Maze is a free browser puzzle where you steer a glowing token from the top-left corner to the door in the bottom-right, counting every step you take. Level 1 is an 8×8 grid; each level you clear rebuilds the maze one size bigger, up to 16×16, and every maze is generated fresh so you can never memorise a route. It runs in the tab with no download and no signup, which makes it equally workable as a two-minute break for an adult and a first logic puzzle for a child who can already handle arrow keys.
How to Solve Any Maze: The Wall-Follower (Left-Hand Rule)
The wall-follower is the one maze technique worth memorising, and it is a single instruction: put an imaginary hand on one wall and never take it off. Playing left-hand, that means at every cell you try to turn left first; if there is a wall to your left, you go straight; if straight is blocked too, you turn right; if all three are blocked you are in a dead end, so you turn around. You do not think, you do not look ahead, you just cycle left → straight → right → back at every single cell. The right-hand rule is the same procedure mirrored, and either one gets you to the door here — the only difference is how much of the maze you tour on the way.
The reason it works in this game is not luck. Every maze here is what mathematicians call simply connected: no passage ever loops back on itself, so all the walls are one single connected surface. Trace that surface with your hand and you eventually run your fingers over every part of it, which means you walk past every cell in the grid, which means you cannot miss the exit. The rule fails only in mazes with loops, where the goal can sit on a detached island of wall your hand never touches — and this game contains no loops at all.
The catch is cost. Wall-following can walk you into every dead-end branch in the grid, so on a 16×16 board it will happily spend a few hundred steps on a route the shortest path covers in far fewer. Treat it as your recovery move: use it the moment you feel lost, ride it until you recognise where you are, then go back to reading the maze with your eyes. It is a guarantee you will finish, not a method for finishing well.
Dead-End Filling and Trémaux: Finding the Shortest Route, Not Just a Route
Dead-end filling is the fastest technique for a maze you can see all of at once, which is exactly the situation here — the whole grid sits on screen. Instead of tracing the route, delete the wrong ones. Scan for any cell with only one opening: that is a dead end, so mentally fill it in. Then walk back up its corridor filling every cell until you reach a junction. Repeat until nothing is left to fill. Because these mazes have no loops, whatever survives is the solution path, and it is the only path — you have not found a good route, you have found the route.
Trémaux's algorithm is the version for when you cannot see ahead, and it is what you should fall back on if you shrink the board or play on a phone where the corridors get small. Charles Pierre Trémaux, a 19th-century French engineer whose method was written up by Édouard Lucas in 1882, formalised it as a marking system: mark every passage as you enter and leave it, never enter a passage that already carries two marks, and when you arrive at a junction you have seen before, turn straight back the way you came if your entry passage only has one mark. Otherwise take the least-marked exit available. Follow it strictly and the passages carrying exactly one mark at the end trace the solution.
On screen you obviously cannot draw on the board, so the practical version is a rule about junctions: whenever you reach a spot with three or four openings, decide which branch you are committing to before you move, and if that branch dumps you at a dead end, walk all the way back to that same junction and cross the branch off. The mistake that burns steps is abandoning a branch halfway and re-entering it from another angle later, which is precisely what the two-mark rule exists to prevent.
Controls and Cutting Your Step Count: Glide Down Corridors, Tap at Junctions
There are three ways to move. Arrow keys and WASD move you exactly one cell per press. The on-screen D-pad under the board does the same, one cell per tap. Swiping or click-dragging across the board itself is the different one: drag at least about 30 pixels in a direction and the token starts moving that way on its own, one cell every 80 milliseconds, until something stops it. It stops on a wall ahead, or the instant it reaches a junction — a cell with three or four openings. A plain corner does not stop it; the wall does, because you cannot keep going in the direction you swiped.
That gives you the technique that actually cuts your time: swipe down corridors, tap at junctions. Long straight or lightly-bending stretches are where key-tapping wastes the most effort, and a single swipe covers them at eight cells per second while you use the time to read ahead. Around branching points, switch to single presses so you land exactly where you meant to. Every glided cell still increments the Steps counter, so gliding does not make your route cheaper — it makes the same route faster to execute.
The Steps number itself is a self-imposed challenge, not a penalty; your points come from clearing levels (a cleared level is worth 20 plus twice the grid size, so an 8×8 is 36 and a 16×16 is 52). The hard floor on any grid is the Manhattan distance from corner to corner: 14 steps on the 8×8, 30 on the 16×16. No maze here lets you hit that, because the generator builds long winding corridors, but the gap between your count and that floor is the honest measure of how well you read the board. Press Finish & see results whenever you want to close the run out.
Perfect Mazes vs Braid Mazes: Why the Hand-on-Wall Trick Works Here
A perfect maze has exactly one path between any two cells: no loops, no isolated sections, no shortcuts. A braid maze is the opposite extreme — every dead end has been removed by knocking through an extra wall, so the whole thing is loops and you can circle forever without covering new ground. Most commercial and hedge mazes are partial braids, deliberately, because loops are what make a maze feel confusing to a person walking through it. Every maze in this game is perfect.
