In the Meadow

Puzzle craft ·

The Puzzle You Solve Is the Last Draft

Behind a clean crossword, Sudoku or puzzle-hunt clue sits a mess of rewrites, failed tests and ideas that had to be thrown away.

A woman seated at a green desk marking printed pages with a red pen
A finished puzzle may look inevitable. The pages that came before it rarely do. Photograph by cottonbro studio via Pexels. Source. Licensed under Pexels License.

In September 2011, Alex Calhoun and his wife were passing the time on a drive down the California coast with a game of silly rhymes. One clue described a dim-witted love god. The answer was “Stupid Cupid”. It was the sort of game you might forget by the next petrol station.

Instead, they tried to turn it into a puzzle for the MIT Mystery Hunt, a sprawling annual competition whose winning team gets the privilege, and considerable work, of creating the next one. Calhoun’s team had won. Now it was their turn.

Four months later, their little car game was in danger of missing the event altogether. Somewhere along the way, a puzzle that was too easy had become one that almost nobody could solve.

How to make a simple idea much too complicated

The couple’s first proposal had a problem. As Calhoun later recounted in his account of the 2012 Hunt, their editor thought the rhyming clues offered too many possible answers. Solvers needed a clearer route, not just a talent for coming up with funny phrases.

So they added a mechanism. They used eye rhymes, pairs of words whose spelling suggests a rhyme that their pronunciation refuses to deliver. Then came poems and limericks. Then references to films and their remakes, with release years to help order the answers.

The original roadside pastime now required a working knowledge of cinema and an appetite for questionable verse. The editor had asked for a puzzle, not a postgraduate entrance exam in bad poetry.

By December, the verdict had reversed. What started out too simple was now cluttered with false leads. The suggestive poems were entertaining, but their naughtiness wasn’t actually a clue. The film references were too subtle. Calhoun and his wife threw out the poems, rewrote them to describe the films more directly and changed the final answer phrase.

They called the new version Bad Poetry. At least the title had survived the editing process with its sense of humour intact.

The people who don’t know what you mean

The next obstacle was an audience with a useful deficiency: none of the test solvers knew the answer.

They tried Bad Poetry and failed to finish it. The authors revised the rhymes to give the poems a consistent pattern. They removed a movie reference that was too vague. Another had to go after a tester found a second remake. A fact-checker caught a supposed eye rhyme that could be pronounced as a real rhyme, undermining the trick. Even the words that looked reliable were making trouble.

This is the central nuisance of puzzle construction. The author knows which details matter. A solver sees all of them, including the misleading ones the author never meant to put there.

Imagine a grid whose answer is BARK. The maker knows it refers to a dog’s noise, but a solver may be thinking about a tree. That ambiguity can be a delightful feint in a crossword. It becomes irritating when the clue offers no fair way to distinguish the two. As A Crossword Clue Is a Tiny Promise argues, misdirection still has to keep its promise.

The 2021 MIT Mystery Hunt took this kind of testing seriously. Its team produced roughly two hundred puzzles, assigned editors to developing ideas and asked test solvers to record not only whether they reached an answer, but how difficult and enjoyable the journey had been. The team aimed for two successful test solves before final production. One person’s wrong turn isn’t automatically a defect. Several people making the same wrong turn are delivering a message. The awkward part is that the message may be about your favourite clue.

A crossword editor may need to move the furniture

An editor is not necessarily someone who comes along at the end to correct spelling. Sometimes the editor has to rebuild a corner of the puzzle.

Raymond Simon, writing about his work at Games World of Puzzles, describes checking crossword clues for accuracy, style and suitability for readers. In a typical American 15-by-15 crossword, that means looking at roughly 75 clues, as well as the answers they depend on.

Change one word in the grid and several crossings may need attention. Replace an outdated reference and a previously easy clue can become hard. The editor can also spot what the constructor, who has lived with the puzzle for weeks, may have stopped noticing: that the same joke appeared recently, or that a supposedly accessible clue expects an alarming amount of opera knowledge.

A crossword can be technically sound and still leave readers feeling they have been asked to guess what is inside the author’s head. That is why editors sometimes alter the puzzle itself rather than merely polishing its instructions.

The work is especially unglamorous when the thing to remove is a clever idea. Writers become attached to their best tricks. Test solvers are under no obligation to admire them.

Not every difficult minute is a good difficult minute

The Japanese puzzle publisher Nikoli makes a similar argument about Sudoku. In its defence of hand-crafted grids, editor Nobuhiko Kanamoto focuses on the experience of solving: the order in which deductions become visible and the satisfaction of making progress.

That is Nikoli’s editorial philosophy, not a proof that a computer can never make an enjoyable puzzle. But the distinction matters. A Sudoku can have exactly one solution and still offer a tedious route to it. A difficult puzzle can also be wonderfully fair if its clues help you work towards each discovery.

One solver may finish in ten minutes because a particular pattern looks familiar. Another spends half an hour looking at the wrong part of the page. A label such as “medium” cannot explain what happened to either of them. More useful questions are where they became stuck, what they tried next and whether the eventual discovery felt earned.

This is part of why some puzzles feel fair. It is also behind the quiet architecture of a Sudoku. You can count the givens, verify the solution and still miss the quality that matters most to the person holding the pencil.

One last change, days before the deadline

By January 2012, Bad Poetry had survived a remarkable number of drafts. Calhoun and his wife worried it would be cut from the Hunt. The team was working against a fixed launch date, and its internal progress chart was nicknamed the “Graph of Doom”. Even the graph apparently needed to contribute to morale.

One final test came tantalisingly close. The solvers identified the films and looked up the years, exactly as intended. Then they stalled at the last step. They had all the information but couldn’t see what to do with it.

The editors suggested adding a row of blanks to show that the solution required extracting one letter from each rhyme. That small change was enough. Bad Poetry passed testing with only days to spare.

A row of blanks is not a particularly romantic ending to months of work. It was, however, the thing the puzzle needed. The writers had learned to stop asking whether their mechanism was clever and start asking whether someone else could find the way through it.

The finished puzzle reached the solvers. The original car game, the discarded poems and the troublesome movie references did not. Those drafts had done their job.

Sources and further reading

Image credit

Photograph by cottonbro studio via Pexels. Used under the Pexels License. The hosted image is resized and compressed by Pexels; no editorial alterations were made.