The New Sudoku Players' Forum

[civiliza]

I had raised this issue earlier and at the time the discrepancies between my Generator and Sudoku Explainer were due to bugs in my Generator.

However, I have just found a non-bug situation where removing givens can actually lower the rating of a puzzle:

On the way to generating a final puzzle, my generator (which works by removing givens from a full answer) created an interim state:

Code: Select all

7..|5.2|498
.45|798|321
982|314|...
-----------
3.6|.85|...
5.1|.2.|..7
4.8|.7.|...
-----------
8..|.5.|..4
15.|.3.|..9
237|849|165


Which needs a BUG+1 (SE rating 5.6) to solve.

The generator continued to remove givens and ended up with the puzzle:

Code: Select all

7..|5.2|.9.
.4.|.98|32.
9..|.1.|...
-----------
..6|.85|...
5..|.2.|..7
4.8|...|...
-----------
...|.5.|..4
1..|...|...
.3.|..9|1.5


Which only needs a Type 1 Unique Rectangle (SE rating 4.5) to solve.

So there is at least one situation where "taking away squares" (for which read givens) actually makes a puzzle "easier".

[dobrichev]

see http://forum.enjoysudoku.com/post26244.html#p26244

[civiliza]

Thanks for finding / identifying that thread. It seens that Uniqueness methods are largely to blame for this.

That being the case it makes sense that the forum that first taught me uniqueness techniques would have already discovered the phenonemon.

Civvy-come-lately strikes again.