ssxsssxs wrote:But I see you often posting puzzles whose SE-rating are higher than 10.0. And from the puzzles you posted, it seems that the sudoku.exe had been searching at least 18 clues, but in my computer, it is at most 13 clues for 5 days. I don't know why.
Sorry, I made a wrong answer to your post, having in mind Sudoku Explainer. (A beach side problem I guess).
Generation of puzzles in the pattern game is somehow the key point. Each of us has it's own process, but for me the scan or the random generation are unefficient processes.
The basic strategy, to be efficient, is the vicinity search. You just change 'n' digits or ('1' to 'n' digits) and you start a new cycle with selected seeds depending on what you try to achieve.
You can also, knowing that in the game you always have a symmetry, try to find some symmetric (why not the so called symmetry of given) distribution of the given, which is much faster than a scan (say less than one hour is you are looking for the symmetry of given). You then have a lot of seeds usually giving a good cover of the field.
Playing with the 2 main parameters of a vicinity search : depth of the search, and selection of seeds gives enough variations from one player to another one.
I tried also some filtered scans, but usually with a poor yield.
My basic process is a vicinity '1' to '3', with deeper searchs depending on the pattern.
The seed selection is generally oriented in my case towards the high ratings, but other players can have other priorities