Wander Star — a chain you can predict before you fire it

The challenge

Wander Star hands the player exactly one action: tap a block. Everything after that is automatic. The chain fires, turns, splits and dies on its own, and the player watches. That makes it a pure planning puzzle — the entire game is the few seconds before the tap, when the player is tracing a path through the board in their head.

Which creates a precise problem: if the player cannot accurately predict what the chain will do, there is no planning, only tapping and hoping. And prediction here is genuinely demanding. Arrows point in eight directions including diagonals, so the chain can turn at any block it lands on. The player has to follow a route through a 7x7 grid, mentally applying each arrow in sequence. One misread square and the whole prediction is wrong.

Split blocks make it harder in a way that is central rather than incidental. A block with two arrows divides the chain into two paths that then travel independently, each turning at whatever it lands on. The player is now tracking two routes at once, and finding the right split point is the actual skill the game is built around. Any ambiguity about what a block does, or where a chain currently is, destroys that skill entirely.

The second problem is the stars themselves, and it is a rules-clarity problem. Stars cannot be destroyed — chains pass straight through them. They are collected by clearing the blocks underneath so they fall to the bottom row. This is not what a player assumes on first contact; the natural guess is that the chain should hit the star. A player operating on the wrong model will make reasonable-looking moves that achieve nothing, and will not understand why.

Third, this is a hand-made puzzle game, which means the content is the cost. Every level has to be authored and verified, with its arrow layout, split placements, star positions and limit tuned so it is solvable and so the solution is findable rather than accidental. Generated levels would be cheap and would produce exactly the noise this genre cannot survive — a chain puzzle where levels are random is a game where the player cannot tell skill from luck.

Fourth, there are two limit types — a move count and a timer — and they pull in opposite directions. A move-limited level rewards thinking as long as you like. A timed level punishes exactly that. Mixing them badly means the game is asking for careful planning in one level and panic in the next, with no signal about which mode of play applies.

Our approach

The board was built so a chain can be traced before it is fired. Arrow directions are unambiguous at a glance, blocks are visually distinct from stars, and the 7x7 grid is small enough to hold in the head — large enough for real routes, small enough that a player can actually follow one. The size is a design decision, not a default: a bigger board would make prediction impossible and turn the game into guesswork.

Split blocks are marked clearly as carrying two arrows, because they are the single highest-consequence block on the board. A player who misses that a block splits will predict one path and get two, which invalidates everything they planned. Making them obvious is what keeps the core skill learnable.

The star rule is taught by the board rather than by text. Stars sit above the columns, visibly waiting, and drop in as blocks clear — so the relationship between clearing a column and releasing a star is something the player watches happen rather than reads about. Because stars are indestructible and chains pass through them, the player quickly learns that the target is never the star itself but the blocks beneath it, which is the one counter-intuitive rule in the game.

Levels are designed by hand, every one. That is the expensive decision and the correct one. In a chain puzzle, a generated level is as likely to be trivially solvable by a random tap as it is to be impossible, and either outcome teaches the player that outcomes are not theirs. Hand-authored levels mean each board has an intended solution and a limit tuned so that finding it matters — which is also what lets difficulty climb smoothly through split placement and star position rather than through arbitrary tightening.

The two limit types are kept as distinct level types rather than mixed into one. Move levels give a fixed number of taps and reward unhurried planning. Timer levels give a fixed number of seconds and test recognition speed on patterns the player has already learned. Separating them means each level makes a single clear demand, and the player knows which kind of thinking a board is asking for before they start.

The three power-ups deliberately do not consume a move, which is what makes them usable. A helper that costs the resource you are short of is not help. Bomb clears every block while leaving stars to fall, for a board that has become unreadable. Swap rearranges the contents of every block, for a layout that has stopped offering routes. Star simply collects one from a column that still holds one, for when a single star is all that stands between the player and the exit.

Controls stayed at one finger and one tap throughout, and the board was kept uncluttered — in a game where the whole activity is reading a grid, every piece of decoration competes with the thing the player is trying to see.

The result

A chain puzzle where the chain can actually be predicted. A 7x7 board, unambiguous arrow directions and clearly marked split blocks mean a player can trace a route in their head before committing — which is the difference between a planning game and a tapping game.

Split blocks are the skill ceiling, and they are made obvious rather than subtle, because a missed split invalidates an entire prediction.

The counter-intuitive rule — that stars are indestructible and must be dropped by clearing beneath them — is taught by watching the board rather than by a tutorial screen.

Every level is hand-designed. In a chain puzzle, generated levels produce the one thing the genre cannot survive: a player unable to tell skill from luck.

Move levels and timer levels are kept separate, so each board makes a single clear demand instead of asking for careful planning and speed at the same time.

The three power-ups cost no moves, because a helper that consumes the resource you are short of is not help.

Live on Google Play, free.