Skip to content

Quickstart

This walkthrough covers the complete pawnbook loop from first start to first drill session.


Step 1 — Start the server

bash
docker compose up

Or for a native install:

bash
npm start

Open http://localhost:3000. The dashboard appears.


Step 2 — The dashboard

On a fresh install the dashboard shows:

  • Elo tile: 1200 — the starting rating before any ranked games
  • Puzzles due tile: 0 — no games played yet, no drill cards
  • Streak tile: 0 days
  • Suggested opponent: appears once the engine availability check completes (a few seconds after startup)
  • Recent games: empty table

The suggested opponent is chosen based on your current Elo. On first launch it defaults to Maia-1500.


Step 3 — Pick an opponent and start a game

Click Play in the navigation. The play page shows the board and an opponent selector.

Choose an opponent from the grid:

  • Maia-1500 is a reasonable starting point for most casual players. It plays like a 1500-rated human — it will blunder, miss tactics, and play somewhat inaccurately, producing a realistic game.
  • Stockfish at calibrated depth plays more precisely at the same rating but with different error patterns. Use it if you want cleaner, more consistent opposition.

Leave the defaults as-is for your first game: White, Ranked, no time control, Coach on.

Coach toggle

The coach is silent until you have 20 confirmed opening book nodes. For a first game it will not intervene even if coach is enabled.


Step 4 — Play a game

Click the pieces to move. The engine responds after a short pause.

After the game ends:

  1. A result card appears on the board — win, loss, or draw. Elo shows while analysis runs.
  2. An analysis progress bar fills in three segments. Pass 1 (the largest segment) evaluates every position; pass 2 re-examines mistakes; pass 3 runs Maia policy probes.
  3. When analysis finishes, game_over is re-sent with the final Elo delta. The Elo tile on the board updates.

The full analysis of a 40-move game takes under 5 minutes on modern hardware if no pre-evaluation was cached. Subsequent games are faster because positions are pre-evaluated during play.


Step 5 — Review and drill

Two buttons appear after the game:

Review — opens the annotated game. You will see:

  • A win-percentage chart across the game
  • Each move colour-coded by classification (blunder, mistake, inaccuracy, good, great, best)
  • A mistake list with findability scores and best-move alternatives

Quiz — opens immediate practice on the mistakes from this game. The quiz does not schedule FSRS cards; it creates cards that enter the normal drill queue from tomorrow.

To drill your queue: click Drill in the navigation. Due cards appear one at a time. Find the best move in each position. The server infers a rating from your response time and correctness — you do not rate yourself.


That's the loop

Every game feeds the analyser. Every mistake feeds the queue. The queue gets shorter as you drill.

The dashboard Elo and strength-estimate tiles update after each analysed game. The due-count tile tracks how many cards are waiting. The streak tile marks days with any activity.

Continue from here:

Released under the MIT License.