2048 AI Game

A Java implementation of the 2048 game with an intelligent AI player using the Expectimax algorithm.

Features

• Complete 2048 Game: Fully functional implementation of the classic 2048 puzzle game
• Intelligent AI Player: Uses Expectimax algorithm with sophisticated heuristics
• Graphical Interface: Clean GUI built with Java Swing
• Victory Detection: Game stops automatically when reaching the 2048 tile
• Human vs AI: Choose between human player or AI player

AI Strategy

The AI player uses an Expectimax algorithm with multiple evaluation heuristics:

• Highest Tile Score: Prioritizes growing the largest tile
• Empty Space Management: Values keeping multiple empty spaces
• Merge Potential: Looks for opportunities to merge identical tiles
• Tile Smoothness: Prefers smooth transitions between adjacent tiles
• Monotonicity: Encourages ordered tile arrangements
• Corner Strategy: Rewards keeping the highest tile in corners
• Future Move Potential: Values having multiple possible moves

How to Build and Run

Compile:

javac -cp src src/ttfe/*.java

Run with AI Player:

java -cp src ttfe.TTFE --player c

Run with Human Player:

java -cp src ttfe.TTFE --player h

Custom Options:

java -cp src ttfe.TTFE --seed 42 --width 4 --height 4 --player c

Game Controls (Human Player)

• Arrow Keys: Move tiles (↑ ↓ ← →)
• Goal: Combine tiles to reach 2048

AI Performance

The AI typically achieves:

• Consistent performance reaching 1024+ tiles
• Good chance of reaching the 2048 goal
• Strategic tile placement using corner strategy
• Efficient empty space management

Technical Details

• Language: Java
• Algorithm: Expectimax with depth-limited search
• Search Depth: Configurable (default: 2)
• Heuristics: Multi-criteria evaluation function
• Architecture: Clean separation of game logic, AI, and UI

Credits

This project implements the classic 2048 game concept with custom AI algorithms. The Expectimax algorithm and evaluation heuristics are original implementations.