RingBRP/docs/scenarios/CLEAN_GROUND_PLAYTEST_1.md

CLEAN GROUND — desk playtest 1 (GM chair)

Scenario version: v0.5, at commit 15da21d Date: 12 September 2026 Method: full session from the GM chair, all six cast pregens played as distinct archetypes. Every contested beat rolled with a seeded roller (mulberry32, seed 4271, rolls consumed in the order printed below) so the run is re-checkable rather than remembered. Banding per rules.mjs: critical under a twentieth, special under a fifth, 96–99 fail, 100 fumbles.

Archetypes to sheets: Bhattacharya = methodical note-taker · Renshaw = roleplay-first talker · Pollard = quiet rules-literate tactician · Ashcroft + