About
Lizzie is a free, open-source desktop application that bridges the powerful Leela Zero Go-playing AI engine with an accessible, real-time graphical board interface. Designed for Go players who want to study and improve their game, Lizzie renders Leela Zero's analysis directly on the board, showing top candidate moves, win-rate estimates, and playouts for each variation as you navigate a game. Built in Java (requiring version 8 or higher), Lizzie runs on all major desktop platforms—Windows, macOS, and Linux—and supports OpenCL-accelerated computation for faster analysis. Users can load their own OpenCL tuning files to speed up the initial setup process. The tool supports building from source for developers who wish to contribute or customize their experience. Lizzie is ideal for serious Go enthusiasts, amateur and professional players, and researchers studying AI decision-making in board games. The project is community-driven, hosted on GitHub with over 1,000 stars and an active wiki for documentation and setup guidance. Its GPL-3.0 license ensures it remains freely usable and modifiable by anyone. Key use cases include post-game review, opening preparation, endgame analysis, and AI vs. human strategy comparison. Because it wraps Leela Zero—itself inspired by DeepMind's AlphaGo methodology—Lizzie gives players access to superhuman-level analysis without any subscription fee.
Key Features
- Real-Time AI Analysis: Displays Leela Zero's move suggestions, win-rate estimates, and playouts directly on the Go board as you review a game.
- Graphical Board Interface: Provides an intuitive visual board where candidate moves and analysis are overlaid, making AI insights immediately readable.
- OpenCL Acceleration: Leverages GPU-based OpenCL tuning for faster neural network inference, with support for user-supplied tuning files.
- Cross-Platform Java App: Built on Java 8+, Lizzie runs natively on Windows, macOS, and Linux with no platform-specific builds required.
- Open Source & Community-Driven: Licensed under GPL-3.0 with full source code on GitHub, an active wiki, and community contributions via issues and pull requests.
Use Cases
- Post-game review for Go players wanting to understand where they made mistakes and what moves the AI preferred.
- Opening preparation by analyzing joseki and fuseki sequences with superhuman-level AI feedback.
- Endgame analysis to identify optimal moves in close territorial situations using win-rate evaluation.
- AI strategy research for academics or enthusiasts studying how neural network-based engines like Leela Zero approach board games.
- Coaching and teaching aid for Go instructors who want to show students AI-recommended moves during lessons.
Pros
- Completely Free: Open-source under GPL-3.0 with no cost, subscriptions, or usage limits—full AI-powered Go analysis at no charge.
- Superhuman Analysis Quality: Powered by Leela Zero, which uses a neural network approach inspired by AlphaGo, delivering top-tier move evaluations.
- Cross-Platform Compatibility: Java-based runtime ensures Lizzie works on all major desktop operating systems without special configuration.
Cons
- Steep Initial Setup: First-run OpenCL tuning can take a long time, and setting up Leela Zero alongside Lizzie requires technical familiarity.
- Requires Java 8+: Users must have a compatible Java runtime installed, adding an extra dependency compared to standalone executables.
- Limited Active Maintenance: The repository has accumulated 193 open issues and 82 pull requests, suggesting development activity has slowed.
Frequently Asked Questions
What is Lizzie?
Lizzie is an open-source graphical interface that lets you analyze Go games in real time using the Leela Zero AI engine. It displays AI move suggestions and win-rate estimates directly on the board.
Does Lizzie require a paid subscription?
No. Lizzie is completely free and open-source under the GPL-3.0 license. You can download, use, and modify it at no cost.
What do I need to run Lizzie?
You need Java 8 or higher installed on your computer, as well as Leela Zero. Detailed setup instructions are available in the project's GitHub wiki.
Why does the first run take so long?
On first launch, Leela Zero runs OpenCL tuning to optimize neural network performance for your specific GPU. This is a one-time process; you can speed it up by supplying existing tuning files.
Can I contribute to Lizzie?
Yes. The project is hosted on GitHub and welcomes contributions. You can build from source, submit pull requests, or report issues following the guidelines in CONTRIBUTING.md.
