Rock-Paper-Scissors AI

An intelligent Rock-Paper-Scissors game that learns from your playing patterns using a neural network. The AI analyzes your previous moves and game outcomes to predict your next move and counter it strategically.

Features

• Interactive Web Interface: Clean, responsive web UI with emoji-based game buttons
• Neural Network Learning: PyTorch-based neural network that learns from gameplay patterns
• Training Mode: Collect training data through interactive gameplay
• AI Prediction: Trained model predicts player moves based on game history
• Dual Play Modes: Random play (untrained) and AI play (trained)
• Real-time Feedback: Live game results and model status updates

How It Works

The AI uses a neural network to learn patterns in your gameplay:

1. Input Features: Previous player move, previous program move, and previous game result (9 features total via one-hot encoding)
2. Neural Architecture:
   • Input layer: 9 neurons (3 for each categorical feature)
   • Hidden layer: 32 neurons with ReLU activation
   • Output layer: 3 neurons (probabilities for Rock, Paper, Scissors)
3. Training: Uses collected gameplay data to train the network with CrossEntropyLoss and Adam optimizer
4. Prediction: Predicts your next move and plays the counter-move to maximize win probability

Installation

Prerequisites

• Python 3.7+
• pip package manager

Setup

1. Clone or download the project files
2. Install required dependencies:

pip install fastapi uvicorn torch

Required Files

Ensure you have all these files in your project directory:

• app.py - FastAPI web server
• main.py - Core game logic and neural network
• index.html - Web interface
• styles.css - Styling
• script.js - Frontend JavaScript logic

Usage

Web Interface (Recommended)

1. Start the web server:

uvicorn app:app --reload

2. Open your browser and navigate to http://localhost:8000

3. Training a Model:

   • Enter the number of training samples (recommended: 50-100)
   • Click "Train New Model"
   • Play the specified number of games by clicking Rock, Paper, or Scissors
   • The AI will collect data and train automatically

4. Playing Games:

   • After training, click Rock, Paper, or Scissors to play
   • The AI will use your gameplay patterns to predict and counter your moves
   • Click "Reset Model" to clear training data and start over

Command Line Interface

Run the standalone version:

python main.py

Follow the prompts to:

• Train a new model with sample data
• Play games against the AI
• Quit the program

Game Rules

• Rock (🪨) beats Scissors (✂️)
• Paper (📄) beats Rock (🪨)
• Scissors (✂️) beats Paper (📄)
• Same moves result in a tie

Technical Details

Neural Network Architecture

Input Layer (9 neurons)
    ↓
Hidden Layer (32 neurons, ReLU)
    ↓
Output Layer (3 neurons, Raw logits)

Training Parameters

• Epochs: 500 (default)
• Learning Rate: 0.005
• Loss Function: CrossEntropyLoss
• Optimizer: Adam
• Minimum Training Samples: 1 (recommended: 50+)

API Endpoints

• GET / - Web interface
• GET /play - Get random move
• GET /evaluate - Evaluate game result
• POST /train_model - Train the neural network
• GET /predict_move - Get AI prediction
• POST /add_datum - Add training data
• POST /reset - Reset model

File Structure

project/
├── app.py              # FastAPI web server
├── main.py             # Core game logic and neural network
├── index.html          # Web interface
├── styles.css          # Styling
├── script.js           # Frontend JavaScript
└── README.md           # This file

Performance Tips

• Training Data: More samples generally improve AI performance (50-100 recommended)
• Consistent Patterns: The AI learns better when you have detectable patterns
• Game History: AI needs at least one previous game to make predictions

Troubleshooting

Common Issues

1. Model not making smart moves: Increase training samples or ensure you have consistent patterns
2. Web interface not loading: Check that all files are in the same directory
3. Training fails: Ensure you have at least 2 samples to create training pairs

Dependencies Issues

If you encounter import errors:

pip install --upgrade torch fastapi uvicorn

Contributing

Feel free to enhance the project by:

• Improving the neural network architecture
• Adding more sophisticated features
• Enhancing the web interface
• Adding game statistics and analytics

License

This project is open source and available under the MIT License.

Future Enhancements

• Add game statistics and win/loss tracking
• Implement different AI difficulty levels
• Add multiplayer support
• Include more sophisticated pattern recognition
• Add export/import functionality for trained models