What it is. The first two minutes of play, engineered. You deliver a taste of the core fantasy before you explain anything, teach the first verb by having the player do it, and hand them a real decision fast enough that they lean in instead of drifting off.

Player fantasy / why it’s fun. The pull is “I already get this, and I already like it.” A good hook lets a stranger feel competent and curious inside a minute — the promise of the whole game, compressed and playable, before the game has earned any patience.

The three beats

The hook is not one moment. It’s a sequence, and the order is load-bearing.

Miss Taste and the player never forms a reason to care. Miss Verb and they wait to be told (and leave). Miss Win and they never close the first competence loop, so nothing pulls them into the second.

Levers

• Time-to-fun — how long until the player does the thing the box promised. Measure it in seconds, not screens. Tetris is fun on piece one. Vampire Survivors is fun the moment a weapon auto-fires and a wave dies. Cut everything between boot and that.
• First-verb clarity — the opening should make one verb obvious and safe to try. Not a moveset — a verb. Celeste teaches “jump” on a flat screen with nothing to hit, then “dash” on a screen where dash is the only way forward. The level is the lesson.
• Early win — a payoff the player caused and can read. It must feel earned, not gifted. Peggle’s first board can’t really be lost; the fireworks still land because the player aimed the shot.
• Reversibility — early mistakes should be cheap. A safe sandbox lets the player poke the verb without fear, which is faster teaching than any prompt. See pattern-fairness-and-trust and system-grace.

How to build it

1. Name the fantasy in one line, then ask what single action is that fantasy. Build the first screen around that action and nothing else. Work from process-core-loop and process-pillars — the hook is the loop’s smallest honest slice.
2. Teach the first verb spatially, not textually. Design a room where the verb is the only path forward. The environment prompts; you don’t. This is the system-onboarding craft.
3. Front-load the taste, back-load the depth. Show the fantasy in beat one; reveal systems on a curve after the hook has bought attention. See pattern-mastery-and-flow and system-mastery-curve.
4. Guarantee the first win, then let the second be earnable. Rig beat three so effort pays; make beat four require a little more. That gap is the pull into system-progression.
5. Withhold the menu. Locked slots, greyed abilities, and choice screens before the first win are noise. See pattern-restraint-and-negative-space.

Hook shapes by genre

Different genres promise different fantasies, so their hooks front-load different beats.

Twist seams

Bend the standard hook when the game’s identity lives in the bend.

• A hook, but the taste is a lie you’ll later invert (subverted-promise vector). Open with a confident, competent fantasy — then reframe it once trust is banked. Braid teaches “rescue the princess,” then rewinds the meaning. Powerful; costs a second act to pay off. See world-narrative-delivery.
• A hook, but the first minute withholds the verb to make you find it (discovery vector). No prompt, no obvious action — the fun is deducing what you can do. Outer Wilds and Return of the Obra Dinn hook on curiosity, not competence. Works only when noticing is itself the core loop; see pattern-emergence.
• A hook, but it teaches by killing you (controlled-failure vector). The first death is the lesson and it must read instantly as fair. Spelunky and Dark Souls front-load a legible loss. Demands airtight readability — see pattern-readability and system-telegraphs — or it curdles into antipattern-rng-frustration.

Where it goes wrong

• The hook oversells. You show a fantasy the game never delivers — a scripted, juiced first level that the real game can’t sustain. The player feels the drop when systems arrive. Promise only what beat 30 also keeps.
• The tutorial ate the hook. Explanation displaces play; the player is told the fantasy instead of tasting it. This is antipattern-endless-tutorial — the single most common way a hook dies.
• The first decision is fake. An early “choice” with one real answer teaches that choices don’t matter. See antipattern-fake-choice.
• The menu came first. A loadout/settings/lore wall before the first verb. See antipattern-decision-paralysis.
• The first win was gifted, not earned. Unearned confetti reads as noise; the player learns their input didn’t cause the payoff. Keep pattern-fairness-and-trust intact even in the sandbox.

Overdone when

• Every session re-runs the hook — the tutorial you can’t skip on run two. Gate it once; see system-save-and-checkpoint.
• The scripted opening is more polished than any later hour, so the game peaks at minute one.
• You add a second hook (a splashy meta-progression tease) before the first fantasy has landed. One taste at a time — see antipattern-second-system.

Composes with

• system-onboarding — the mechanics of teaching; the hook is its opening move.
• antipattern-endless-tutorial — the failure mode this pattern exists to prevent.
• pattern-mastery-and-flow — where the player goes after the hook lands.
• pattern-pacing-and-tension — the hook is beat zero of the game’s whole rhythm.
• system-reward-schedules — beat three’s early win is the first entry on the schedule.
• process-the-twist — if your game is “X but Y,” the hook must taste the Y, not just the X.

Prove the first-two-minutes feel with sandboxes/ in isolation before wiring the full game; verification lives in docs/VERIFICATION.md, not here.