 . . | 9 3 . | . 7 4
. 2 . | . . 7 | . . .
. . . | . 1 . | . 8 .
------+-------+------
9 3 . | . 4 . | 8 . .
. 1 . | . . . | . 6 .
. . 6 | . 8 . | . 3 9
------+-------+------
. 7 . | . 6 . | . . .
. . . | 5 . . | . 4 .
5 9 . | . 7 1 | . . 6


[gsf]

I'm running over gfroyle's 18s with (H4||T4||W4) (no swordfish) right now

superior plus: 2814 puzzles
no swordfish (quads/jellyfish only): 501 puzzles
22M puzzles ~35 min @ 3.2Ghz per run

[fermat]

I'll keep trying.

Code: Select all

. . . | . . 8 | 9 . .
. . . | . 9 . | . 2 .
. . . | 2 . . | 6 . 1
---------------------
. . 5 | 6 . 3 | . . 8
. 1 . | . 2 . | . 9 .
2 . . | 4 . 9 | 3 . .
---------------------
8 . 7 | . . 6 | . . .
. 2 . | . 3 . | . . .
. . 3 | 8 . . | . . .


This one is not an effortless, many techniques work and eliminate others.
I think that the swordfish usually is a factor.

That's beyond the scope of this thread fermat. You've gone from one extreme to another.

You could use my software to decide if your puzzle qualifies:
http://uk.geocities.com/aidan_001/software.html

Simple Sudoku solves this. Effortless extremes rules say that makes it too easy. (Do humans use colouring as anything but a last resort?)

So, it is too easy for them but way to hard for you? Did you mean it is way too easy?

Edited to remove an "ad hominum".

[Ocean]

I'm running over gfroyle's 18s with (H4||T4||W4) (no swordfish) right now

superior plus: 2814 puzzles
no swordfish (quads/jellyfish only): 501 puzzles
22M puzzles ~35 min @ 3.2Ghz per run

Thanks for the statistics. The distribution is in the same range as observed in my (much smaller) collections. (Except that I did not catch the Jellyfish yet). A few questions:
Do you know how many of the 501 that are Jellyfish and how many are Quads (and eventually how many are both)?
Did any of the S+ have symmetric clue patterns (eventually after transformation)?
Did you check for (or find) 'combinations' (ex. JF + HQ), and 'multiples' (ex. 2 x JF, or 2 x HQ) ? [A list of such puzzles would be interesting].

[gsf]

Do you know how many of the 501 that are Jellyfish and how many are Quads (and eventually how many are both)?

there were 2814 superior+ puzzles in the 22M sudoku18 catalog
2332 required swordfish

Code: Select all

sudoku -S -f- -F%a -e "W3" superior18.dat

391 required naked quad

Code: Select all

sudoku -S -f- -F%a -e "T4" superior18.dat

10 required hidden quad

Code: Select all

sudoku -S -f- -F%a -e "H4" superior18.dat

100 required jellyfish

Code: Select all

sudoku -S -f- -F%a -e "T4" superior18.dat

Did any of the S+ have symmetric clue patterns (eventually after transformation)?

I haven't coded a symmetrizer given a valid sudoku -- has anyone?

Did you check for (or find) 'combinations' (ex. JF + HQ), and 'multiples' (ex. 2 x JF, or 2 x HQ) ? [A list of such puzzles would be interesting].

1 required two separate steps with superior+ quad techniques (not counting swordfish, this one takes 2 separate naked quads)

Code: Select all

sudoku -S -f'# %Q\n%#gg' -e "(W4+T4+H4)>1" superior18.dat
#   296 FNBT      C18.m/F12.40/N6.23/B2.4.2.2/T3.9.1.0.2

. 7 .  . 4 .  . . .
. . .  . 3 .  . . 5
. . .  8 . .  1 . .

4 . 3  . . .  . 2 .
. . .  5 . 1  9 . .
6 . .  . . .  . . .

. 5 .  9 . .  7 . .
2 . .  . . .  . 4 .
. . .  . . .  . . .


19 required a combination of swordfish and 1 quad (naked hidden jelly)
a separate filter confirmed only 1 swordfish needed in the combinations

Code: Select all

sudoku -S -f- -F%a -e "W3&&(W4+T4+H4)" superior18.dat

226 required 2 separate swordfish

Code: Select all

sudoku -S -f- -F%a -e "W3>1" superior18.dat

there were no other combinations required

[daj95376]

Excellent analysis gsf! I was particularly interested in your results for Hidden Quads.

