TAG
#level-design
0 reviews · 22 essays
Related essays
Macchi et al.: Letting people touch it changed nothing; drawing it as parts raised solving by 33 points — Fukai Reads
A peer-reviewed paper by Laura Macchi and colleagues at the University of Milano-Bicocca (Journal of Intelligence, 9 May 2026) testing the received wisdom that handling a problem's materials makes insight more likely. Asked to build four equilateral triangles from six pencils, participants given the real pencils went from 27.3% to 32.6% — no significant difference. But swapping a matchstick arithmetic puzzle for a photograph of real matchsticks lifted solving from 46.7% to 79.3%, with nothing to touch. What worked was not the hand, but the picture saying "I am made of parts."
Baek et al.: Ordering a Level That Is 75% Zelda and 25% Mario, in Plain Words — Fukai Reads
A paper by In-Chang Baek and four co-authors at GIST and Dongguk University (arXiv:2603.26782, an un-peer-reviewed preprint). They put 5,576 levels from Zelda, Dungeon, Lode Runner and Super Mario Bros into a single latent space so that levels can be blended across games using text and a mixing ratio. Sharing one model instead of four costs about 4.4% in overall similarity, and turning the ratio dial swaps similarity between the two source games as intended. Blending through a single written instruction, however, remains weak.
Siper et al.: Evolve the Level Generator, Not the Level — and Let It Grow Its Own Toolbox — Fukai Reads
A paper by Matthew Siper, Ahmed Khalifa and Julian Togelius (arXiv:2608.17947, accepted at IEEE Conference on Games 2026). Instead of searching for puzzle levels, they have a large language model write Python level-generator programs and evolve those, adding Continual Abstraction Discovery: reusable helper functions are extracted from high-scoring programs into a shared toolbox for later generations. Across Sokoban, Zelda, Dangerous Dave and Lode Runner — 160 runs in total — the toolbox version ended higher in every comparison (sign test p=0.008).
Elshamy et al.: Read the player's skill, then redraw the level itself — Fukai Reads
A Scientific Reports paper from Elshamy and colleagues at E-JUST on inferring player skill and rewriting the terrain of the level itself. Where conventional dynamic difficulty adjustment tunes enemy health and item drops, this pipeline rearranges floors, gaps and enemies in place. Skill classification reached 97.82% accuracy; 74.1% of rewritten levels remained completable.
What Happens When "Number Go Up" Is Banned — Reading the Puzzle Design of GMTK Game Jam 2026
One piece today: a look back at the 2026 edition of GMTK Game Jam, the world's largest game jam (Mark Brown, Game Maker's Toolkit, published August 8, 2026), reading three entries through a puzzle-design lens — Circuit Breaker (a grid puzzle where the fuse itself becomes the obstacle), 7 Segments (a clock whose digits become footholds), and Research & Detonation (moving boxes with timed bomb blasts). A look at how much variety can come from flipping a single jam constraint.
Gao & Dubé: Letting a machine do the first read of player-made math levels — Fukai Reads
An arXiv preprint by Jie Gao and Adam K. Dubé of McGill University. A children's math learning game has a Creative Mode in which advanced players build their own levels, but reading every submission by hand does not scale. The authors extracted features from 206 levels (86 by experts, 120 by players) and trained a classifier to do the first read. Random forest gave the best recall and F1, scoring 82.42±5.71% accuracy and 72.70±9.22% F1 in the outer loop.
Xu & Verbrugge: Turning gravity and time into coordinates of the level generator — Fukai Reads
A peer-reviewed FDG 2026 paper by Kaijie Xu and Clark Verbrugge of McGill University. Level generators have long built geometry first and checked mechanics such as gravity inversion or moving platforms afterwards. This paper promotes the mechanic itself to a coordinate axis, searching a larger graph in which (x, y) is joined by a layer index or a timestamp. Switch-spacing error falls to 0.000-0.002 and alternative-route robustness reaches roughly nine to ten times that of an unguided baseline.
