Little Sorry

A Rust library for exploring regret minimization algorithms, with a focus on game theory applications.

Features

• Regret matching implementation
• Rock Paper Scissors (RPS) example game
• Highly performant using ndarray for numerical operations
• Thread-safe with no unsafe code (except for carefully bounded enum conversions)

Getting Started

Add this to your Cargo.toml:

[dependencies]
little-sorry = "1.0.0"

How It Works

The library implements regret minimization algorithms, which are used in game theory to find optimal strategies in imperfect-information games. The core algorithm tracks:

1. Action probabilities for each possible move
2. Cumulative regret for not taking alternative actions
3. Strategy updates based on regret matching

The RPS example demonstrates these concepts in a simple zero-sum game setting.

Building and Testing

# Run all tests
cargo test

# Run benchmarks
cargo bench

# Build in release mode
cargo build --release

License

Licensed under the Apache License, Version 2.0.