Difficulty Curves and Flow State in Game Design

Q: What is the flow state in games?

Flow is Mihaly Csikszentmihalyi's term for deep absorption where time seems to disappear. It happens when challenge roughly matches skill: too much challenge causes anxiety and too little causes boredom. Between those sits a flow channel. Because players get better as they play, difficulty has to rise over time to keep them inside that channel, which is why it should be designed as a curve.

Q: What is a good difficulty curve shape?

A sawtooth is my default: rising peaks of challenge separated by short rests. Each peak tests something the player has learned, and each rest gives them time to consolidate before a slightly higher peak. Flat curves turn dull, straight ramps get exhausting over a long campaign, and cliffs, where one level is far harder than the one before, are the most common quit point I see in play-tests.

Q: What did Jenova Chen's Flow in Games thesis argue?

Jenova Chen's 2006 thesis, the work behind the game flOw, applied Csikszentmihalyi's flow theory to game design. His abstract argues against passive dynamic difficulty adjustment that guesses at skill from incomplete in-game data. Instead it proposes giving players choices, often subtle ones, so they actively shape their own experience. That idea still guides how I think about difficulty options today.

Q: Should I use dynamic difficulty adjustment?

Be careful with hidden adjustment. If players suspect the game quietly eases off when they fail, wins start to feel fake and their sense of competence suffers. Visible options like Celeste's Assist Mode or Hades' God Mode avoid that because the player chooses. I only automate small, local help: fewer hazards right after a checkpoint, longer warnings on a first encounter or slightly longer rests after repeated failures.

Q: How do I author a difficulty curve in Godot or Unity?

Keep the curve in one place as data. In Godot 4, export a Curve resource on an autoload, draw the shape in the inspector and read it with sample(). In Unity, use an AnimationCurve and Evaluate(). Derive speed, spawn interval and other values from that single curve so they cannot drift apart, and measure hazard spacing in seconds so faster play does not cram hazards closer together.

Q: What changed in PawScooty's difficulty design?

Earlier builds had three modes with speed multipliers of 1, 1.25 and 1.5625. A commit on 16 September 2026 replaced them with one balanced difficulty. Difficulty now changes inside a run: speed patches step cruise speed up or down by 20 percent, clamped between 0.8x and 1.8x, and every speed-up is followed by a slow-down within 150 metres. Retention data for this change is not available yet.

Q: How do I tell a difficulty problem from a teaching problem?

Watch fresh players and note every death. If they understand the rule but fail the execution, the challenge is too steep and you should adjust the curve, telegraphs or rests. If they fail because they never learned the rule, making the game easier will not help; teach the mechanic alone in a safe spot first. PawScooty's potholes failed because they were tested before they were taught.

Q: Can AI balance difficulty for me?

AI helps with analysis, not judgement. It can cluster play-test deaths by hazard and level, spot cliffs in your data and write curve tooling. It cannot tell whether a hard section feels like a fair challenge or a cheap shot, because that depends on readability, timing and recent attempts. I would also keep AI out of hidden live difficulty decisions, since players lose trust once they suspect them.

Satyam Kumar

Back to Blog

architecture

Difficulty Is a Shape, Not a Number: Designing Curves That Keep Players in Flow in 2026

By Satyam KumarOctober 26, 202611 min read

difficulty curve flow state games game balancing dynamic difficulty adjustment jenova chen flow in games csikszentmihalyi flow channel sawtooth difficulty level pacing godot 4 curve resource unity animationcurve celeste b-sides hollow knight breath of the wild hades god mode indie game design playtest analysis pawscooty

Difficulty Is a Shape, Not a Number: Designing Curves That Keep Players in Flow in 2026

Frequently Asked Questions

Share this article

Twitter LinkedIn WhatsApp

Satyam Kumar

Founder & AI Architect, AppScale LLP

AI & Cloud Architect. Helping teams build systems that scale to millions.

LinkedIn GitHub

Difficulty Is a Shape, Not a Number: Designing Curves That Keep Players in Flow in 2026

Frequently Asked Questions

Share this article

Comments

Leave a comment