The New Sudoku Players' Forum

[dobrichev]

Full question: According to your understanding what a "Human-style" Sudoku Solving Technique is, are these techniques invariant to puzzles' isomorphic transformations?

This is an attempt to cure Marvin's syndrome.

[Leren]

Well, I would provisionally say yes, as long as they are correctly and fully coded for all of the relevant Isomorphic transforms.

One fun case was my (partially implemented) human style PX solver. Interestingly it found a Suq de Coq move for one PX puzzle.

But what about the F transform of the same puzzle ? What does the F Transform of a Sue De Coq look like? I don't even want to think about that

Leren

[tarek]

In principle an isomorphic transformation does not destroy the technique.

The solving eye however can miss the same technique when an isomorphic transformation takes place. That may end up causing the solution to become shorter or longer!!!

Computers performing Human style solving will encounter this issue all the time unless they apply all available techniques at the same time (depth first rather than breadth 1st).

gsf’s sudoku has a -j option which transforms the puzzle isomephically to compare solving paths. I used that in the pure x-wing/swordfish ... etc to show the of the technique is absolutely needed even with isomorphic transformations using a digital human-style solver

Tarek

[dobrichev]

The solving eye however can miss the same technique when an isomorphic transformation takes place.

That is my point.
Computer will solve in the way it is programmed - nothing interesting.

Are there examples for techniques that are far below the personal questionable zone human <-> non-human, but are within or above this zone when the puzzle is scrambled?

One fun case was my (partially implemented) human style PX solver. Interestingly it found a Suq de Coq move for one PX puzzle.
But what about the F transform of the same puzzle ? What does the F Transform of a Sue De Coq look like? I don't even want to think about that

It must have equivalent in the "PX" world and in all worlds where the F transformation is valid. Is it simpler is subject of doctorate thesis.
But going this way we can rapidly reach ambiguity in investigating the null puzzle consisting of 0 x 0 cells and having arbitrary symmetries and no givens

[qiuyanzhe]

Things are basically the same after transformations, but how the solving methods are seen may vary. If we have givens r1c3467=1234and r3467c1=5678 in a puzzle, the Naked Single is more obvious than r1c3467=5826 and r3467c1=4193. Similar things also exist for subsets or chains, etc. As for computers, we could normalize the puzzle first to avoid any difference caused by transformations.

Any solving method that does not hold after.transformations is pseudoscience ⊙﹏⊙

[eleven]

I can't see much difference for the usual basic and advanced techniques, though for beginners the pattern they learn is easier to spot than a scrambled version.
But a big difference is for digit symmetric puzzles. If i don't know that there is one, i would never spot it, if it is not in normal form.
On the other hand most advanced puzzles are solved with chains, where scrambling is no problem at all.

[champagne]

If I except rules based on the uniqueness property, classical logical rules can be applied one by one or in block without effect on the overall "difficulty"
So morphing a puzzle does not change the "difficulty" subject to all remarks made above on the way a player looks at the picture.

In the "so called low ratings", the problems starts with the URs ULs, the first rule using the uniqueness property
Applying or not a UR/UL changes significantly the remaining difficulty
When several interleaved URs/ULs are there the way/order in which they are processed changes the final difficulty

Similar comments could be done on BUGs ...

Chains and complex chain nets don't use the uniqueness property, so even if they are "very complex", they don't affect in theory the overall difficulty,
But then, player or computer, the chances to cover all the field are close to nil, so morphing will in fact often change the apparent difficulty

And, I fullly agree with eleven, if the puzzle is not in the best form, the symmetry of given (uniqueness rule) will not be seen

[dobrichev]

Just to mention that initially this was posted in Coffee bar section and I don't expect much serious discussion. More like "See what once happened to me...".

Any solving method that does not hold after.transformations is pseudoscience ⊙﹏⊙

Agree, but the question makes sense to me because the definition of "human-style solving method" isn't scientific.

[StrmCkr]

yes, all moves are mutable to transformations.

however they may not exhibit the same number of eliminations or even have any eliminations based on the new layout.

some moves do not survive transformation on specific puzzles as the "eliminations" zones can land on cells that are already eliminated or solved

I've seen this for every human style of move: fish, almost locked sets, subsets, chains etc where some/all of the eliminations are no longer applicable, but that doesn't mean the move is invalid as the construct is still valid, instead it becomes ineffective.

[eleven]

I don't understand that. What transformations do you mean?
Can you give an example, please ?