![[info]](http://l-stat.livejournal.com/img/userinfo.gif){kind=small}
Here is me solving a very easy (36 given) sudoku very fast (58 seconds).
Here is me describing a set of ways to get started solving any sudoku using this very easy one. There are a couple subtleties that would come out if the puzzle were harder, but in hearing my own description, I'm beginning to realize that I'm getting even better at the puzzle than I was last year at the time of my world championship as I am really formalizing for myself the best approaches to the puzzle.
As I also brought up in my seminar on Sunday, I never start searching by number (looking for 1 to 9 placements in order by focusing on specific digits), but instead look for one of two things when I get started. These are almost exclusively - (1) repeated digits in chutes (which I'll then crosshatch as needed), and then (2) interesting geometries that seem to have a lot of forced placements or forced either/or situations. Included in interesting geometries are obviously rows/columns/boxes that are almost full, and if they are very full, I will always start by looking there and storing the information they might provide.
I like to think of my approach to a puzzle as being an organic, thoughtful consideration of what I see, what from experience seems the best place to work, and then starting there. Whenever I make a placement, I tend to find something that placement must have forced which means I might jump from a digit to another digit, or place more instances of that single digit. Progress leads to progress, so wherever you place a number (that row/column/box) is a good place to keep looking. Anyway, my fluid process seems to be one that should beat any more automated process a solver uses, even if they are capable of writing very fast and never missing obvious placements in their searches.
Here is me describing a set of ways to get started solving any sudoku using this very easy one. There are a couple subtleties that would come out if the puzzle were harder, but in hearing my own description, I'm beginning to realize that I'm getting even better at the puzzle than I was last year at the time of my world championship as I am really formalizing for myself the best approaches to the puzzle.
As I also brought up in my seminar on Sunday, I never start searching by number (looking for 1 to 9 placements in order by focusing on specific digits), but instead look for one of two things when I get started. These are almost exclusively - (1) repeated digits in chutes (which I'll then crosshatch as needed), and then (2) interesting geometries that seem to have a lot of forced placements or forced either/or situations. Included in interesting geometries are obviously rows/columns/boxes that are almost full, and if they are very full, I will always start by looking there and storing the information they might provide.
I like to think of my approach to a puzzle as being an organic, thoughtful consideration of what I see, what from experience seems the best place to work, and then starting there. Whenever I make a placement, I tend to find something that placement must have forced which means I might jump from a digit to another digit, or place more instances of that single digit. Progress leads to progress, so wherever you place a number (that row/column/box) is a good place to keep looking. Anyway, my fluid process seems to be one that should beat any more automated process a solver uses, even if they are capable of writing very fast and never missing obvious placements in their searches.
2 comments | Leave a comment