Myth Jellies wrote:I am not trying to demean Sherlock in any way.
I understand. And I love your POM method. I never knew that it existed until this discussion. That is the type of creative thinking that I enjoy.
Now for Sherlock. Although it is poignant, I hope that you take it in the humor that it is meant.
My Sherlock method of solving Sudoku puzzles requires a slight overhaul in the way that you think about the problem. Normally you are finding all of the digits that possibly fill a cell and then trying to eliminate all choices except for the correct one. When you switch to Sherlock mode you find all of the valid patterns for numbers within a line couplet and try to eliminate all patterns except for the correct one. It turns out that this change in approach can be a powerful tool for solving many, otherwise intractible, puzzles without resorting to forced chains.
The first step is to solve the puzzle as far as you can go using whatever normal methods you have mastered. Personally I solve for everything up to and including XY-wings, although I can miss an X-wing or two if I am really honest. The fewer possibilities you have for a given digit, the easier the next step becomes. Discovery of all XY-wings, swordfish, and other denizens of the deep, may fall out naturally during the next step of the process anyway, but then again, as the next step limits itself to two rows, it may not.
Step two is to take all the possible cells for a line couplet and, by applying the Sudoku rules, find all of the possible unique patterns for that line couplet. Any line pattern that does not contain a valid pattern can be removed as a possibility. I like to create a separate column of patterns for each line, X out any lines where that pattern is not possible, and represent each pattern with a unique letter. You can do this process for each line. Each line couplet that is uses that pattern will contain that pattern's letter. After you practice this method a few times, you can get a feeling for which lines will provide results in the next step quickly and just do those to save time. At this point, we can deduce whether the numbers remain valid - the number can be removed as a possible from any cell that is not contained in a valid pattern. If removing the possibles from the unused cells does not break things open then procede on to step three.
Step 3 uses derived algebraic-logic relationships between patterns of different lines and relies on the fact that only valid solution patterns will propigate smoothly when you merge these patterns together. Assuming you can find all of the logic equations you need, invalid patterns will ultimately be crossed out after the merging process. However, this is theoretical, as I have not yet taken it to its logical endpoint.
Myth, I hope that you have enjoyed this explanation. I will give a more serious address to your concerns at the weekend. But before I do, could you give me a generalised definition of "trial and error"? I think that the argument of whether Sherlock is trial and error will only be solved if we can agree on a tighter definition, i.e. one that I cannot apply to simpler methods. Until your last post, I thought that you were arguing that determining all permutations of a subset of cells constituted trial and error.
Brendan