Pacing From Player Data: Find the Boring Bits
Compare how players really spend their time with what you planned, and look for long quiet stretches that people leave right after. Quiet alone is not a problem, since someone reading a good scene is quiet too, so what counts is who stops afterwards.
How do you see pacing in a list of events?
Orbit Laundry is an invented game about running the laundry on a space station, and these numbers are made up for this example. You fold, deliver and chat with the crew. Before looking at any data, its maker wrote a plan for how the time should be split. Then he looked at 60 sessions.
The middle session lasted 22 minutes, and half of them were between 14 and 31 minutes. The middle is used and not the average, because one person who wandered off and left the game open would drag an average up.
| Activity | Share of playing time | Plan |
|---|---|---|
| Waiting for the dryer | 34% | Under 15% |
| Folding | 28% | 30% to 40% |
| Deliveries | 24% | 30% to 40% |
| Talking to the crew | 14% | 10% to 20% |
A third of the time goes on watching a machine, and it has pushed the fun down. To find when it hurts, he looked at long gaps in the log, moments when nothing happened for over a minute.
| Quiet after | Sessions | Usual length | Stopped within a minute after |
|---|---|---|---|
| dryer-start | 41 | 4 min 10 s | 19 |
| talk-start | 33 | 1 min 30 s | 1 |
These two gaps look alike and mean opposite things. After the dryer starts, 19 of 41 sessions end within a minute of the wait being over. That is 46%, very likely between 32% and 61%. After a talk begins, the gap is someone reading a conversation: 1 of 33 stopped. The log cannot tell a thrilled reader from someone making a sandwich. Only your plan and the stops afterwards can.
The list of who was still playing agrees.
| Minute | Still playing |
|---|---|
| 0 | 60 |
| 5 | 57 |
| 10 | 51 |
| 15 | 38 |
| 20 | 33 |
The biggest fall comes between minutes 10 and 15, which is when the second load of laundry goes in.
How do you fix a slow stretch?
- Give the wait a job. The maker added a small task during the dryer: a lost sock to find.
- Shorten it, or let players leave. Could the wait be 90 seconds? Could they do a delivery and come back?
- Keep the good quiet. Do not fill the crew talks with noise. A rest is part of a rhythm.
- Change one thing at a time. If the wait gets shorter and a job is added, you will not know which helped.
Watch out for two traps. Long sessions are not automatically good, since waiting makes a session longer. So do not chase session length. And what people do last is only a clue, because everybody has to stop at some point. See time to first fun for the start of the game, and the session pacing recipe for this whole reading in one report. With fewer than about 30 sessions, treat it as a hint and go watch someone play.
Paste this to your AI
Tell me where my game drags. The event log is in [file or folder].
1. Ask me how I want the time to be divided, activity by activity. Then draw the rhythm I want as a Mermaid state diagram: the order of activities, and how long each should last. Mark any wait or stretch that is on purpose.
2. Name which event starts each activity. Check that every activity has one. If not, tell me what to add.
3. Use get_recipe with session-pacing if Spielerando is connected. Show: session length with the middle and the quarters, share of time per activity next to my plan, how many are still playing every 5 minutes, and quiet gaps of over a minute, with who stopped soon after.
4. For each quiet gap, say whether my plan expects it. Separate gaps people stay through from gaps people leave after, and show the counts.
A log cannot show what someone was thinking, so do not claim to know. Give me the one stretch most worth changing, and one way to change it. Do not use session length as proof that the game is good. With under 30 sessions, say this is only a hint.
Or skip this and connect the MCP: your AI then finds all of this by itself.Connect the MCP
How to check it worked
- Play it with a stopwatch. Time the stretch your AI named. Is it as long as it says?
- Sit through it without your phone. What are you doing in those minutes?
- Add up the shares. All activities should come to 100%. If not, something has no event.
- Measure after the change. The fall in the still-playing list should move or shrink, not just the share of time.