Personal Project
Pattern Forge
Open sourcePattern Forge is a chess training application inspired by the Woodpecker Method: you repeat tactical puzzles in deliberate cycles to build pattern recognition, speed, and recall. Training sets, timed sessions, and mistake review keep the loop tight—move, feedback, next—with progress stored locally in the browser.
The app is actively used for personal training and continues to evolve around the practical workflow of deliberate chess repetition.
Why I built it
Pattern Forge started after reading The Woodpecker Method by Axel Smith and Hans Tikkanen. I liked the training idea, but doing it manually was slow: setting up positions on a physical board, tracking cycles, recording mistakes, and reviewing progress took too much time away from the actual training.
Together with João, I started building Pattern Forge to automate that workflow: reusable training sets, repeatable cycles, session tracking, mistake review, and analytics to understand whether recognition and speed are improving over time.
In the app
Training view showing a single puzzle inside a cycle. The goal is to reduce setup friction so the focus stays on solving, repetition, and pattern recognition.
What it does
Training sets
Curated puzzle groups are the unit of practice. You choose a set, then work it in full cycles rather than one-off random positions.
Cycles
Each full pass through a set is a cycle. Comparing cycles surfaces whether recognition and speed are actually improving.
Sessions
Work is organised into sessions so you can train in focused blocks, pause, and resume without losing where you left off.
Mistake review
Failed or slow puzzles feed a dedicated review path so you close gaps before the next cycle, not after you have already moved on.
Progress and reflection
After completed cycles, you can see how performance shifted over time—not only a single session, but the arc across repeats.
Local-first storage
Training history and puzzle state live in IndexedDB via Dexie. Everything runs in the browser; your data stays on your machine.
Architecture
- Frontend foundation
Next.js, React, TypeScript, Tailwind
- Chess engine & board
chess.js, chessground
- Local-first persistence
Dexie, IndexedDB
- Analytics & progress
Recharts
- Testing
Vitest, Testing Library, Playwright
How the training works
- Select a training set to anchor the cycle.
- Run a cycle: solve each puzzle in the set with immediate move feedback.
- Review mistakes and weak spots before or between cycles.
- Repeat the cycle on the same set and compare passes over time.
- Use progress after completed cycles to see recognition and speed improve.
Pattern Forge is built around a simple idea: tactical improvement comes less from endless novelty and more from structured repetition, reflection, and returning to the same patterns until recognition becomes instinctive.