Where these ten puzzles came from

The rough order of events

Two of these puzzles are Victorian, four are products of the 1970s and 80s boom in boxed and printed puzzles, and the rest arrived with home computing. The dates below are the generally accepted ones; where a date is disputed, the section further down says so.

The two Victorian puzzles

Tower of Hanoi

Invented by the French mathematician Édouard Lucas and sold as a toy in 1883, complete with an invented legend about a temple, sixty-four discs and the end of the world. The legend was marketing. The mathematics was real, and the puzzle has been a standard teaching example of recursion ever since — the recursive solution is four lines long and the iterative one is not obvious at all.

The fifteen puzzle

A craze in the early 1880s, and the origin of one of the more persistent falsehoods in puzzle history. It is generally credited to Noyes Chapman, a postmaster in New York State, in the 1870s. Sam Loyd, a prolific and self-promoting puzzle writer, claimed for decades to have invented it; the claim is not supported by the record, and puzzle historians have been correcting it ever since.

Loyd did, however, popularise the version with two tiles swapped and offer a prize for solving it. That version is provably impossible, and the impossibility was already published mathematics when the prize was announced. The parity argument is short enough to follow in one sitting.

The boxed and printed era

Word Search

Usually credited to Norman E. Gibat, who published one in a small free-circulation paper in Norman, Oklahoma, in 1968. It spread through schools first, because teachers found them useful, and from there into the puzzle sections of newspapers everywhere.

Connect Four

Sold as a boxed game by Milton Bradley in 1974, though the underlying idea — four in a row, gravity-fed — is older than the box. Its interesting date is 1989, when it was solved independently by more than one researcher: with perfect play the first player wins, and the winning opening move is the centre column. Why the centre matters follows from counting the winning lines through each square.

Sokoban

Created by Hiroyuki Imabayashi and published in Japan in 1982 by Thinking Rabbit. The name means “warehouse keeper”. It has had an unusual second life as a research problem: because pushes cannot be undone and the arrangements multiply so fast, it is a standard hard benchmark in automated planning. What makes a level hard goes into why.

Nonograms

Invented in Japan in 1987, and unusually, invented twice at almost the same moment by two people working separately — Non Ishida and Tetsuya Nishio. Ishida's name is the source of the word nonogram. They travelled internationally under several names, including picture crosswords and Picross, before settling down.

The ones that arrived with computers

Minesweeper

Games of this shape existed on home computers through the 1980s, but Minesweeper became universal because it was bundled with a desktop operating system in the early 1990s and therefore installed on an enormous number of office machines. It is probably the most-played puzzle on this site's list by a very large margin, and almost nobody who played it ever chose to install it.

It is also the only puzzle here that can force a genuine guess, which is a property of the game rather than a defect. Choosing that guess well is a real skill.

Lights Out

Released as a handheld electronic toy by Tiger Electronics in 1995, following earlier toys built on similar switching ideas. Its afterlife has been mathematical: the puzzle is a system of equations where every value is zero or one, and solving it by hand is parity rather than trial and error.

Memory

The odd one out, in that nobody invented it. Matching pairs from a face-down layout is a traditional card game with no identifiable author and no fixed date, played with ordinary playing cards long before anyone printed a dedicated set.

Sudoku, and the twenty-five year delay

Sudoku's history is the most misreported of the ten, mostly because it has three separate beginnings.

The mathematical ancestor is the Latin square — a grid where each symbol appears once per row and column — studied by Euler in the eighteenth century. That is genuinely related, but it is not the puzzle: Sudoku's extra constraint is the box, and the box is what makes it solvable by human reasoning rather than search.

The puzzle itself, box constraint and all, was published in an American puzzle magazine in 1979 under the name Number Place. It is attributed to Howard Garns, a retired architect, who died before it became famous and never knew what he had made.

It was then picked up by the Japanese publisher Nikoli in the 1980s, given the name it now carries, and refined in ways that matter: symmetrical patterns of givens, and a guarantee of a unique solution. It was popular in Japan for two decades while the rest of the world ignored it.

The global craze arrived in 2004, when a New Zealander named Wayne Gould persuaded a British newspaper to print puzzles from a generator he had written. Within a year it was in newspapers worldwide. The puzzle was twenty-five years old at the time.

Why these ten, and not new ones

Every puzzle here is old, freely reimplementable, and belongs to nobody. That is deliberate. A classic has been played by millions of people, which means its rules have been sanded smooth by use: there are no ambiguous edge cases left, and the difficulty curve is understood.

It also means the interesting work is not inventing a mechanic. It is in the parts nobody sees — whether the generator gives you a fair board, whether the puzzle is playable from a keyboard, whether the explanation of how to solve it is any good. How we prove a board is fair covers the first of those.

We have not tried to improve any of these games. What is on this site is our implementation of them, written from scratch, and the games themselves are not ours to claim.