The New Sudoku Players' Forum

[Pupp]

I've done enough puzzles, with enough variations on the patten, to say it works every time I've used it.

Doesn't seem to matter if the pattern if 4 cells are grouped right next to each other (plus the 2 tails someplace aligned), or it the 4 cells are some distance apart. Doesn't seem to matter how close or far the tail cells are far from the square/rectangle cells of the pattern.

In any event, the pattern always consists of 6 cells. If a similar pattern exists with more cells, I'd name it something a bit differentt. I haven't looked into higher orders though.
Not really sure what to call it.

Perhaps Intersecting Diamond pattern?
Sorta like 2 diamonds on top of each other with the tail cells in different, 90 deg. orientations. (Looks more like a kite with 2 tails), but kite was already used as Sudoku patterns).

[denis_berthier]

I've done enough puzzles, with enough variations on the patten, to say it works every time I've used it.
[...]
Not really sure what to call it.

Don't worry, you don't need to call it any name. It doesn't work. I mean this pattern, with no additional conditions, doesn't allow to prove any elimination.

[Pupp]

The last puzzle had one instance where it didn't wotk, but I've done a bunch of puzzles where it worked flawlessly.

It's frustrating that I've used the technique well over a dozen times, maybe over 20 times and it worked flawlessly.

[denis_berthier]

The last puzzle had one instance where it didn't wotk, but I've done a bunch of puzzles where it worked flawlessly.

It's frustrating that I've used the technique well over a dozen times, maybe over 20 times and it worked flawlessly.

It doesn't work in the general case, when there are no block constraints, such as below:

Code: Select all

+-------------------------------+-------------------------------+-------------------------------+
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 23456789  123456789 !
! 123456789 23456789  23456789  ! 23456789  123456789 23456789  ! 23456789  123456789 23456789  !
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 23456789  123456789 !
+-------------------------------+-------------------------------+-------------------------------+
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 23456789  123456789 !
! 23456789  23456789  23456789  ! 23456789  123456789 23456789  ! 23456789  123456789 23456789  !
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 23456789  123456789 !
+-------------------------------+-------------------------------+-------------------------------+
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 23456789  123456789 !
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 23456789  123456789 !
! 123456789 123456789 123456789 ! 123456789 23456789  123456789 ! 123456789 123456789 123456789 !
+-------------------------------+-------------------------------+-------------------------------+

In cases you add block constraints (2 rows in the same floor or 2 columns in the same tower) , there are some trivial whip[1] eliminations, but they are not due to any new pattern.

[Pupp]

I was fun thinking I might have invented something, but it wasn't necessary to complete puzzles.

I've been rolling along Sudoku 10,000 Hard 1: I can finish at least 75% of the puzzles in under 10 minutes now.

I haven't quite mastered finned X-wings, which is usually the sticking point. I can easily solve finned X-wings, but frequently not easy to spot. Especially if the number has a lot of pencil marks on the board still.

[Pupp]

I don't know too much about statistics, outside of the normal statistics that you come across in college when getting a business degree. (Which wasn't much more than what I learned in high school to be honest.)

But I got to thinking, the reason I was having to much success was that the more pencil marks in a box, the more likely deleting the pencil mark in question would leave the puzzle still solvable.

That is, if there is a 50/50 chance with only two pencil marks in the cell, if you increase the number of other pencil mark numbers in a cell, obviously it becomes a much greater chance of being true for every added pencil mark number in the cell.

[aricjoshua]

You're correct that there are two comparable patterns at right angles, but one of them isn't a finned x-wing(*), and neither of them can delete anything in this PM on their own.
It's also true that on this grid, r4c6 is equal to 1.
Setting r4c6 = 1 does, in fact, imply a contradiction (using just Naked and Hidden Pairs involving other portions of the grid). This necessitates the application of the T&E(S2, n1r4c6) method.
However, I don't understand why n1r4c6 could be eliminated based only on the existence of the two orthogonal patterns (independent of other sections of the grid).

[denis_berthier]

In cases you add block constraints (2 rows in the same floor or 2 columns in the same tower) , there are some trivial whip[1] eliminations, but they are not due to any new pattern.

