Chessly
A production-grade, client-side replica of Chess.com's Game Review & Analysis Mode featuring client-side engine evaluation, Danny Rensch commentary, and Polyglot opening classification.
Year
2026
Status
Open Source (MIT)
Role
Full-Stack Architecture & Frontend Engineering
The Challenge
Chess.com's Game Review is one of the most beloved coaching tools in online chess, but it runs on heavy proprietary cloud infrastructure. Most open-source alternatives are either basic PGN viewers or clunky engine wrappers that lack the conversational warmth and tactical clarity of modern game review.
The engineering objective for Chessly was to recreate the full Chess.com review experience entirely in the browser—with zero server dependencies, zero latency, and zero tracking.
- Executing deep position analysis and move classification on client hardware without freezing the main thread
- Parsing and searching a massive 235,000+ line Polyglot opening book in real time with minimal memory overhead
- Generating conversational, human-like coaching advice for every move classification without requiring paid API tokens
Architecture & Implementation
To guarantee a silky 60 FPS UI experience during intensive engine calculations, the evaluation pipeline was isolated into a background Web Worker. The evaluation combines PeSTO (Piece-Square Tables) middle-game and end-game values with an alpha-beta minimax search.
Opening detection is powered by standard Polyglot 64-bit Zobrist hashing. The entire 3.7MB binary book is fetched as an ArrayBuffer and searched via binary search in O(log N) time—requiring at most 18 comparisons to identify the exact opening variation.
Move accuracy is computed using the CAPS2 winning probability formula, mapping centipawn evaluations through a sigmoid curve to classify moves into 10 discrete tiers (Brilliant, Great, Best, Excellent, Good, Book, Inaccuracy, Mistake, Miss, and Blunder).
Key Results
Chessly delivers an end-to-end game review experience with authentic Neo piece sets, audio pooling for realistic sound effects, interactive board themes, and an interactive Danny Rensch commentary coach.
The application runs entirely client-side on the web and is accessible via the live deployment.
- Sub-18 comparison binary search over 235,000+ Polyglot opening positions
- Web Worker multi-threading preventing any UI micro-stutters during 40+ move game reviews
- 100% serverless static deployment runnable anywhere offline or on the edge