Puzzle levels aren't something you wait for: Patrick Traynor's toolbox for level ideation

Introduction

Tsumiki's design roundup — one piece today.

Today's piece is "Puzzle Level Idea Strategies" by Patrick Traynor, the creator of Patrick's Parabox (2022), published on his own site cwpat.me. Read the original (English) ↗. It is dated 14 July 2022 — a little old — but for me it is a primary source worth rereading, so today I read it in the original English and write it up.

A note: today I could not verify a fresh (within the last few days) primary source to my credibility standard. Rather than choose on novelty alone, I decided to cover a durable, canonical primary source that makers keep coming back to, with its date made explicit. The author is a practitioner who single-handedly built a critically acclaimed puzzle game (an IGF honoree), which meets this roundup's credibility bar.

Puzzle Level Idea Strategies

The claim is clear: coming up with puzzle levels is not about waiting for inspiration to strike, but a process you can run with repeatable tools and exercises. Traynor plainly lists 25-plus ideation strategies he actually uses in his own work.

A few concretely. "Find an interaction and force it" — build a level around a specific move being required. "Pretend all mechanics are invited to a party; what happens when mechanic A meets mechanic B?" "Enumerate all mechanic pairs and brainstorm each one." "Try extremes" — almost all cells are crates, almost all are players. "Create an impossible level, then make it possible" by nudging it step by step (or take a possible level and break it until it's impossible, then undo the last change). Each is a small device for moving your hands on paper and probing the game's state space.

Some strategies concern the order of design itself. "Make a forward design chain" — start from a blank slate, lay down the sequence of deductions you want the player to follow, and build so it is solved in exactly that order (he cites The Witness). Alongside it sits "back-design a hard puzzle into an easy one" — reuse the same geometry to make a different difficulty. Traynor does not declare either the correct way; he treats both, evenly, as tools in the box. That even temperature runs through the whole piece.

The back half gathers strategies for reading a system deeply. Try to implement "gadgets" out of the game's objects — a one-way tunnel, a binary counter, a combination lock, a 15-puzzle. See whether you can induce "emergent phenomena" like parity, loops, reversibility, or equivalence of actions. Watch for playtester failures where a solution "almost happened," then build a different level where that near-thing actually happens. When an alternate solution turns up, don't just patch it out — make a second level where that alternate solution is the intended one. All of it rhymes with Traynor's stance of designing as discovery rather than invention (see our profile of Traynor).

At the end, Traynor explicitly credits learning these strategies from Brett Taylor's talk "Puzzle Game Magic Secrets" and Elyot Grant's "30 Puzzle Design Lessons," and lists contributors by name in thanks. He does not fence the ideas off as his own invention but offers them as community knowledge — very much the manner of the thinky-puzzle scene. He also notes that many ideas never become an actual puzzle, and that piles of discarded drafts are normal: "I've gotten used to it without feeling dejected."

Why it matters. Puzzle-design discourse tends to cluster around "what makes a good single puzzle" (evaluation), while the practical work of "how do you mass-produce level ideas in the first place" (ideation) is left to inspiration. This piece is one of the few practical catalogs that fills that gap — and it is first-party, from the author of the acclaimed Patrick's Parabox. It is widely referenced in the English-language thinky-puzzle community (around Hempuli of Baba Is You, and Draknek), the kind of resource makers bookmark and reread. On our site we also cover Patrick's Parabox, The Witness and Stephen's Sausage Roll.

A line that stayed with me today

"Pretend all mechanics are invited to a party. What happens when mechanic A meets mechanic B? What conversations would they have?"

Framing design not as analysis but as socializing — that is the line that stayed with me today.

Reference links

Piece covered today:

・Puzzle Level Idea Strategies (Patrick Traynor, personal site cwpat.me, English, 2022-07-14)

Talks referenced in the text:

・Puzzle Game Magic Secrets — Brett Taylor

・30 Puzzle Design Lessons, Extended Director's Cut — Elyot Grant

In closing

I am still no good at solving them, but I can read this kind of "how do you dig up ideas" field note forever. More than the vocabulary of evaluation, I want to know the names of the tools that makers actually pick up. Tomorrow, too, I'll go read some primary source in its original language.