That single property is what makes the hand-on-wall rule safe here, and it is also why dead-end filling terminates on the correct answer rather than an arbitrary one. In a braid maze both techniques break: wall-following can trap you circling one loop's outer wall forever if the exit sits on a wall island your hand never reaches, and dead-end filling does nothing at all because there are no dead ends to remove. If you have ever tried the left-hand rule in a hedge maze and had it fail, a loop is why.
The practical read is that anything that looks like a loop on this board is a misread, not a design feature. If you seem to be arriving somewhere you have already been by a different route, you did not — you backtracked without noticing, most likely by swiping past a junction. Stop, find the last junction with an unexplored opening, and restart your search from there.
Mazes for Kids: Which Grid Size Fits Which Age, and What It Actually Builds
The size ramp does the age-matching for you, because the levels step 8 → 10 → 12 → 14 → 16 and hold at 16×16 from level five on. The 8×8 opening level is the right entry point for roughly ages four to six: 64 cells, a visible corner-to-corner diagonal, and a route short enough to hold in the head. Around seven to nine, the 10×10 and 12×12 boards start to demand actual branch tracking rather than following whatever passage looks open. From ten up, 14×14 and 16×16 are where the technique sections above stop being optional and start being how you finish without wandering.
What a maze trains, concretely, is visual path-planning: scanning ahead of your hand, holding a route in working memory while you execute it, and recognising a dead end as information rather than a failure. It is also one of the few puzzles where a child can be given a genuinely complete method — the left-hand rule — and watch it work every time, which is a cleaner first taste of an algorithm than most things on a screen. Watching the Steps counter across a few tries makes improvement measurable in a way that a win/lose result does not.
For a young player, one adjustment helps more than any other: get them using the D-pad or arrow keys rather than swiping. Swiping glides several cells at once and children lose track of where the token went; single steps keep the movement matched to the decision. Have them say the direction out loud before pressing, and let them finish a dead end and walk back out rather than restarting — the walk back is the part that teaches.
Why Every Maze Here Has Exactly One Solution
The generator is a recursive backtracker, also called randomised depth-first search, and it is worth knowing how it works because it explains how the mazes feel to play. It starts in the top-left cell with every wall in the grid standing. It looks at the neighbouring cells it has not visited yet, picks one at random, knocks out the wall between them, and moves there — repeating until it reaches a cell with no unvisited neighbours. Then it backs up along the route it came by, one cell at a time, until it finds a cell that still has an unvisited neighbour, and starts carving again from there. It ends when it has backed all the way out to the start.
Because a wall only ever comes down when the algorithm moves into a cell it has never visited, no passage can ever close a loop. On an n×n grid it carves exactly n²−1 passages to connect n² cells — 255 passages for the 256 cells of a 16×16 board — which is the definition of a spanning tree. One consequence matters at the keyboard: there is exactly one route from your corner to the door, so there is no such thing as a clever shortcut here, and any move that is not on that route is a move you will have to undo.
The backtracker's signature is long, snaking corridors and comparatively few dead ends, because it keeps pushing forward as long as fresh territory exists and only branches when it is forced to. That is why the solution path winds so far past the straight-line distance, and why swipe-gliding pays off so much on these particular mazes. Every level generates a fresh one, so there is nothing to memorise between runs.
Questions people ask
Is the game free, and do I need to download anything or make an account?
It is free and it runs in your browser tab — no download, no install, no signup. There are no accounts, so nothing is stored on a server; your best result is kept locally on the device you played on.
What are the controls?
Arrow keys or WASD move one cell per press. The on-screen D-pad below the board does the same for touch. Swiping or click-dragging across the board itself glides you down the corridor in that direction until you hit a wall or reach a junction with three or more openings.
Can a maze ever be unsolvable, or trap me in a loop?
No. Every maze is generated as a perfect maze — one single path between any two cells, no loops anywhere. If it feels like you are going in circles, you backtracked without noticing, usually by gliding past a junction. Put your hand on one wall, follow it consistently, and you will reach the door.
How big do the mazes get, and does it ever end?
Level 1 is 8×8 and each clear steps you up: 10×10, 12×12, 14×14, then 16×16, which is the cap from level five onward. Levels keep coming after that with a new 16×16 maze each time, so the run ends when you press Finish & see results.
Do fewer steps get me a better score?
Points come from levels cleared — 20 plus twice the grid size per level, so 36 for an 8×8 and 52 for a 16×16. The Steps counter is there as a personal target: the absolute minimum is 14 steps on the 8×8 and 30 on the 16×16, and cutting waste means more levels cleared in the same sitting.
How to play Maze Escape
- Swipe or use arrow keys to move through the maze.
- Reach the glowing exit to finish the level.
- Every level generates a bigger maze. Speed counts!
Maze Escape is free to play online at Trivio — no download, no signup, works on desktop and mobile. Your best score is saved in your browser.