Online Maze Puzzle

What makes a perfect maze different?

A maze can contain loops, disconnected rooms or multiple equally valid routes. A perfect maze is stricter. Every cell belongs to one connected network and the passage graph contains no cycle. In graph terms it is a spanning tree. That gives it a useful property for a daily brain game: from the start to the goal there is exactly one simple route, so the puzzle never depends on hidden tie-breaking or an arbitrary “preferred” answer.

V0.129 does not infer that property from a screenshot or from whether the browser can reach the exit. The project-local generator freezes sixty source mazes, and a separate verifier independently checks reciprocal passages, full connectivity and the exact edge count of cells minus one. It also recomputes the start-to-goal path and compares it with the stored route.

The result is a maze that remains simple to understand but has a stronger release contract than “it seems solvable.”

Three sizes, one consistent control model

Easy uses an 11×11 logical cell grid and is meant for quick orientation practice. Standard expands to 17×17 and introduces more junctions and deeper wrong branches. Challenge uses 25×25, creating a substantially longer route while keeping the board usable on a narrow phone screen.

You can move with arrow keys or WASD. On touch devices, tap an open neighboring cell, swipe a direction, or drag through a corridor. The same board also supports mouse interaction. The controls never change the maze itself, so a seeded Practice URL reconstructs the same source, transform, start and goal regardless of input method.

Breadcrumbs show the route you are currently following. Fog mode visually reduces distant cells without changing walls or scoring. Both are local display preferences rather than separate puzzle variants.

Measure detours instead of inventing a mystery score

The live monitor exposes the verified shortest-path length before you finish. Your own movement is counted separately: total steps, backtracks, dead-end entries, visited cells and unique junctions reached. When you solve, route efficiency is simply shortest-path steps divided by your actual steps.

That makes the result easy to interpret. A 100% route means you followed the unique solution without any detour. A lower percentage means you explored extra branches. There is no opaque rating formula that changes because of device speed, animation timing or an undisclosed server model.

Backtracking is not treated as failure. It is normal maze-solving behavior and is recorded as descriptive telemetry. The goal is to help you compare attempts and notice whether you are learning to reject unproductive branches earlier.

How to solve a maze more deliberately

A common method is wall following: choose one side and keep that hand connected to the same wall. In a simply connected perfect maze it will eventually lead you through the network, although it may take many detours. It is reliable but rarely shortest.

For a cleaner solve, pause at junctions and build a small mental map. Notice whether a branch immediately narrows into a dead end, whether it moves you away from the broad direction of the goal, and whether you have already explored a neighboring corridor. Breadcrumbs can reduce memory load while you practice this scanning habit.

On larger boards, do not try to memorize the entire maze at once. Work in local chunks: current corridor, next junction, branch choice. This keeps the visual task manageable and is more useful than repeatedly zooming the whole board.

Verified Route Hint is intentionally limited

If you get stuck, Verified Route Hint highlights only the next adjacent cell on the unique path from your current position to the goal. It does not teleport the player, fill the remaining route or reveal an entire solution overlay.

The hint comes from the same independently verified tree path used in release QA. Requesting it marks the run Assisted. That separation lets DailyBrainArc preserve honest Clean bests while still giving players a way to learn rather than abandon a difficult maze.

Because a perfect maze has one route between every pair of cells, the hint remains well-defined even after you have explored a wrong branch. From any current cell there is still exactly one path to the goal.

Deterministic Practice and browser-local autosave

Opening Practice without a seed stabilizes the URL with an s= value. Reopening that URL selects the same frozen source and symmetry transform. This makes a good board shareable and testable without a remote puzzle API.

Movement state is autosaved locally: current position, active route, visited cells, step count, backtracks, dead-end history, elapsed time, display options and Assisted status. Close the tab and reopen the same seeded puzzle to continue. Clearing browser storage can remove those records, so the product does not pretend local progress is cloud-backed.

Daily Maze uses the same runtime but a date-stable selector and separate specialist history. It remains outside the frozen Today’s 6 rotation.

Print the same verified board

The playable page includes a print-friendly layout. Navigation, controls and live telemetry disappear in print while the maze walls, start and goal remain. This gives the same verified board a worksheet use case without creating a family of thin “printable 11×11,” “easy maze,” or “maze for mobile” doorway pages.

Search expansion stays focused on distinct intent. V0.129 adds this general Maze guide and a separate Daily Maze surface because they answer different user needs: learning and unlimited Practice versus one date-stable returning challenge.

Why the release verifier matters

Generated content can fail in subtle ways: a one-way wall, an isolated cell, an accidental loop, a stored route that no longer matches the board after a transform, or two source entries that become identical. V0.129 checks those properties independently of the UI.

The verifier rechecks all sixty sources and all eight symmetry transforms for each source. It validates thousands of deterministic Practice selections and every 2026 Daily selection. It also measures distinct transformed starts instead of assuming that a Daily selector produces 365 different boards.

This keeps runtime lightweight while making the release claim specific: bundled mazes are frozen, connected perfect mazes with a mathematically unique route and deterministic replay.