About Atlas
The engine behind the board.
The Engine
Atlas is a high-performance, UCI-compliant chess engine written entirely in Go from scratch. It utilizes advanced bitboard architectures, including pre-calculated magic bitboards for sliding pieces, to achieve extreme move generation speeds.
The engine is powered by an Iterative Deepening Negamax search framework utilizing Alpha-Beta Pruning, Quiescence Search, Null-Move Pruning, and a massive Transposition Table.
Self-Learning
Atlas isn't just a static bot. It features a fully integrated self-play loop capable of producing hundreds of thousands of games. Using Texel's Tuning Method, the bot subjects its game history dataset to Logistic Regression gradient descent algorithms, iteratively adjusting its internal piece valuations to minimize its Mean Squared Error (MSE) and become mathematically stronger.
The Tech Stack
- Backend Engine: Go (Golang)
- Frontend Web UI: Next.js, React, TypeScript
- Styling: Pure Vanilla CSS (Glassmorphism, animated orbs)
- Board Engine: react-chessboard & chess.js
Created by Multix
Atlas was architected and built by Multix to push the boundaries of backend engine speed paired with sleek, modern UI aesthetics.