Sudoku Origin: Where the Puzzle Really Came From

The sudoku origin story is usually told wrong in one of two directions. One version makes it ancient — a Japanese puzzle of great antiquity, or something Euler invented in the eighteenth century. The other makes it a corporate product with a clean birthday. Neither is right. What actually happened is stranger and more interesting: a mathematical structure studied for centuries, an American magazine puzzle from 1979 whose creator was never officially named, a Japanese publisher who gave it a name and two extra rules, and a retired judge from New Zealand who spent six years writing software before talking a London newspaper into printing it.

Here is that sudoku history in order, with the disputed parts flagged as disputed. Where the record is genuinely uncertain — and a surprising amount of it is — this page says so rather than picking the tidiest version.

The short answer

Sudoku descends from Latin squares, was published in essentially modern form as "Number Place" in an American puzzle magazine in 1979, was named and refined in Japan from 1984, and became a global phenomenon after appearing in British newspapers in late 2004.

Latin squares: the ancestor, not the origin

A Latin square is an n×n grid filled so that every symbol appears exactly once in each row and each column — sudoku's constraint minus the boxes.

Arrangements of this kind turn up in many places long before anyone thought of them as puzzles: in decorative and mystical grids, in the magic squares of several mathematical traditions, and in the combinatorial curiosities that circulated among mathematicians. They were objects of study, not entertainment.

The formal treatment belongs to Leonhard Euler, who worked on them in the 1770s and 1780s and gave them the name we still use — he wrote the symbols as Latin letters, which is precisely why they are "Latin" squares. His famous problem of thirty-six officers, posed in 1782, asked whether six regiments of six ranks each could be arranged in a 6×6 grid with no rank and no regiment repeated in any row or column. He conjectured it was impossible; a proof came much later, and it held.

Two things need saying clearly here, because this is where most sudoku origin accounts go astray.

First, Euler did not invent sudoku. He studied the mathematical structure underneath it. That's a genuine ancestor, not an origin — roughly the relationship between the study of graphs and the invention of a maze puzzle.

Second, a Latin square is not a sudoku. Sudoku adds a third constraint that Latin squares don't have: the nine 3×3 boxes. That extra rule is what makes the puzzle solvable by human-scale reasoning rather than brute search, and it is not present in Euler's work at all. The full statement of what the modern grid requires is set out in sudoku rules.

The French near-misses of the 1890s

In the 1890s, French newspapers published 9×9 number puzzles that came remarkably close to modern sudoku — close enough that some are effectively the same puzzle.

This part of the story was largely unknown until researchers went back through the archives in the 2000s. Papers including Le Siècle and La France ran number-placement puzzles derived from magic squares, and over a few years the format drifted toward something familiar: a 9×9 grid, some numbers given, the rest to be deduced. At least one puzzle from La France in 1895 used a grid divided into 3×3 regions with the digits 1 to 9 in each, which is functionally the puzzle we play now.

These puzzles disappeared around the First World War and left no visible line of descent. Nobody involved in the 1979 American revival appears to have known about them. So they're best understood as an independent invention that didn't take — a real part of the sudoku origin story, but a branch rather than the trunk.

There's a lesson buried in that dead end, and it recurs at every later stage: distribution matters as much as design. A good puzzle without a durable publishing habit around it simply vanishes, however well built it is.

Number Place, 1979

The modern puzzle first appeared as "Number Place" in the May 1979 issue of Dell's Pencil Puzzles & Word Games, and the rules were already exactly the ones we use today.

There was no fanfare. It was one puzzle type among many in an American newsstand magazine, sitting alongside crosswords and word searches, with the same three constraints — rows, columns, boxes — and the same one-solution requirement. Dell ran it steadily but it never became a headline feature. For most of the 1980s it was, in the United States, an ordinary puzzle of no particular distinction.

Howard Garns: credited, never confirmed

Howard Garns, a retired architect from Indianapolis, is widely credited with creating Number Place — but Dell never officially confirmed it, and he died before the puzzle became famous.

The attribution rests on a piece of circumstantial detective work done years later. Dell magazines listed contributors without saying which puzzle each person made. Researchers noticed that Garns's name appeared on the contributor lists of issues that contained Number Place, and was absent from issues that didn't. That correlation is the evidence. It is suggestive and it is widely accepted, but it is not a confirmation, and Dell has never made an official statement.

Garns died in 1989, sixteen years before sudoku became a worldwide craze. If the attribution is right, he never knew what he had made, never earned anything beyond a magazine contributor's fee, and never saw his puzzle on the front of a newspaper. That's the honest version — and it's worth stating plainly, because a lot of writing about the sudoku origin presents the Garns attribution as settled fact when the record genuinely isn't.

