Slitherlink Puzzle Online
Draw one continuous loop through a dot lattice while every numbered cell counts exactly how many of its four surrounding edges belong to that loop.
What makes Slitherlink different from a normal grid puzzle?
Slitherlink is solved on the edges between dots rather than by filling the cells themselves. A numbered cell gives a local edge count: a 0 means none of its four borders can be part of the loop, a 1 means exactly one border is used, and so on through 3. Those clues overlap because neighboring cells share edges, so one deduction immediately changes what is possible in more than one place.
The second rule is just as important as the clues. Every dot touched by the loop must have exactly two line segments. A dot cannot have a dead end with only one line and it cannot branch into three directions. This vertex rule turns a local clue puzzle into a network puzzle. An edge that looks possible from one numbered cell can still be impossible because it would leave a nearby dot with the wrong degree.
The final rule is global: every used line segment must belong to one continuous closed loop. You are not allowed to finish two separate loops, and you cannot close a small loop early while unsolved line segments remain elsewhere. That single-loop requirement is why Slitherlink can stay interesting even when many numbered clues have already been satisfied.
How the controls are designed for desktop and mobile
This implementation uses explicit Line, Cross and Erase tools rather than forcing every tap through a three-state cycle. On a phone, choose Line and tap edges you believe belong to the loop. Choose Cross to mark edges that cannot be used, or Erase to clear a mark. On desktop, right-clicking an edge also places a cross. Keyboard shortcuts L, X and E switch the active tool, while H requests a logic hint.
Cross marks are optional working notes. A completed puzzle only needs the correct loop edges, but crossing out impossible edges makes the remaining choices easier to read. The browser saves both line and cross states so a refresh does not erase the reasoning you have already recorded.
Easy, Medium and Hard are different clue densities
The frozen source bank contains 5×5, 6×6 and 7×7 boards. Easy sources keep more numbered cells visible, Medium removes more clues, and Hard keeps the smallest clue set that the deterministic generator accepted while preserving one exact solution. Difficulty is therefore connected to the information actually shown on the board rather than being a decorative label attached after generation.
Practice selection is deterministic once the URL seed is fixed. The seed chooses one source from the requested size and difficulty pool and then applies one of eight geometric symmetry transforms. Reloading the same seeded URL restores the same source orientation, while Fresh Practice creates a new seed. This makes a board reproducible for debugging, sharing and local resume without needing a server-side puzzle ID.
How uniqueness is verified
The generator first creates a connected region with no holes and uses that region's boundary as the target loop. It calculates the full Slitherlink clue set from the boundary, then removes clues while an exact solver continues to report one solution. A candidate is rejected if the clue set admits zero solutions, more than one solution, a branched network, a dead end, or multiple disconnected loops.
Release verification does not simply trust the generator. A separate JavaScript verifier reads the committed frozen bank and solves each puzzle again from the public rules: numbered cell edge counts, vertex degree zero-or-two and one connected closed loop. It also transforms every source through all eight rotations/reflections and verifies those transformed boards independently. That gives 45 source checks and 360 variant checks before deterministic Practice and Daily selection tests are counted.
Using Hint, Check and Undo without mixing records
Hint fills the next edge according to the verified solution, choosing either a line or a cross. Check looks only for contradictions that are already provable from the current partial state, such as too many lines around a clue or a completed dot with an impossible degree. Undo and Redo restore exact edge states, including crosses, so you can inspect a deduction chain without manually reconstructing the board.
Hint, Check, Undo, Redo and Restart mark the current run Assisted. The puzzle remains fully playable, but an assisted completion does not overwrite a clean personal best. That separation keeps local records meaningful: a fast solve after several hints is useful practice, but it is not presented as the same thing as an unaided clean solve.
Local autosave and privacy
Practice and Daily progress are stored only in this browser. The autosave includes the exact source, symmetry, edge states, move count and Clean/Assisted status. No account, cloud database or paid API is required. Reopening the same Practice seed or Daily date restores the board that belongs to that key.
Slitherlink is a Practice specialist and does not enter the platform's existing Today’s 6 rotation. Completing a Slitherlink board therefore cannot change the six-game Daily Score. Daily Slitherlink has its own browser-local history and streak so people who want a recurring loop puzzle can build that habit without silently changing the older daily contract.
A useful solving order
Start with 0 clues because all four surrounding edges can be crossed immediately. Then inspect 3 clues, especially near borders and corners, because the vertex degree rule sharply limits how their three required lines can turn. After every line placement, scan its two endpoint dots: if a dot already has two lines, every other incident edge can be crossed; if it has one line and only one undecided continuation, that continuation is forced.
Next revisit neighboring numbered cells that share newly decided edges. A clue with its full line count reached lets you cross every remaining border. A clue that still needs exactly as many lines as undecided borders forces all of them. Finally, watch connectivity. If a move would close a small loop before the rest of the board is attached, that move cannot be correct even if its local clue counts still look legal.