holdout wrote:Each test you choose involves a cost (computer time).
I would agree with you when we were looking for the fastest way to solve a sudoku. However, this has never been my intention. What I am looking for is the shortest solving path that can be presented to a human player in the form of a series of hints or a solver log. The more complex sudokus (see my example) tend to produce very long chain reactions in the solver's log. (see rubylips' reply, I have not posted my log entries yet, they are too embarrassing).
If the cost to eliminate 7 candidates is big and the relative cost of eliminating 2 candidates is very small, then the 2-candidate test should be preferred.
For simplicity, yes. For finding the shortest path, we should look at the after-effects of both techniques. If the 7 eliminations break the puzzle into singles only, that path is shorter than the easier 2-candidate path that requires a more advanced technique before it finally breaks.
Who would do anything else except perfrom [SIC] eliminations based on singletons when faced with a new puzzle.
We are not looking at those difficulty levels here. Many solver programs can handle puzzles with singles only. However, once some T&E-based methods are required, results show greater variation. My aim is to find the shortest route to the solution, with human-style techniques only.
The last 2 points about ineffective test are also not relevant to the problem at hand. They are indeed relevant for solver-programmers and have been discussed in-depth on the sudoku programmers forum.
I do agree it is complicated.
"It is a strange fate, that we should suffer so much fear and doubt over so small a thing". (JRR Tolkien)
Ruud.