Nikoli and the name

The Japanese publisher Nikoli picked up Number Place in 1984, gave it the name we know, and turned it from a filler puzzle into a craft.

Nikoli published it in April 1984 in its magazine under the title Sūji wa dokushin ni kagiru, usually rendered as "the digits must be single" or "the numbers must occur only once." Nikoli's president, Maki Kaji, shortened it to sudoku — su for number, doku for single. Kaji, who died in 2021, is often called the godfather of sudoku, though he was careful throughout his life to say that he named the puzzle rather than invented it.

The two rules Nikoli added

In 1986 Nikoli introduced two conventions that shaped the puzzle as most people now experience it:

1. A cap on the number of givens, generally cited as no more than 32.
2. Rotationally symmetric clue patterns — the arrangement of givens looks the same when the grid is turned 180 degrees.

Neither is a rule of the puzzle in any logical sense. A sudoku with 40 asymmetric clues is perfectly valid and perfectly solvable. But the conventions gave published puzzles a recognizable elegance, and they signaled something more important: at Nikoli, puzzles were hand-made by named setters and judged on the quality of the solving path, not just on whether they had a unique answer. That editorial culture is a genuine part of why the puzzle survived to be exported.

Japan also happened to be an unusually good market for it. Crosswords work poorly in Japanese, where the writing system makes letter-by-letter grids awkward, so there was an existing appetite for a puzzle that used no language at all. The property that made sudoku welcome in Tokyo in 1984 is the same property that later let it cross every border it met without a translator.

One consequence of Nikoli's involvement persists today. Nikoli holds the trademark on the name "Sudoku" in Japan, which is why other Japanese publishers still call it Number Place, usually shortened to nanpure. In Japan, the puzzle's American name stuck.

Wayne Gould and the 2004 boom

Sudoku went global because a retired Hong Kong judge saw one in a Tokyo bookshop, spent six years writing a program to generate them, and then persuaded a newspaper to print the output.

Wayne Gould, born in New Zealand, came across the puzzle on a visit to Japan in 1997. He was struck by it, and — this is the part that makes the story unusual — he responded by teaching himself to write software that could generate and grade puzzles automatically. That project ran for roughly six years.

Then he did the other necessary thing: he sold it. Gould approached newspapers with a proposition that worked in their favor, supplying puzzles at little or no cost in exchange for a credit and a pointer to his own site. The Times of London began publishing sudoku in November 2004. Other British papers followed within weeks, and there's a reasonable case that at least one had run a version slightly earlier under a different name, which is why "the first newspaper sudoku" is a claim worth hedging.

What followed in 2005 was genuinely extraordinary. British papers fought a circulation war with competing sudoku sections. American papers picked it up. Books charted. Television produced sudoku programming. Puzzle apps arrived with the smartphone a couple of years later and never left. By March 2006 the first World Sudoku Championship was held in Italy, and the puzzle had a competitive scene.

Two structural facts explain the speed of it. The puzzle needs no language, so it crossed borders without translation — a decisive advantage over the crossword. And Gould's generator meant that supply was effectively unlimited and effectively free, so any newspaper that wanted one could have one tomorrow — no commissioned constructor, no lead time, and a rule set short enough to introduce in a caption rather than an article. The same computational approach is what lets the free generator at puzzlemakeronline.com produce fresh graded puzzles today.

This is the part that gets mislabeled most often. 2004 is routinely described as the puzzle's invention. It wasn't. It was the moment the puzzle's distribution problem got solved — the same problem that had quietly killed the French grids a century earlier.

What the mathematicians settled afterwards

Once the puzzle was culturally significant enough to be worth serious computing time, two long-open questions about the grid were closed.

In 2005, Bertram Felgenhauer and Frazer Jarvis calculated how many valid completed 9×9 grids exist. After accounting for the symmetries that make one grid a relabeled copy of another, the count of essentially different grids is 5,472,730,538. In 2012, a team led by Gary McGuire used an exhaustive computer search to prove that no sudoku with a unique solution can be built from fewer than 17 given digits — a question that had resisted every attempt at an elegant proof for years, and which was settled in the end by brute force rather than insight.

Neither result changes how anyone solves a puzzle. Both are worth noting for what they say about status: a puzzle only attracts that kind of effort once enough serious people care about it, and in 1979 nobody would have spent a supercomputer year on Number Place.

Competitive solving formalized on the same timeline. The first World Sudoku Championship in 2006 created a demand that standard grids couldn't meet on their own — when every finalist can finish a hard 9×9 quickly, organizers need variants that still separate them — which is a large part of why the variant family grew as fast as it did.