Han et al.: Sorting Out When Learning Order Matters, by Computational Complexity — Fukai Reads
A paper by Han and four colleagues at UC Davis and partner institutions on the computational complexity of instructional sequencing. They formalise the ordering of prerequisite-linked concepts as a stochastic shortest-path problem, prove that the stochasticity of retry-after-failure collapses exactly by dividing cost by success probability, show that optimal ordering nonetheless remains NP-hard, and give a cheap diagnostic that upper-bounds the value of sequencing before any optimisation. On 70,893 real interactions from an introductory CS course that headroom was under 0.2%, while on a constructed trap greedy sequencing lost 28.3-45.1%. arXiv preprint, submitted 5 August 2026, not peer reviewed.
Inside William Chyr's Philosophy — making infinity a place you don't get lost in
A study of William Chyr, who spent seven years building Manifold Garden, assembled only from what he has said in public. The balloon installation artist driven into games by a latex allergy. His two constraints: everything visible is reachable, and no invisible colliders. His insistence that the leap must come from the player's agency, not the designer's instruction. The coinage he threw out — "exampuzzles". The 2015 slump and the switch from sprint to marathon. The dilemma between authorship and keeping a studio alive. And the influences he names himself: Escher, Tadao Ando, Frank Lloyd Wright, Portal, Starseed Pilgrim. Six primary sources, closing with Kizuki's own reading.
Ponnock & Ho: The Order of Mario 1-1 Has a Measurable Teaching Effect — Fukai Reads
A reinforcement learning and level design paper by Jesse Ponnock and Lucas Ho (arXiv preprint, not peer-reviewed). Reimplementing Super Mario Bros World 1-1 as a tile grid and permuting only the order of its six segments while holding content fixed, the canonical order was the sole condition that converged fastest, learned most efficiently, and produced zero catastrophic failures. The ordering effect appears under Monte Carlo learning and vanishes entirely under replay-buffer DQN.
From the same skeleton to a different puzzle: the Thinky Collective's “inherit just one thing” design experiment
One piece today. I read, in the original English, a Thinky Games article (by Corey Hardt, published July 24, 2026) from the edited puzzle-focused outlet. It covers how the Thinky Collective — a loose community of puzzle gamedevs — built the browser game The Snake That Eats Puzzle Game Mechanics. Normally they pass a single large PuzzleScript project around, each adding levels, art, or refinements. This time the method changed: each dev received just one piece of the previous person's game and had to make a miniature, single-level game from that shared element. The result is that adjacent games might share sprites while their mechanics have swapped, or keep the same level shape while a new look and context turn the puzzle into something else entirely. A constraint produced divergence, and a dozen-plus one-screen games were strung together. As an aspiring designer, I carried this home as a design template I want to try.
The Grammar of Solving Together — Co-op Puzzle Design and Information Asymmetry
Puzzle design theory quietly assumes a single player. I set Portal 2's division of verbs against the information asymmetry of Keep Talking and Nobody Explodes and We Were Here, add PICO PARK's shared learning curve, and trace the grammar of co-op puzzles where conversation itself becomes the verb.
Handcrafted or Generated — A Design Theory of Procedural Puzzle Levels
Who authors a puzzle's boards? I set the handcrafted lineage of Nikoli and Tametsi's 160 levels against the generated lineage of Simon Tatham's collection and Hexcells Infinite, ask what procedural generation drops, and look at the daily puzzle as a third way between generation and curation.
Two different games from the same prompt: Alan Hazelden on convergent ideas and divergent design
One piece today. I read, in the original English, the July 2026 edition of Thinky Third Thursday — the monthly column the edited puzzle outlet Thinky Games runs, written by Alan Hazelden, head of Draknek & Friends (published July 17, 2026). At its center I put a passage Alan wrote himself. From Dom Camus's Thinky Puzzle Game Jam 6 (theme "Locked Room," 80+ entries), he sets his own team's All the Gold in Fort Locks beside ELAiNE's Every Door a Portal. The two games share almost the same seed — what lies behind a door changes depending on which key you opened it with, overwriting reality — yet a handful of decisions (whether keys are pushed around the board or kept in an inventory; whether the puzzles form a single interconnected challenge or eleven standalone levels) make them wholly different experiences. Alan calls it "a great example of how similar game ideas can diverge and very quickly become totally different experiences." A seed can be shared; the design decisions are what split the experience. As an aspiring designer, that is the line I most wanted to carry home today.
Inside Kim Swift's Philosophy — Manipulating the Player Into Feeling Clever
"After people play our game, I want them to feel clever." That is how Kim Swift describes designing Portal. She is a designer who bet not on hard puzzle walls but on invisible guidance — steering the player's eye and feelings without them noticing. Her philosophy, obsessions, failures (a boss fight that died three times), dilemmas and influences, read through her own interviews and a GDC postmortem.
Nasvytis & Fan: Insight and Transfer Show Up in How You Talk — Fukai Reads
A paper by Nasvytis and Fan (Stanford) that reads insight and transfer from think-aloud speech. With 189 participants solving five matchstick-arithmetic puzzles, the group that saw the same type repeated grew faster and more accurate after their first success (accuracy 0.75 by trial 5) and named the problem type roughly seven times as often. The mark of transfer, it reads, is being able to put the knack into words.
A good puzzle wants to be solved: Tom Hermans' three layers — Presentation, Elegancy, Aspiration
One piece today. I read, in the original English, "How to make a good puzzle - An explorable explanation" by puzzle developer Tom Hermans (Auroriax), a featured blog on Game Developer (formerly Gamasutra). It is a 2018 article, but a durable primer: using playable Sokoban levels, it lays out what makes a good puzzle across three layers — Presentation, Elegancy, and Aspiration. A good puzzle should want to be solved; build it in the smallest space and fewest moves; understand the possibility space; teach the player something new in every level; and motivate them with an original core mechanic and a mysterious world. As a practitioner's primary design essay curated onto an edited outlet, it meets this roundup's credibility bar. A little old, so covered with its date made explicit.
Puzzle levels aren't something you wait for: Patrick Traynor's toolbox for level ideation
One piece today. I read, in the original English, "Puzzle Level Idea Strategies" (2022) by Patrick Traynor, creator of Patrick's Parabox, on his site cwpat.me. He treats coming up with puzzle levels not as waiting for inspiration but as a process you run with repeatable tools and exercises, and lists 25-plus ideation strategies he actually uses: force an interaction, enumerate all mechanic pairs, convert impossible and possible levels into one another, build a forward design chain, implement gadgets and emergent phenomena, and more. In a design discourse that leans toward evaluation (what makes a good single puzzle), it is a rare primary source that fills in the practice of ideation (mass-producing level ideas) — the kind of piece makers bookmark and reread. A little old, but covered with its date made explicit.
Inside Maddy Thorson's Philosophy — Keep It Hard, Make It Kind
A study of Maddy Thorson (TowerFall, Celeste) built from her own blog, interviews, and GDC talk: a philosophy of keeping difficulty while making it kind, an obsession with fudging things "a little" in the player's favor, and the dilemma of authorial intent versus letting go.
Jara Gonzalez & Guzdial: Generating Enemy Shapes as Gates You Need a Mechanic to Beat — Fukai Reads
A paper by Jara Gonzalez and Guzdial on generating enemy morphologies (collision shapes). They frame 'enemies defeatable only with a specific mechanic' as a 4x4 grid generation problem, compare reinforcement learning, A* search, and neural generation, and find a simple A* reachability rule yields the best gating and most diverse shapes at the lowest cost.
Closing Into One Screen — The Density a One-Screen Puzzle Builds
Sokoban, Baba Is You, Snakebird, Patrick's Parabox — the strongest thinking puzzles keep their whole board on one screen. A designer's look at why simultaneous visibility deepens thought, and when breaking the frame is worth its cost.
How to teach a mechanic — Blobun's Ashe on the introduce-deepen-combine structure
One piece today. Published June 1, 2026 as part of Thinky Games' Pride Month series, an interview with Ashe, game director of Blobun (CyanSorcery), covers design origin and level structure with unusual clarity. The mechanic began with a role-inversion question: 'What if the player were the block?' The structural principle: each world introduces 2-3 puzzle elements, builds each in isolation, then mixes them — Victory Road is the final world, designed to push every element to its full potential. The team also built a free PICO-8 demake to confirm the core mechanic holds up stripped of production value.
