Low-inspiration puzzlehunt puzzle ideas

30 Dec, 2024

(This post features mild spoilers for the first steps of various puzzles from puzzlehunts I've helped write, by which I mean I helped write the hunts, not all the individual puzzles)

Sometimes as a puzzlehunt writer, you just need another feeder puzzle or two to round out a meta that takes some number of answers, but your puzzle-ideas.txt has run out. What to do?

Here are a few places to start brainstorming that I've picked up over the years, roughly increasing by how much inspiration you'll still need to dredge up:

Clues about slightly modified things. Take a set of things (e.g. songs, musicals, video games). Apply some simple wordplay transformation to them (e.g. add/change/delete a letter). Clue the transformed things (e.g. contrived crossword-style clues, MS Paint drawings, text that's related to the object in some more diegetic/thematic way like a review if it's a movie). The space of possibilities is enormous.

IMO these are as free as uninspired puzzles can get, and honestly I often think they're kind of boring, but I think many other solvers do enjoy them. And, if you can take the opportunity to either introduce the solver to a funny dataset, or simply write funny clues, that can considerably elevate the puzzle.

A prototypical example (albeit probably overclued compared to what I'd include in other hunts) is DP Puzzle Hunt's Smash Biros.

Generalized dropquotes. It's widely known that "random anagrams" are a puzzle no-no; the space of possible anagrams for a given string is too large. On the other hand, dropquotes, where you "anagram" columns of letters, are usually fine (as long as the columns aren't too tall), because the space of anagrams of columns has more structure to it. Dropquotes themselves I also sometimes find boring, but I think there are many plausible variations of "restricted anagrams" that are underexplored.

One trick I've used more than once is the "dual dropquote" or "shredder": Similar to a dropquote, you write some text into a grid and then cut the grid into columns. However, instead of anagramming within each column, you preserve the order of letters within each column and shuffle (or sort) the columns. The solver has to reorder the columns.

Other puzzles I've written out of bespoke rearrangement methods include the custom Topsy Turvy, from 2023, and 2024 Galactic Puzzle Hunt's Exchange.

For the math nerds, I would succinctly describe this as: Take some text, choose a small subgroup of the permutation group of the text's letters, and shuffle the text with a permutation from that subgroup.

More people should try these!

Generalized crosswords. Why do crosswords "work"? Why are they fun to solve? One reason is that crossword clues are often not solvable by themselves, at least not unambiguously, but become solvable when you have crossing entries, allowing you to iteratively home in on the overall solution. In standard American crosswords, every square is "double-checked", and I think this is worth preserving for most squares to allow clues to be more interesting. But if you're looking for a wacky puzzlehunt puzzle idea, there's no reason the answers (or more precisely the sequences of blanks that answers are written letter-by-letter into, sometimes called "lights" in crossword lingo) have to go neatly across or down on a rectangular grid. You can make the lights go in opposite directions or slither or loop, you can write multiple letters or things that aren't letters or nothing in cells, you can make some lights larger or differently-shaped, you can break apart the clues or partially disassociate them from the lights in various ways. Some generalized crosswords achieve escape velocity and become genres in their own right, e.g. Rows Gardens. Others are completely bespoke and might just arrange their lights in some thematic way, e.g. 2018 GPH's An Eccentric Crossword or 2021 Mystery Hunt's Fret Not!.

The above descriptions sort of implicitly assume that blanks have unique, well-defined locations in 2D space and lights are vaguely coherent patterns of blanks in that same 2D space, but even that doesn't have to be true; really, I want you to think of blanks as an arbitrary set $B$ and lights as arbitrary finite sequences¹ of elements of $B$ — its Kleene star, if that's helpful.² The double-checking convention just says that every blank appears in exactly two lights (or even more?). Acrostics, then, are generalized crosswords where there's one big light for the quote that contains every blank, and every other light just double-checks and is double-checked by the big one. The "letter webs" in Jack Lance's R.T.3 SEARCH represent generalized crosswords by assigning a 2D position to each blank-light incidence, arranging the incidences of the same light in a line, and connecting all the incidences of the same blank with line segments. There's definitely more room for innovation. And, I have not mathematically formalized any generalization of how clues can be broken apart or dissociated from lights.

Identifying things from restricted aspects/communication. Pick a data set and depict or clue things from it after reducing/censoring/"down-projecting" it in a strange way, or with strange restrictions on what you're allowed to say. Coming up with a strange reduction method or set of restrictions does still require inspiration/creativity, but I think it's a decent source of ideas.

Sometimes the restriction can be purely mechanical or deterministic, and it's the puzzle author's job to assemble the puzzle from things that can be identified under this restriction. For example, DPPH's Bottom Lines only shows you the bottom of each letter; the author can make some global decisions (what the cluephrase is, how much of the bottom of each letter to show, what font to use), but the puzzle is fully determined by those choices.

Other times, the restriction leaves you with a lot of freedom. They may feel like a "restricted communication" board game, where the relationship between puzzle writer and solver feels more collaborative than adversarial. Think of playing Pictionary with your friends. For example, 2018 GPH's Emotion Pictures depicts plots of various movies with only emoji. The authors had freedom to pick movies they thought would be easy to clue (and that worked with the puzzle extraction), but after picking the movie, they also had a lot of freedom to choose specific emoji to clue it.

Typically, of course, things will be kind of in-between. For example, in 2019 GPH's Cross Lines, you're asked to identify country flags from a thin stripe. In this case the puzzle author had freedom to choose which countries to use and which stripe to take from the country's flag, but not total freedom to use any colors/shapes they liked to clue any country. Or, 2021 Mystery Hunt's Back School To reduces the things being clued to their enumerations plus one emoji of the author's choice. Reducing things to their enumerations is a fully deterministic process the author has no control over; the choice of emoji was up to the author, but they were limited to one emoji, and there don't exist arbitrary emoji for arbitrary concepts.

Identifying things described from a strange perspective. The line between this and the previous is blurry, and coming up with a "strange perspective" can require a lot of inspiration (which is why it's last on this list), but I think this is still a fruitful direction to brainstorm.

A popular subgenre is unnaturally depicting things through 2D graphs/diagrams. Examples might include DPPH's Now I Know..., Silph Puzzle Hunt's A Lot of Research... , and 2021 MH's Enclosures.

In general, though, you can try to write a puzzle of this kind by staring at a standard, unremarkable dataset and asking yourself, what if I pretended this were something completely different? This is how I wrote 2021 MH's Countries.