Thirty-Three People and a Dog Just Out-Designed AAA
July 11, 2026 · Casey Brooks, A Theory of Fun for Game Design~8 min read
Here's the tell. Sit a stranger down with Clair Obscur: Expedition 33 and watch their thumbs during a boss fight — not the cutscene, the fight. They're leaning in. Every enemy swing, they're timing a dodge or a parry, reading the wind-up, getting punished when they read it wrong. It's a turn-based RPG, the genre where you're supposed to sip your coffee between moves, and instead they've forgotten the coffee. That's a roughly 33-person studio doing that to a AAA veteran's hands. And this March, that same little game walked into the 2026 Game Developers Choice Awards with 8 nominations and walked out with 5, including Game of the Year — beating teams with a thousand times the headcount.
30-second read
Clair Obscur: Expedition 33 (Sandfall Interactive) led the 2026 Game Developers Choice Awards with 8 nominations and won 5 — Game of the Year, Best Debut, Best Visual Art, Best Audio, Best Narrative.
The team: about 33 in-house developers. "33 team members and a dog" — Hideo Kojima called that size ideal. They beat studios with thousands.
The design trick is one idea: bolt real-time dodge/parry/aim timing onto turn-based combat, so every fight still demands skill.
Raph Koster's frame explains why it works: fun is the brain learning a pattern. Add a new timing variation every fight and the learning never flatlines — that's sustained fun in the exact, technical sense.
1The upset
A studio you could fit in a bus just swept the awards the giants show up to win
The scoreboard isn't close, and that's the whole point.
Let me lay the facts flat first, because they do the arguing for me. Clair Obscur: Expedition 33, made by Sandfall Interactive and published by Kepler Interactive, went into the 2026 Game Developers Choice Awards — the one voted by people who make games, not by fans or critics — with 8 nominations. It won 5: Game of the Year, Best Debut, Best Visual Art, Best Audio, and Best Narrative. Best Debut on top of Game of the Year, in the same night, from a first-time team. That doesn't happen by accident and it doesn't happen on marketing.
Now the headcount. Somewhere around 33 in-house developers built it. The line going around is "33 team members and a dog," and Hideo Kojima — a man who has shipped some of the most expensive games ever made — said that team size is basically ideal. External partners handled QA, marketing, and localization, the way a small studio rents the machinery it can't own. But the game itself, the thing you actually play, came out of a room you could photograph in one shot.
The usual AAA answer
Thousands of staff. Hundreds of millions. Scope as the strategy — more map, more systems, more hours — and quality assumed to follow the budget.
Sandfall's answer
~33 people. One combat idea executed cleanly. Depth as the strategy — the same fight staying interesting on the fortieth hour, not the map getting bigger.
I've sat on the other side of that budget. I've watched a hundred-person team spend a year making a thing wider when what it needed was to be deeper. So when a game this small out-designs the giants at the thing giants are supposed to be best at — combat feel, minute to minute — that's not luck. That's a team that knew exactly which one thing to get right, and spent all thirty-three of their people on it.
2The one idea
They took the one boring thing about turn-based combat and made it demand your hands
Every design has a load-bearing idea. This one you can say in a sentence.
Here's the mechanism, plain. Turn-based combat's oldest weakness is that once you've solved the math, execution is free — you pick the optimal move and press a button, and the same button works the tenth time and the hundredth. The challenge is all up front, in figuring it out; after that you're doing data entry. Clair Obscur keeps the turn structure — you still choose actions, still build a party, still read a damage sheet — but bolts a live layer on top: when the enemy attacks, you're not watching, you're dodging, parrying, and aiming real-time inputs, frame by frame, during their turn. Read the wind-up right and you take nothing. Read it wrong and you eat it.
▶
Think of it like: a fighting game and a turn-based RPG shaking hands. The RPG half is the chess — your build, your resource math, the plan. The fighting-game half is the reflex — the parry window, the tell you learn to read. Old turn-based games gave you the chess and then let your hands go to sleep. This one keeps waking them back up, every single turn.
That's the whole trick. Not a bigger world. Not more systems bolted on until the box looks heavy. One structural idea — the fight never stops asking for skill — executed cleanly enough that a thirty-three-person team could actually finish it and polish it. Scope would have buried them. A single sharp idea, sharpened all the way, is a thing a small team can win with.
3Why it stays fun
Fun is the brain learning a pattern — and this game refuses to stop teaching
There's a twenty-year-old idea that explains the whole thing, and once you see it you can't unsee it.
