Classic minesweeper game, redefined by AI.
Our minesweeper game is one of its own kind in the market. We leverage AI technology to achieve various improvement on top of the classic minesweeper game.
- Exercise your brain and logic reasoning skills
- 4 difficulty options
- Fun experiences to unlock and daily rewards to claim.
- Classic minesweeper redefined by AI
* Redefined by AI *
1. Model the minesweeper game as a Constraint Satisfaction Problem (CSP) [a]
2. Precompute the heuristic score for each valid move using multiple state-of-the-art algorithms [b], to yield a stack rank of 10 (or less) best moves.
3. Based on the heuristic difference between the optimal move and the actual move by the user, we use our pre-defined probability (non-uniform distribution) to control the mine position (by ad-hoc switching cells on the board).
4. At the end of the game, further tune the mine-switching probability using win/lose ratio as an input [c], to ensure a custom tailored gaming experience that’s challenging but not frustrating.
References:
[a] https://inst.eecs.berkeley.edu/~cs188/fa18/assets/slides/lec4/FA18_cs188_lecture4_CSPs_6pp.pdf
[b] https://dash.harvard.edu/bitstream/handle/1/14398552/BECERRA-SENIORTHESIS-2015.pdf?sequence=1
[c] https://en.wikipedia.org/wiki/Reinforcement_learning
Our minesweeper game is one of its own kind in the market. We leverage AI technology to achieve various improvement on top of the classic minesweeper game.
- Exercise your brain and logic reasoning skills
- 4 difficulty options
- Fun experiences to unlock and daily rewards to claim.
- Classic minesweeper redefined by AI
* Redefined by AI *
1. Model the minesweeper game as a Constraint Satisfaction Problem (CSP) [a]
2. Precompute the heuristic score for each valid move using multiple state-of-the-art algorithms [b], to yield a stack rank of 10 (or less) best moves.
3. Based on the heuristic difference between the optimal move and the actual move by the user, we use our pre-defined probability (non-uniform distribution) to control the mine position (by ad-hoc switching cells on the board).
4. At the end of the game, further tune the mine-switching probability using win/lose ratio as an input [c], to ensure a custom tailored gaming experience that’s challenging but not frustrating.
References:
[a] https://inst.eecs.berkeley.edu/~cs188/fa18/assets/slides/lec4/FA18_cs188_lecture4_CSPs_6pp.pdf
[b] https://dash.harvard.edu/bitstream/handle/1/14398552/BECERRA-SENIORTHESIS-2015.pdf?sequence=1
[c] https://en.wikipedia.org/wiki/Reinforcement_learning
Show More
