by **r.e.s.** » Sun Oct 30, 2005 10:30 pm

I was a bit surprised that by hand these can be rather tediously difficult to make so as to have enough blanked-out cells to be interesting, yet not so many that the resulting "partial candidate grid" no longer has a unique solution. (If it does have a unique solution, it's reasonable to call the partial candidate grid a "partial sudoku", since its solution is the same as that of the original sudoku.) As the solving of a given original sudoku proceeds step-by-step, the corresponding partial candidate grids will eventually cease having a unique solution -- for some sudokus, this occurs at the very start (so there is no "partial sudoku" possible at all), and for many others it occurs before enough cells get solved to provide enough blanks for an interesting puzzle.

On the other hand, it should be fairly easy to produce this type of puzzle as part of a standard sudoku solver program ... At any stage of the step-by-step solving, the program could check whether the partial candidate grid (with all single-candidate cells blanked) has a unique solution, and if so to offer the "partial sudoku" as an optional puzzle.