There is a family of games that all look different and are secretly the same game: 2048, the fifteen puzzle, water sort, most "remove the block" boards. They are all state-space puzzles, and the mathematics underneath them is older than any of the screens they run on.

The puzzle that broke a lecture

In 1879 a puzzle maker in New York sold a small tray with fifteen numbered tiles and one empty square. The task was to arrange them in order. It became a genuine craze — and then someone noticed that half the arrangements people were buying could never be solved.

The proof was published by William Woolsey Johnson in 1879, and it is a beautiful piece of reasoning. Take the tiles in their reading order, skipping the empty square, and count how many pairs are out of order — that is the permutation's parity. For the fifteen puzzle, every legal move changes the parity of the tile order and the parity of the empty square's row at the same time. The two effects cancel. So a deal whose starting parity does not match the target's is not difficult; it is impossible.

Half of the boards sold in 1879 were not hard puzzles. They were proofs that the puzzle had been printed wrong.

That is worth sitting with, because it changes how you read a failure. Some players spent weeks on an unsolvable tray and concluded they were bad at puzzles. They were not. They had been handed a broken board.

Why 2048 feels different

2048 is the same family — a board, tiles, moves that shift everything in one direction — but it removes the trap. There is no fixed target arrangement, only merges, so almost any sequence of reasonable moves produces progress. The puzzle is not "reach this state", it is "manage this space".

The design lesson is precise: a sliding-tile game is interesting when the goal is a property, not a position. A property can be approached from many directions. A position can be unreachable, and then the game has wasted your time.

That is also why the fifteen puzzle needs a solver guarantee. If you ship a shuffle that produces a dead board, your players do not experience a hard puzzle — they experience a bug that looks like their own incompetence.

Parity is not the only ceiling

State-space puzzles have a second constraint that matters more for design than for players: the size of the space.

The fifteen puzzle has about 10^13 reachable states. That is large enough to be hard and small enough to be searchable — which is why it has been solved optimally, and why a computer can tell you the minimum number of moves from any position.

2048's space is larger and its branching factor is tiny: four moves, most of them equivalent. A game with few choices per turn and a huge space is a game of long-term planning rather than calculation. A game with many choices per turn and a small space is a game of search. The feel of a puzzle comes from which of those two ratios it has.

What players actually learn

The interesting thing about parity puzzles is that they teach a specific mental habit: checking whether a goal is reachable before optimising the route.

That habit transfers. Most people, given a task, immediately start moving. The player who pauses to ask "is this actually possible, and how would I know?" saves themselves the worst kind of work.

A note on respect for the board

There is an odd dignity to these games. They do not pretend to be worlds. They are a rule, a space, and a goal, and everything that happens is a consequence of those three things. When you fail, it is either because the board was broken or because your reasoning was — and both are knowable.

That is more honesty than most games offer.