To be more precise, starting from exactly the same pattern of 1s as Pupp (a special case with 2 columns in the same tower), with no other clue:

Code: Select all

   +-------------------------------+-------------------------------+-------------------------------+
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   ! 23456789  23456789  23456789  ! 123456789 23456789  123456789 ! 23456789  23456789  23456789  !
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   +-------------------------------+-------------------------------+-------------------------------+
   ! 23456789  123456789 23456789  ! 123456789 23456789  123456789 ! 23456789  23456789  23456789  !
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   ! 123456789 123456789 123456789 ! 23456789  123456789 123456789 ! 123456789 123456789 123456789 !
   +-------------------------------+-------------------------------+-------------------------------+
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   +-------------------------------+-------------------------------+-------------------------------+

we get the trivial eliminations:
whip[1]: r2n1{c6 .} ==> r3c5 ≠ 1, r1c5 ≠ 1
whip[1]: b8n1{r9c5 .} ==> r6c5 ≠ 1, r5c5 ≠ 1

Code: Select all

   +-------------------------------+-------------------------------+-------------------------------+
   ! 123456789 123456789 123456789 ! 23456789  23456789  23456789  ! 123456789 123456789 123456789 !
   ! 23456789  23456789  23456789  ! 123456789 23456789  123456789 ! 23456789  23456789  23456789  !
   ! 123456789 123456789 123456789 ! 23456789  23456789  23456789  ! 123456789 123456789 123456789 !
   +-------------------------------+-------------------------------+-------------------------------+
   ! 23456789  123456789 23456789  ! 123456789 23456789  123456789 ! 23456789  23456789  23456789  !
   ! 123456789 123456789 123456789 ! 23456789  23456789  23456789  ! 123456789 123456789 123456789 !
   ! 123456789 123456789 123456789 ! 23456789  23456789  123456789 ! 123456789 123456789 123456789 !
   +-------------------------------+-------------------------------+-------------------------------+
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   ! 123456789 123456789 123456789 ! 23456789  123456789 23456789  ! 123456789 123456789 123456789 !
   +-------------------------------+-------------------------------+-------------------------------+

and that's all we can get.

[Pupp]

After taking a short break, I started hard 1 puzzles again.

Doing much better. If I get stuck, I hit the analyze button.

So far I haven't messed up a puzzle, and usually I missed an X-wing, or missed a claiming.

Sadly, Sudoku 10,000 just kinda lumps certain techniques under the "claiming" umbrella with no indication of what you need to do to eliminate more pencil marks.

I did eventually realize that doubles, triples and quadruples are under that umbrella term. It's easy to spot a quadruple when watching a youtube video... it's often already highlighted by the person explaining. Much more difficult when your looking at an entire board filled with mostly pencil marks and few solved cells outside of the starter cells.
But practice makes perfect, so I'm just rolling along, doing all the Hard 1 puzzles on Sudoku 10,000.

It did get easier when I realized that. Knowing what to look for is over half the battle.

[Pupp]

I was stuck on a puzzle with the same pattern.

I figured out I was probably using a forcing chain to figure out which cell to fill in.

So there is no specific cell that's correct every time. But I do look at 2 different paths. One for each rectangle+1, then use that to find the common denominators. I actually end up with 2 cells I know I can fill in. That being said, I think it's better described as a simplified pattern for a forcing chain.

Once I identify the cells, it's easy to mentally blitz the forcing chain to double check my train of thought.

It only works as an end stage for a number, and it's a last resort thing. Sorta like the BUG+1 is only useful at a late stage.

But, I'll have to come across this many more times to be certain it's not an educated guess. But this represents a small progress on this thread. At least I figured out that it's not an exact cell every time I come across the pattern.

I do know that I should have been able to finish the puzzle with far more basic techniques. But after staring at the puzzle too long, I got tired of looking at it.

I want to mention, so I don't forget: This time, the correct cells did NOT leave 2 separate rectangles of said number.

(No idea about prior attempts, since I didn't take notes.)

I'm not saying this works every time, but it works enough to pique my curiosity to see if I can nail down what elements of pre-existing techniques I'm using. The pattern doesn't show up too often. So this might take a very long time to fully investigate.

EDIT: after some digging I think it's possibly related to a 3 link X-chain. Next time I come across the pattern, I'll watch videos on both the Turbot Fish, and the 2 String Kite.

[Pupp]

I'm getting better at puzzles, so I think getting stuck and having to look at that pattern at all, will go away.

I do still see the pattern, but I ignore it l, unless I'm really stuck without seeing a path forward.

The last puzzle I did today had that pattern, but I finished the puzzle in a little over 3 minutes, and I just skipped the pattern and went to a different number.