LikeLike

]]>LikeLike

]]>https://github.com/EdLogg/MineSweeper

LikeLike

]]>LikeLike

]]>LikeLike

]]>I have updated my code to use a 64 bit pseudo random number generator (KISS) and sped up the program. I can give you a more updated numbers but for the expert game using 60,000 sample puzzles using 600 random seeds:

Pos Wins Percent est. Dev.

——————————————

3,3 28297 47.16% .21%

3,2 28172 46.95% .20%

2,3 28218 47.03% .20%

4,3 28030 46.72% .20%

3,4 28011 46.69% .20%

So there is really three best starting places in the upper left quadrant. So my original issue of why 3,2 was better than 2,3 was really a case of small test samples.

LikeLike

]]>I was thinking of writing a solver myself – and started down the same path, of deductive thinking. I didn’t get to code, but tried what I thought out manually on a few expert boards. And then I found this: https://massaioli.wordpress.com/2013/01/12/solving-minesweeper-with-matricies/ IMO, it sort of models deductive thinking in a better way than using rules as a human would like them. In each step, you just have to write equations for all already uncovered squares. Fortunately, solving sytems of at best thousands of linear equations is an extremely light task for modern computers. You can, of course, optimize by always running a simple, straightforward step of marking all squares which _have to_ be mines and which are certainly without mines after each solving of the linear system.

LikeLike

]]>LikeLike

]]>LikeLike

]]>