Back to Blog

What makes a game fun

ยท 3 min read

  • Fun is the brain's intrinsic reward for learning: making progress on prediction (ie. reducing uncertainty).

  • Loops are how that learning happens.

      1. perceive 2) hypothesize 3) act 4) observe 5) update
    • This loop repeats over new situations, until the system becomes predictable.

  • Feedback is what closes the loop.

    • Players need to know: what they can do, that they did it, how state changed, and whether it helped relative to the goal.
    • Feedback operates on multiple timescales:
      • Dense, moment-to-moment feedback guides immediate action (eg. hit sparks).
      • Medium-term rewards shape tactics and builds (eg. unlocking a new weapon).
      • Sparse, long-term rewards keep you exploring and prevent local optima (eg. story beats, rank climb).

  • Uncertainty is what fuels the loop.

    • Good games stay uncertain for a while, then become learnable.
    • If a problem has only one answer, it's closer to a puzzle. Once you solve it, it gets boring.

  • Variation and escalation maintain that uncertainty.

    • A game that's too easy offers nothing to master; a game that's too hard offers no sense of progress.
    • The best games keep you near the edge of mastery, with breathers and peaks.
    • Difficulty sliders and multiple paths can let different players climb that curve at different speeds.

  • Loops can be nested and composed.

    • Small loops chain into value chains (one loop outputs tools for the next) and web into economies.
    • Many "currencies" aren't money: HP, ammo, cooldowns, attention, time, etc.

  • Most feedback mechanics already exist; novelty often comes from recombination and reskinning.

    • Many problems are isomorphic under different dressings, but aesthetics change how the exact same problem feels.
    • eg. dressings, art, audio, story, setting, UI, and metaphors
    • A useful workbench is a catalog of problem types you can combine.

  • Different modalities produce different kinds of fun.

    • Video games are often partially observable and non-Markovian:
      • you can't solve the game from the current screen alone
      • you must track hidden variables like cooldowns, off-screen threats, momentum, and timers.
    • Board games are often closer to perfect information where the full state is visible:
      • the fun comes from reasoning about the enormous space of what could happen next.
    • Both reduce prediction error, but via different mechanisms: discovery of hidden state vs. mastery of visible complexity.

References