----- ----- ----- ----- ----- ----- -----

Related to the topic but not to gsf's or anyone else's work.

When I place Naked Pair/Triple/Quad detection before Locked Candidates (1&2), many more n-tuples are reported by my solver than if I place Locked Candidates (1&2) detection before these n-tuples. It's something to consider.

[Ocean]

there were 2814 superior+ puzzles in the 22M sudoku18 catalog ...

Thanks for the thorough answer - both the numbers and the methods!

[gsf]

there were 2814 superior+ puzzles in the 22M sudoku18 catalog ...

Thanks for the thorough answer - both the numbers and the methods!

I just re-read the post and should clarify that the posted -e filter expressions assume
that the superior+ filtering was already done (i.e., superior18.dat contans only the 2814 superior+ puzzles)

I used this to do the initial filter

Code: Select all

sudoku  -f'# %Q\n%v' -q-XYG -e 'V&&(T4||H4||W3||W4)' sudoku18/sudoku18.dat > superior18.dat


[Ocean]

... the superior+ filtering was already done (i.e., superior18.dat contans only the 2814 superior+ puzzles)

Of course!

daj95376's comment reminded me of a question I forgot to ask:

Could any of the S+ puzzles be solved with alternative S+ techniques (ex. SW or HQ)?

[gsf]

Could any of the S+ puzzles be solved with alternative S+ techniques (ex. SW or HQ)?

the default order, which can be changed with -q, checks T4H4W4 in order,
so if W4 and !T4 !H4 were used then W4 was the only S+ technique
if H4 were used then it might be that W4 would also advance the solution

for the 391 that were solved by T4 (naked quads), 386 solve with H4 (hidden quad)
with the S+ order H4T4W4

T4H4 before W4 (jellyfish) tends to quash W4
with T4H4 off 1082 solve with W4

so the rarity of H4 and W4 in the puzzles I filtered is mostly an artifact of
the default T4H4W4 technique application order

[Ocean]

for the 391 that were solved by T4 (naked quads), 386 solve with H4 (hidden quad)
with the S+ order H4T4W4

... almost as expected...

T4H4 before W4 (jellyfish) tends to quash W4
with T4H4 off 1082 solve with W4

... but quite unexpected, given the previous statistics. I can not see how the number for W4 (previously 100) can possibly increase with more than 401 (391+10) with t4/h4 turned off, unless also cases where W4 can replace W3 are included. Maybe there is something I don't understand here.
P.S. Will be off the next two weeks, so eventual further comments from my side will be delayed...

[gsf]

T4H4 before W4 (jellyfish) tends to quash W4
with T4H4 off 1082 solve with W4

... but quite unexpected, given the previous statistics. I can not see how the number for W4 (previously 100) can possibly increase with more than 401 (391+10) with t4/h4 turned off, unless also cases where W4 can replace W3 are included. Maybe there is something I don't understand here.
P.S. Will be off the next two weeks, so eventual further comments from my side will be delayed...

aha, you get what you ask (filter) for
disabling T4H4W4 shows 2313 solved by W3 (swordfish)
but some of those swordfish come at later steps than the W4 (jellyfish) if it were endabled
in other words, disabling jellyfish and quads allow swordfish to solve some
puzzles that would otherwise be attributed to W4

[tso]

This puzzle -- rejected from the Effortless Extreme thread -- fits here. The very first move is Swordfish. Three Swordfish pairs are available at the start, though only one is required for the solution.

Code: Select all

+-------+-------+-------+
| 9 . . | 4 . . | 7 . . |
| . 6 . | . 3 . | . 5 . |
| . . 1 | . . 8 | . . 4 |
+-------+-------+-------+
| 2 . . | 8 . . | 1 . . |
| . 3 . | . 4 . | . 7 . |
| . . 7 | . . 9 | . . 5 |
+-------+-------+-------+
| 5 . . | 6 . . | 2 . . |
| . 2 . | . 8 . | . 9 . |
| . . 6 | . . 1 | . . 3 |
+-------+-------+-------+


The diagonal patterns are good places to find Swordfish. Out of 100 created with this pattern at random, 21 required Swordfish, two of which fit the rest of the criteria for this thread:

