Difficulty Curves Are a Lie: Design the Saw, Not the Slope

Game designers obsess over the difficulty curve — that mythical line that should rise smoothly from tutorial to final boss. But as Dave from Just Things Made By Dave points out, the curve is a lie. What actually works is a sawtooth: a series of small curves, each tied to a new mechanic, that together create the feeling of progress without the frustration of a single relentless slope.

The core insight comes from a tweet by Russell Kay, who worked on Lemmings: instead of one long smoothed curve, break the game into chapters, each with its own difficulty ramp. Each chapter introduces a new mechanic, teaches it, then lets the player master it under pressure. The overall difficulty doesn't monotonically increase — it resets and re-ramps with each new mechanic.

The Difficulty Saw

Dave's model is simple: each mechanic gets its own curve. A new enemy, a new ability, a new puzzle type, a new rule — each one starts easy, gets harder, and then plateaus. The game's overall difficulty is the sum of these sawtooth waves, not a single line. This matches how players actually learn: they struggle with a new concept, then it becomes second nature, and then they're ready for the next one.

This isn't about making the game easier overall — it's about making the difficulty legible. Players want to feel like they're getting better, not just that the game is getting harder. A sawtooth curve gives them that feeling: each new mechanic is a fresh challenge, but the mastery of the previous one carries over.

Keeping Players on the Curve

Dave offers a practical list of dos and don'ts for keeping players in the flow zone:

• Do offer optional collectables, alternate routes, and multiple paths so players can self-select difficulty.
• Do dynamically adjust difficulty based on player performance — the 'magic sauce' that deserves its own talk.
• Do give players multiple ways to learn: text, voice, visual cues, environmental storytelling.
• Don't punish failure by making future attempts harder.
• Don't let players skip areas without learning the skill they need.
• Don't hide the main mechanic of a puzzle — that's just obfuscation.

The key is to give players agency over their own difficulty. Let them choose the hard path, the collectable, the optional boss. Don't force a single gradient on everyone.

Measuring Your Curve

Once you've designed your sawtooth, you need to validate it with data. Dave suggests several methods:

• Death maps (like Mario Maker) — if deaths cluster in one spot, that part is above the curve.
• Level completion stats — a big drop-off between levels signals a difficulty spike or a boredom valley.
• Structured playtesting — ask specific questions, record answers, and look for outliers.
• Streaming and expos — watch real players, but remember the observer effect.

The goal isn't to eliminate deaths or failures — it's to spread them evenly. A level where everyone dies at the same jump is a level with a spike. A level where deaths are scattered is a level that's challenging but fair.

The Takeaway

Difficulty design isn't about making a game that's hard for everyone — it's about making a game that feels rewarding for your target audience. Pick your market, design your sawtooth, and measure the results. The curve is a tool, not a law.