2048 Python with AI

This project is a Python implementation of the popular game 2048 by Gabriele Cirulli. The objective of the game is to slide numbered tiles on a grid to combine them to create a tile with the number 2048. This version uses TKinter for the graphical interface and includes several AI strategies to play the game.

Getting Started

To start the game, run:

$ python3 puzzle.py

AI Implementations

This project includes several AI strategies to play the game:

1. Greedy AI: Selects the move that maximizes a heuristic value.
2. Expectimax AI: Uses the expectimax algorithm to choose the best move.
3. Monte Carlo Tree Search (MCTS) AI: Uses MCTS to simulate and select the best move.

Running the AIs

To run the Greedy AI:

$ python3 greedy_ai.py

To run the Expectimax AI:

$ python3 expectimax_ai.py

To run the MCTS AI:

$ python3 mcts_ai.py

Heuristics

Several heuristics are implemented to evaluate the board state:

• Score Heuristic: Sum of all tile values.
• Open Cells Heuristic: Number of empty cells.
• Max Tile Heuristic: Value of the highest tile.
• Tile Sum Heuristic: Weighted sum of all tile values.

Results and Analysis

After running the AI implementations, various graphs have been generated to analyze their performance.

Score Comparisons

Key Tile Achievement Rates

Average Time per Move

Average Time per Game