Raph Koster wrote it down in A Theory of Fun for Game Design, and it's the most useful sentence I know for this job: fun is the pleasure your brain takes in learning and mastering a pattern. Not in winning. In learning. That little hit of satisfaction when the shape of a problem clicks into place — that's the whole engine. Which leads straight to the bad news for designers: the moment a player has fully grokked your pattern, when there's nothing left to figure out, the fun drains out. Not because the game got harder or worse. Because there's nothing left to learn. A solved pattern is a boring one. Good design, in Koster's frame, is a machine for feeding the brain new variations so the learning curve never goes flat.
A learning-curve chart. Fun, in Raph Koster's sense, is the brain's pleasure in learning a pattern — so both a static turn-based game and Clair Obscur's reactive combat climb fast at first. But once a static system's pattern is fully grokked, the curve flatlines and boredom sets in; Clair Obscur bolts active dodge/parry timing onto every turn, so each new fight adds a variation to learn and the curve keeps stair-stepping up. Corner: Sandfall Interactive — about 33 developers plus a dog — led the 2026 Game Developers Choice Awards with 8 nominations and won 5 — including Game of the Year. Framing: Raph Koster, A Theory of Fun for Game Design. Source: gdconf.com (GDCA 2026). Illustrative curve, not measured data.
Now watch what the reactive layer does to that curve. In a static turn-based game, you crack the combat math a few hours in, and from then on you're executing a solved pattern — the learning's done, and Koster says the fun goes with it, no matter how much map is left. But when every enemy attack demands a live read, every fight hands you a fresh pattern to learn: this boss's tells, that one's rhythm, a new wind-up you haven't seen. The turn-based layer keeps the strategy deep; the reactive layer keeps the learning fresh. So the curve never flatlines. That's not a vibe. That's Koster's mechanism, running exactly as written — and it's why a small game can stay fun past the point where a bloated one has gone quiet.
4What this means
What this means for you — whether you play them or make them
One question does the work on both sides of the couch.
If you make games: stop budgeting for breadth and start budgeting for the learning curve. Before you add another zone or another system, ask the Koster question about the thing you already have — is the player still learning something here, or have they solved it and now they're just clocking in? A solved system doesn't get saved by making it bigger; it gets saved by giving the brain a new variation to chew. Sandfall didn't out-spend anyone. They found one system that keeps teaching and spent everything on it. That's a move a two-person team can make. Scope is the thing that eats small teams alive; a single learnable idea, sharpened all the way, is the thing that lets them win.
If you play games: trust your thumbs over the trailer. The tell for a game that'll last isn't the size of the world or the length of the feature list — it's whether you're still learning forty hours in, or just going through motions you mastered in hour three. Watch for the moment a game hands you a new thing to read instead of a bigger room to walk across. That's the difference between a game that stays alive in your hands and one that goes quiet the second you've solved it. Thirty-three people and a dog just proved the whole industry can be beaten on that one axis. The learning is the game. It always was.
Clair Obscur: Expedition 33 (Sandfall Interactive / Kepler Interactive) led the 2026 Game Developers Choice Awards with 8 nominations and won 5 — Game of the Year, Best Debut, Best Visual Art, Best Audio, and Best Narrative — per gdconf.com. Team size of roughly 33 in-house developers ("33 team members and a dog") and Hideo Kojima's remark that the size is ideal, per VGC; external partners handled QA, marketing, and localization. Framing: Raph Koster, A Theory of Fun for Game Design — fun is the brain's pleasure in learning a pattern, and a game grows boring once the pattern is fully grokked; good design keeps feeding new variations. Honest limits: the learning-curve figure is illustrative, not a measurement, and "about 33" is the widely-reported core team, not a precise payroll count.
现在看那层反应式操作对这条曲线做了什么。在静态回合制里，你几小时就把战斗数值算破了，从那之后你就是在执行一套已解开的模式——学习结束了，科斯特说好玩也跟着走了，不管地图还剩多少。可当敌人每一次出手都要你临场读一遍，每一场仗都塞给你一套新模式去学：这个 boss 的起手动作、那个的节奏、一个你还没见过的挥砍。回合那一层把策略留得深；反应那一层把学习保持新鲜。于是曲线永远不走平。这不是玄学。这是科斯特的机制，一字不差地跑着——这就是为什么一个小游戏，能在一个臃肿大作早已安静下来的那个点之后，还好玩。
まず事実を平らに並べます。事実が代わりに語ってくれるので。『Clair Obscur: Expedition 33』、Sandfall Interactive 開発、Kepler Interactive 発売。8 ノミネートで 2026 Game Developers Choice Awards に入りました——ファンでも批評家でもなく、ゲームを作る人が投票する賞です。受賞は 5 部門：Game of the Year、最優秀デビュー、最優秀ビジュアルアート、最優秀オーディオ、最優秀ナラティブ。Game of the Year に加えて最優秀デビューまで、同じ夜に、初めて作ったチームから。これは偶然では起きないし、マーケティングでも起きません。