2048 Game Bot Project

$ Source Code

toy_archive/2048_ai at main · hekoh99/toy_archive

$ Overview

The 2048 Game Bot project involves developing a bot that plays the 2048 game using four distinct heuristics to optimize its gameplay. The bot calculates the best possible move within a time limit of 1/12 of a second.

Heuristics Used

1. Big:
   • Evaluates boards where the sum of all numbers is higher, giving a higher score to such boards.
2. Emptyness:
   • Favors boards with more empty spaces, as they offer more potential for future moves.
3. Monotonicity:
   • Scores boards higher where larger numbers are positioned towards the corners and arranged in descending order.
4. Smoothness:
   • Boards with smaller differences between adjacent tiles are given a higher score, indicating a smoother tile distribution.

Algorithm and Time Constraint

• The bot operates with a time constraint of 1/12 of a second to evaluate and decide the optimal move for the maximum possible score during that time frame.
• Utilized Minimax and Expectimax algorithms to make optimal decisions based on the heuristics.

Key Learnings

• Virtual Environments: Created and managed Python virtual environments, including handling package installations and version conflicts.