Yes, you can play Lights Out without downloading anything. The full puzzle runs directly in your browser at Lights Out, with no installer, app store account, browser plug-in, or sign-up required. The 4x4 board loads as a self-contained HTML page, stores all 16 light states as local booleans, and never uploads gameplay data to a server. You click or tap any tile to press it, watch the affected cross update on the same screen, and aim to turn every one of the 16 lights off. Because nothing leaves your device, you can play on a school computer, a work laptop, a phone, or a tablet without administrator privileges, and you can take as long as you want without a timer pressuring your score. The same browser session also keeps the keyboard controls and the boss-key gesture working between visits, so you can switch to a spreadsheet view with a double-Escape tap and return to the board the same way.
The toggle rule, in plain words
Lights Out is a binary toggle puzzle. Every tile is either on or off, and pressing a tile reverses its own state together with the states of its immediate up, down, left, and right neighbors that are still inside the board. Diagonal tiles do not change. The same rule is documented in the GNOME Lights Off help and independently confirmed by UC Berkeley GamesCrafters, so the cross-neighbor behavior is well established across published sources, not invented for this build.
The number of tiles that change on a single press depends only on the cell's location, which makes the geometry easy to read at a glance.
| Press type | Cells affected | Count |
|---|---|---|
| Corner tile | Self + 2 in-bounds orthogonal neighbors | 3 |
| Non-corner edge tile | Self + 3 in-bounds orthogonal neighbors | 4 |
| Interior tile | Self + 4 in-bounds orthogonal neighbors | 5 |
A corner press on this 4x4 board therefore touches three tiles, an edge press touches four, and an interior press touches five. The selected tile carries a visible outline, and the other cells in its affected cross receive a lighter-border preview. That preview is explanatory only; hovering never changes the state of a tile, and the affected cells are always the same orthogonal cross.
How to play Lights Out without downloading
- Open the browser puzzle at /games/lights-out/ in any modern browser; the board, status row, and controls render in a single page.
- Read the starting pattern: six lights are on at zero-indexed cells 1, 2, 3, 5, 9, and 10, and the cursor starts at cell 0.
- Select any tile by clicking, tapping, or pressing the Arrow keys; the selected cell carries a visible outline and announces its row, column, and current state in its accessible label.
- Press the selected tile by clicking the cell, tapping it, choosing Press selected, or pressing Space; the tile and only its in-bounds orthogonal neighbors flip on the same animation frame.
- Use the lighter-border cross preview around the selected cell to confirm which other tiles will change before you commit to a press.
- Think in press parity: pressing the same cell twice restores its prior light pattern, so only the parity of every press count affects the final board.
- After at least five deliberate presses, choose Check board, press Enter, or press C to evaluate the run.
- If every one of the 16 lights is off, the run scores exactly 1,000 points and the controls freeze.
- If any light is still on, the run records one mistake and stays editable; press additional tiles, reach all off, and check again to recover.
- A second wrong complete check deadlocks the run; press Restart to restore the exact six-light initial pattern, the cell-zero cursor, zero moves, zero mistakes, and zero score.
Why the optimum is exactly five moves
The browser Lights Out fixture is paired with an independent mathematical oracle, and the oracle's numbers are public. It builds one 16-bit effect column for each board cell, runs binary Gaussian elimination to find matrix rank 12, and so derives nullity 4 (since 16 - 12 = 4). It then enumerates all 65,536 binary press masks and counts 4,096 reachable effect states. For the disclosed initial board, exactly 16 of those masks produce an all-off result, which agrees with the rank-nullity calculation that 2^4 = 16 solutions must exist for any reachable target.
Among those 16 all-off solution masks, the Hamming weights are 5, 7, 9, or 11. Exactly one mask has weight 5, which means the shortest solution is unique and cannot be beaten by any shorter path. The oracle also confirms that no mask of weight 0 through 4 produces an all-off board for this fixture. That same proof is what justifies the Check board boundary: a check during moves zero through four is objectively incomplete, so the runtime records no mistake at that stage.
