7.7 — Level Studio II + Cross-Playtest

From yesterday's walk ritual and your written plan (7.6), you already know which two beats need work. Polish means these four things, and nothing else:

• Sightlines: trim or add tiles until the next beat is glimpsed from the current one. Every removed tile is a question the player stops asking.
• Telegraphing: the turret is visible at rest before it fires — exactly as you tuned it in 3.4. A threat the player never saw is not difficulty, it is an ambush.
• Reward placement: coins on the route you want walked. A coin trail is a dotted line the player follows without ever noticing they are being led.
• Death-audit repair: the beat where dying costs the most gets a shorter approach, or a checkpoint you build yourself — a trigger that stores a position and a respawn that reads it. Your respawn loop from 2.2 is all you get unless you build more.

Do not polish everything. Two beats, deeply. That is how studio time works under a deadline, and your deadline is the cross-playtest, which starts in twenty minutes.

Every field that watches people has had to invent a unit of measurement. Read each of these for what it does in isolation, then work out which column of your playtest table it is telling you to be careful about.

Example 1 — A hospital's hand-hygiene audit:

// Count per station, never per shift. One number for the whole
// ward hides the one sink nobody uses.

Example 2 — A supermarket's checkout timing study:

// Time per basket, not per queue. The average is a story;
// the distribution is the finding.

Example 3 — An airport security throughput trial:

// The stopwatch starts when the tray reaches the belt,
// not when the passenger walks up.

Example 4 — A traffic engineer's junction survey (comment only):

// Count the near-misses, not only the collisions. No crashes
// is not evidence of a safe junction.

Example 5 — A usability lab's think-aloud protocol (comment only):

// Record what the participant DID before what they SAY they
// did. The two disagree constantly.

• Example 1 refuses to average over a whole ward. Your table has one row per beat — what would a single total for the cave hide about your level?
• Example 3 defines exactly when the clock starts. When does your stopwatch start for each beat, and would two testers agree on that moment?
• Example 4 counts near-misses, not only disasters. A beat where nobody dies but everybody hesitates is a near-miss — what on your table would record it?
• Example 5 separates what people did from what they said. Which column of your table holds observations, and which one is in danger of holding your interpretation?

New partner — not yesterday's. Silent watching, per the rules you set in 2.4. The full cave, entrance to goal. Record per beat, not per session, because the beat is the unit of analysis. This table is the exact habit that becomes a dataset in 9.5.

Fill the "tester said / did" column with what you observed, not with what you assume they meant. "Stopped for four seconds at the ladder base and looked left" is data. "Confused by the ladder" is a story you invented.

Then swap. You play their cave with the same table. You are both designer and tester today — notice how different the two jobs feel, and notice which one you are better at.