The digital era, and what it replaced

The stage after the newspaper boom gets less attention than it deserves, because it changed how most people now meet the puzzle at all.

Before software, a sudoku was a scarce object. You got one a day from your paper, or you bought a book and worked through it in order. Scarcity shaped the habit: you solved the grid you were handed, at whatever difficulty it happened to be, and if you got stuck you either persisted or lost the day. Generation removed the scarcity completely, and three things followed from that.

Difficulty became a choice rather than a fact. When a program can grade a grid by which technique it requires, you can pick the tier you want to practice instead of accepting what arrives. Deliberate technique practice is a modern habit; it was barely possible when your supply was one puzzle a morning.

The bookkeeping moved off the solver. A digital grid updates its candidate marks after every placement. That is the single largest practical difference between solving now and solving in 2005, and it quietly changes what the puzzle tests. On paper, a meaningful share of the difficulty is clerical. On screen, almost none of it is, so what's left is closer to pure pattern recognition.

Variants proliferated. Once grids were generated rather than drawn by hand, changing the constraints became cheap. Diagonal grids, irregular boxes, larger boards and the sum-based variants could all be published at volume, and sudoku stopped being one format and became a family.

Something was lost in the exchange. A Nikoli grid built by hand has a deliberate solving path that tends to flow, where a generated grid of the same rating can feel scattered by comparison. Enough solvers notice the difference that hand-constructed puzzles still have a following, which is a fair verdict on what the craft was actually contributing.

The request itself, though, hasn't changed since 2005. Whether someone types free sudoku, online sudoku, sudoku online free or free online sudoku into a search box, they want the same thing a newspaper reader wanted: a grid to solve now. Search suggestion boxes append text to whatever you have already typed, which is why the logs are full of doubled phrases like sudoku sudoku online and even stacked ones like sudoku online sudoku online sudoku online — autocomplete artifacts rather than anything anybody meant. Only the delivery mechanism moved, from newsprint to a browser tab.

Sudoku origin myths worth retiring

Almost every popular claim about where sudoku came from is wrong in a specific, checkable way.

The last three are the ones to be careful about, because they're repeated confidently in places that ought to know better.

Why this particular puzzle lasted

Plenty of number puzzles have come and gone. A few properties of this one made it durable.

It is language-independent, so it needs no localization. Its rules fit in one sentence, so it needs no tutorial. It is infinitely generable and gradeable by machine, so supply never runs out and difficulty can be matched to the solver. It rewards pattern recognition rather than knowledge, so a beginner improves visibly within a week. And it has real depth — the techniques go from crosshatching all the way to inference chains, which means it holds people who would have abandoned something shallower.

That combination is rare. It is also, notably, nothing anybody designed on purpose. The puzzle acquired those properties incrementally across a century of accidents: a mathematician's combinatorics, a French newspaper's magic squares, an anonymous American contributor's grid, a Japanese editor's naming and craft standards, and a retired judge's home programming project. If you want to see where the family tree goes next, the variants that grew out of it are covered in killer sudoku and jigsaw sudoku.

Frequently asked questions

Where did sudoku originate?

The modern puzzle originated in the United States, published as "Number Place" in a Dell puzzle magazine in May 1979. It was picked up in Japan by the publisher Nikoli in 1984, renamed sudoku, and returned to the wider world through British newspapers in 2004.

Is sudoku Japanese?

The name is Japanese and Japan is where the puzzle was refined and popularized, but the game itself was first published in America. Calling it a Japanese puzzle is about half right — the naming and the editorial craft are Japanese, the grid is not.

Who invented sudoku?

Howard Garns, a retired Indianapolis architect, is widely credited with creating the 1979 Number Place puzzle, based on his name appearing in exactly those Dell issues that contained it. The attribution has never been officially confirmed, and Garns died in 1989 without seeing the puzzle become famous.

Did Euler invent sudoku?

No. Euler studied and named Latin squares in the 1770s and 1780s, which share sudoku's row and column rules but have no 3×3 box constraint. That box constraint is what makes sudoku a solvable puzzle rather than a combinatorial object, and it isn't in his work.

When did sudoku become popular?

Late 2004 and through 2005. Wayne Gould's generated puzzles reached The Times of London in November 2004, other papers followed within weeks, and the craze spread worldwide during 2005. The first World Sudoku Championship took place in Italy in 2006.

Why is it called sudoku?

It's a contraction of the Japanese phrase Nikoli originally published it under, sūji wa dokushin ni kagiru — roughly "the digits must occur only once." Nikoli's Maki Kaji shortened it to sudoku, from su (number) and doku (single).