Code: Select all

X-wing, Swordfish
+-------+-------+-------+
| 8 . . | 6 . . | 4 . . |
| . 9 . | . 1 . | . 3 . |
| . . 7 | . . 8 | . . 5 |
+-------+-------+-------+
| 6 . . | 4 . . | 1 . . |
| . 1 . | . 2 . | . 8 . |
| . . 8 | . . 9 | . . 6 |
+-------+-------+-------+
| 5 . . | 9 . . | 2 . . |
| . 6 . | . 3 . | . 7 . |
| . . 4 | . . 5 | . . 3 |
+-------+-------+-------+

Naked Quads (or hidden pairs), X-wing, Swordfish (only the Swordfish is required for the solution)
+-------+-------+-------+
| 2 . . | 4 . . | 3 . . |
| . 7 . | . 8 . | . 4 . |
| . . 4 | . . 7 | . . 1 |
+-------+-------+-------+
| 4 . . | 1 . . | 2 . . |
| . 5 . | . 6 . | . 3 . |
| . . 7 | . . 8 | . . 9 |
+-------+-------+-------+
| 5 . . | 8 . . | 7 . . |
| . 6 . | . 5 . | . 9 . |
| . . 1 | . . 9 | . . 6 |
+-------+-------+-------+


The EXTREME setting of older versions of Simple Sudoku (4.1s for example) create puzzles on demand that fit this thread. They *all* have at least one Swordfish, though some also need colors and would therefore be disqualified.

(Do humans use colouring as anything but a last resort?)

Yes. I find Swordfish much more difficult to find than exclusions by simple coloring.


The current version's extremes also include xy-wings and multiple colors - swordfish appear less frequently. Puzzles are much less likely to fit this thread.

Here are a few at random from 4.1s:

Code: Select all

+-------+-------+-------+
| 9 . 3 | 8 . . | . 6 . |
| . . . | 6 . . | . 8 . |
| . . . | . . . | 9 . 5 |
+-------+-------+-------+
| . . 2 | 3 . . | . 5 . |
| 4 . . | . 6 . | . . 1 |
| . 5 . | . . 7 | 8 . . |
+-------+-------+-------+
| 3 . 5 | . . . | . . . |
| . 7 . | . . 3 | . . . |
| . 6 . | . . 1 | 2 . 9 |
+-------+-------+-------+

+-------+-------+-------+
| . . . | . . 3 | . 6 5 |
| 5 . . | 6 . . | 7 2 . |
| . . 6 | . . . | . . 3 |
+-------+-------+-------+
| 2 3 . | . 6 7 | 5 . . |
| . . . | 8 . 4 | . . . |
| . . 8 | 3 2 . | . 9 7 |
+-------+-------+-------+
| 6 . . | . . . | 1 . . |
| . 5 7 | . . 6 | . . 8 |
| 3 1 . | 5 . . | . . . |
+-------+-------+-------+

+-------+-------+-------+
| . . 7 | . . . | . . . |
| 5 3 . | 1 . . | . . 7 |
| 9 . 1 | . . . | 5 4 . |
+-------+-------+-------+
| . . . | 6 . 7 | . 3 5 |
| . . 3 | . 5 . | 9 . . |
| 8 6 . | 3 . 4 | . . . |
+-------+-------+-------+
| . 2 8 | . . . | 3 . 9 |
| 3 . . | . . 2 | . 7 4 |
| . . . | . . . | 1 . . |
+-------+-------+-------+


[fermat]

(Do humans use colouring as anything but a last resort?)

Yes. I find Swordfish much more difficult to find than exclusions by simple coloring.

OK, thanks.

I look at colouring as something to do when all patterns you know are not present. The swordfish I use is a "bent" x-wing.

When I see a "long jump" on a number I look for a match. Bent x is the one I spot easily.

Code: Select all

. . . | . . . | . . .
. . a | . A . | . . .
. . . | . . . | . . .
---------------------
. . . | . . . | . . .
. . . | . A . | A . .
. . . | . . . | . . .
---------------------
. . . | . . . | . . .
. . . | . . . | . . .
. . a | . . . | A . .


[ab]

here's a nice puzzle with 2 swordfish and several naked pairs. you only need one of the swordfish though!

Code: Select all

 6 . . | . 9 . | 3 . .
 . . 7 | . . . | 6 2 .
 . 2 . | . . . | . 1 5
 ------+-------+------
 . . . | 5 . 4 | . . .
 7 . . | . . . | . . 3
 . . . | 1 . 2 | . . .
 ------+-------+------
 4 1 . | . . . | . 7 .
 . 3 6 | . . . | 8 . .
 . . 5 | . 6 . | . . 9