Somewhat OT... Computer playing Minesweeper

[Paul Rubin]

Heiko Wundram <heikowu at ceosg.de> writes:
> That's why I'm running a Monte-Carlo simulation on the whole thing, using a 
> general algorithm for solving a given mine-field, and testing for "many" 
> distinct fields. But, the thing is: KMines tells me that on simple setting 
> the solvability rate is somewhat near 80% for 200 runs, while my algorithm 
> gives a solvability rate of 55-65% for 200 runs.

I don't know how either your program or KMines works, but with the
Windows program I used to use, you had to clear a few squares by hand
before the algorithm could infer anything.  Sometimes you blew
yourself up while clearing those first few squares.  There's not
enough information to avoid that.  I don't know if KMines counts those
explosions against it's algorithm's success rate.