You do not need to follow the Gaussian elimination to play well. A useful mental shortcut is press parity. Pressing cell 0 (a corner) flips cells 0, 1, and 4, three tiles. Pressing cell 0 a second time flips those same three tiles back, so the net effect after two presses of cell 0 is zero flips. Therefore only the odd-or-even parity of every press count matters for the final board, even though the displayed move counter records every physical press.
Scoring, repair, and the two-error state machine
A successful Check board at or beyond five moves awards exactly 1,000 points. The score does not change with speed, device, input method, or the number of extra moves you used, because correctness is the only scoring predicate. The runtime stores the 16 current light booleans rather than a hidden five-button answer, so any valid all-off state reached after the proven minimum is accepted, and longer valid solutions remain valid as standard toggle algebra requires.
A check during moves zero through four records no mistake, because the oracle has not yet reached the proven lower bound and the board is mathematically incomplete. At five or more moves, a board with any remaining light represents a complete but incorrect submitted path. The first such wrong complete check records one mistake and leaves the run editable: you can keep pressing tiles, reach all off, and check again to repair the run. A second wrong complete check deadlocks the run until you press Restart.
Both success and deadlock are terminal states. The runtime freezes clicks, Space presses, cursor movement, selection changes, and further checks after either outcome, and these guarantees live in the pure state-transition functions covered by tests rather than depending only on disabled visual controls. The first-error repair path is itself tested by submitting a deliberate five-move failure, canceling its residual parity, applying a valid solution, and checking again.
Browser play versus a downloaded app
| Aspect | Downloaded Lights Out app | Lizely browser Lights Out |
|---|---|---|
| Install required | Yes | No |
| Account or sign-up | Sometimes | None |
| Board content | Random or paid levels | Fixed 4x4, six lights on at cells 1, 2, 3, 5, 9, 10 |
| Solvability proof | Varies by source | Independent GF(2) oracle |
| Shortest path | Varies | Five moves, unique among the 16 solutions |
| Score | Varies | Exactly 1,000 for any all-off state at or beyond five moves |
| Wrong-check recovery | App-dependent | Repair after first wrong check, deadlock after second |
| Privacy | May upload state or require login | Local only, no upload |
| Mobile fit | App store required | Designed for a 390-pixel viewport |
Keyboard, touch, and accessibility
Full keyboard control covers every interaction. Arrow keys move the selected cell through the rectangular grid without wrapping. Space presses the selected tile. Enter or C runs the check. The global repeat flag is ignored, so holding Space cannot produce an uncontrolled burst of parity-changing moves and holding an Arrow key cannot skip through the puzzle. Touch users tap any tile to press; switch-control and keyboard users can rely on Press selected, which performs the same operation as clicking the board and helps keep a stable cursor on the same screen.
Each cell exposes its row, column, on-or-off state, and selection state in an accessible label, and the words On and Off as well as surface contrast communicate state without relying on color alone. The shared double-Escape gesture activates the spreadsheet-style boss screen, and a second double Escape returns to the board. The square four-column grid, concise status row, large native buttons, and wrapping controls are designed to fit a 390-pixel viewport without horizontal scrolling, so the same browser page works on a phone, a tablet, and a desktop without a separate mobile build.
Why this edition is not the GNOME or Berkeley level
The starting pattern, size, check policy, score, and two-error state machine are original product fixtures. They are not copied from GNOME, UC Berkeley, the historic handheld game, a puzzle book, or a daily challenge. GNOME's maintained Lights Off help is cited for the selected-plus-non-diagonal-neighbor behavior, and UC Berkeley GamesCrafters independently confirms both the orthogonal toggle rule and the all-off goal that define the genre. The Lizely fixture is a recreational fixed puzzle, not an intelligence test, cognitive or medical assessment, educational placement tool, or employment screen, so the 1,000-point score measures puzzle completion rather than ability, and time is not scored because repeated play naturally becomes easier through memory.