Your playtester quits 38 minutes in. The combat works, the controls feel good, nobody hit a bug. When you ask why they stopped, you get the answer every designer dreads: "I don't know, I just kind of drifted off."
That is almost never a mechanics problem. It is a pacing problem, and pacing is the one thing most small teams never write down. Level layouts get blockouts, enemies get stat tables, economies get spreadsheets — but the emotional shape of the hour, the rise and fall of tension, usually lives in somebody's head as a vibe. Vibes do not survive scope cuts, new hires, or six months of production.
Here is a way to make pacing an artifact you can review, argue about, and test.
Score your beats on a five-point scale
A pacing pass starts by breaking your content into beats — the smallest chunk a player would describe as "a thing that happened." A fight, a traversal section, a conversation, a puzzle, a shop visit, a cutscene. Anywhere from 30 seconds to 4 minutes each.
Then give every beat an intensity score. Five points is enough; ten invites bikeshedding.
| Score | Player state | Examples |
|---|---|---|
| 1 | Idle, safe, no failure possible | Hub walking, inventory management, shop, ambient exploration |
| 2 | Low engagement, no threat | Light platforming, dialogue with a choice, a gentle puzzle |
| 3 | Active but comfortable | A standard fight against known enemies, timed traversal |
| 4 | Pressured, failure likely | New enemy combo, resource-starved fight, a chase |
| 5 | Peak — full attention, failure expensive | Boss, scripted set piece, one-life-left escape |
Now lay the beats out in order and read the numbers as a line. A mid-game level in a 3D action game might look like this:
| # | Beat | Length | Intensity | Note |
|---|---|---|---|---|
| 1 | Arrive at gatehouse, look at vista | 1:00 | 1 | Establish goal visually |
| 2 | Two patrolling grunts | 1:30 | 3 | Re-teach the core loop |
| 3 | Locked door, find the lever | 2:00 | 2 | Rewards exploration |
| 4 | Ambush: grunts + first shield enemy | 2:30 | 4 | New verb introduced under pressure |
| 5 | Corridor, audio log, ammo cache | 1:00 | 1 | Rest + restock |
| 6 | Arena: shields + archers | 3:00 | 4 | Combination test |
| 7 | Collapsing bridge chase | 1:30 | 5 | Spike |
| 8 | Camp: save, upgrade, NPC scene | 2:30 | 1 | Reset before next region |
Read down the intensity column: 1, 3, 2, 4, 1, 4, 5, 1. That is a healthy shape — it climbs, it breathes, it peaks, it lands. Compare it to 3, 3, 4, 3, 4, 3, 4, 3, which is the numeric signature of the level your tester drifted out of. Nothing is wrong with any individual beat. The problem is only visible in the sequence.
This is the whole trick: pacing bugs are invisible at the beat level and obvious at the sequence level. You have to look at the list.
Three rhythms that actually work
Not every game wants the same curve. Three patterns cover most of what small teams build.
The sawtooth
Tension climbs across three or four beats, peaks, then drops hard to a 1. Repeat. This is the classic action-game rhythm — Half-Life 2 and Resident Evil 4 both ride it, and it is the safest default for a linear game because the reset beat gives players a natural place to stop playing without quitting, which sounds bad but is exactly what makes them come back tomorrow.
Target: a peak every 8–12 minutes, and never more than two consecutive beats at 4+.
The wave
Slower oscillation, wider amplitude, used in exploration and survival games. Long 2s punctuated by sudden 5s. The tension comes from not knowing where the spike is. This one is fragile: if your spikes become predictable — every third room, always after a save point — the wave flattens into a sawtooth the player can schedule around.
The escalating run
Roguelikes and arcade games: intensity rises monotonically across a run and only resets at death or victory. Hades does this, and so does every good wave-defense mode. Your rest beats have to be inside the climb — a shop, a boon choice, a corridor between rooms — because you never get to drop to 1 until the run ends. Budget 20–40 seconds of decision-making between each fight, and protect that time; it is the only breathing room you have.
Pick one deliberately and write it at the top of your pacing document. "This level is a sawtooth with a 10-minute period" is a testable statement. "This level should feel exciting" is not.
Rest beats are content, not gaps
The most common mistake in indie pacing is treating 1s and 2s as filler — corridors to be shortened, dialogue to be skippable, shops to be menus. Then the difficulty tuning pass makes everything harder, the rest beats shrink, and the curve flattens at the top.
Rest beats do three jobs, and if you cut them you lose all three:
- Recovery. Attention is a resource. Sustained 4s stop reading as intense after about ten minutes; they read as exhausting, then as boring.
- Contrast. A 5 only feels like a 5 relative to what preceded it. A boss after fifteen minutes of grinding is just more grinding.
- Consolidation. Players form plans in quiet moments. If they never get one, they never build a strategy, and combat degrades into button-mashing.
Protect them explicitly. In the beat list above, beats 3, 5 and 8 are load-bearing. Mark them that way so the next scope cut takes something else.
Where pacing quietly breaks
A short field guide to symptoms you will actually see in playtests:
- The flat middle. Everything at 3 for twenty minutes. Usually caused by content built by different people in parallel with no shared curve. Fix: insert one 5 and two 1s; you rarely need new content, just reordering.
- The premature peak. Your best set piece lands 25 minutes in and nothing tops it. Fix: move it later, or raise the stakes of the ending, not the spectacle of the opening.
- Difficulty spikes read as pacing spikes. A beat scored 4 that is actually a 5 because players fail it six times. Intensity is felt, not authored — so verify it.
- The tutorial hangover. A long 1–2 opening followed by an abrupt 4. Players who trained on low-pressure beats have not practiced under pressure. Fix: put one 3-intensity beat in the tutorial with real failure stakes.
- No landing. The game peaks at the final boss and then dumps you to the menu. A 1-intensity closing beat — walking out, a last line of dialogue — is what converts "that was hard" into "that was good."
Verify the curve against real players
Your authored intensity is a hypothesis. Test it cheaply:
- Record the session with timestamps. Even a phone recording of the screen works.
- Every 2 minutes, ask one question: "one to five, how tense are you right now?" It interrupts flow slightly; the data is worth it.
- Log every death and every time the player stops moving for more than 15 seconds without a menu open. Long stationary stretches are either confusion or boredom — the video tells you which.
- Plot their reported curve next to your authored curve.
Gaps between the two lines are your work list. A beat you scored 2 that testers report as 4 is a readability or difficulty problem. A beat you scored 5 that reports as 3 is a set piece that is not landing — usually because the player did not understand the stakes before it started.
If you keep your levels, objectives and playtest notes in GameDesignerX, the level and timeline modules are a reasonable home for this: put the beat table on the level, the intensity score in a field you can sort, and attach playtest sessions to the same beats so the authored and observed curves live side by side instead of in two different documents.
A pacing checklist
Before you call a level done:
- Every beat has an intensity score, and the sequence is written down somewhere a teammate can read.
- The intended rhythm (sawtooth, wave, escalating) is stated at the top.
- No more than two consecutive beats at 4 or higher.
- At least one beat at intensity 1 every 10–12 minutes.
- Rest beats are marked as protected from scope cuts.
- The peak lands in the final third, not the first.
- There is a closing beat below 2.
- At least three playtesters' reported curves have been compared to the authored one.
It takes about an hour to score a level you have already built. It is the cheapest design pass available to you, it requires no new art, and it fixes the failure mode that never shows up in a bug report — the tester who was having a perfectly fine time and simply stopped.