Overview
Chessiac reads every game you have ever played on Chess.com, classifies each one by ECO code, and draws the result as an illuminated skill tree — so you can see at a glance which openings you own and which ones have been quietly losing you games.
The problem it addresses is that a rating is one number standing in for a thousand different positions. It cannot tell you that your Ruy Lopez is carrying you, that the Sicilian has been bleeding you for months, or that you have played the French exactly four times and have no idea whether it suits you.
No account, no upload. It reads the public Chess.com API and parses everything in the browser.
- 500+ECO codes mapped
- 2Mirrored trees
- 6Time epochs
- 0Games uploaded
01 — How it works
Three steps, no account
-
Name thyself
Type a Chess.com username. It's remembered locally, so the next visit skips straight to the tree.
-
The archives are read
Monthly archives are fetched from the public API with live progress, then parsed in the browser. Nothing is sent anywhere.
-
The tree grows
White branches left, Black branches right, from a shared centre. Click any node to open its variations.
Design decisions
Verdicts, not just numbers. Wins, draws and losses aggregate up the tree, so a branch that looks healthy at the family level often has one variation dragging it down. Openings above 65% are marked Strong, below 35% Weak — with a five-game floor so noise never earns a verdict.
Aged parchment, gold leaf, iron gall ink. The chronicle metaphor isn't decoration — it's what makes a dense statistical tree feel worth reading. Cinzel for display, Crimson Text for body, on an 8-point rhythm.
Five states, always. Every surface is specified for default, hover, focus, loading and empty. Motion has an off switch and honours reduced-motion preferences.
02 — Presentation
The showcase
A scrolling walkthrough of the product, in its own parchment-and-gold visual language — the problem, the tree, every interface surface, and the design system underneath. Includes live interactive components.
Outcome
What was produced. A working prototype and a full showcase page — the ECO classifier, the mirrored tree, per-node win-rate aggregation, filters and verdicts, plus a documented design system of palette, type and interaction states.
What I learned. The classifier was the real design problem, not the visuals. Deciding where a family ends and a variation begins determines what the tree can tell you, and no amount of illumination rescues the wrong hierarchy.
What could be improved. Deep repertoires get wide fast. The tree needs better density handling — collapsing, searching and pruning — before a player with ten thousand games can read it comfortably on a phone.
Where it goes next. Opponent-side analysis: the same tree built from someone else's archive, so you can prepare for a specific player rather than only auditing yourself.
Links
- Product showcase Full walkthrough
- Live app Prototype, not yet deployed
- Source code Private