This is an opensource AI project to solve Minesweeper as fast (slow) as possible

This is an opensource project I made when I had some free time. The goals were to:

• Have a basic algorithm to solve Minesweeper
• Use screen capture and mouse-clicks
• I used https://minesweeperonline.com
• I worked on it for ~3 days

• Iterative process for:
  • calculate probabilities of bombs
  • flag squares with probability 1
  • expose certainly non-bomb squares
  • if no 100% certainty - mark squares with the highest bomb probability

• main.py has 2 run options:
  • main(board_level=) - run the program on the given config
  • read_board(board_level=BeginnerMinesweeperOnline) - read and print the board on the screen

• Reading screen pixels isn't ideal, maybe I have offsets here and there and in expert mode the board doesn't always match what's on the screen.
• Improve middle-clicks - only when necessary
• Completely ignore blank screen - no need to sample them over and over
• REFACTORING! (make main more OO, migrate logic to board.py,... )
• TESTS!
• Maybe get pixels from browsers? from DOM? would be more accurate

• [email protected]
• https://www.linkedin.com/in/sagydr/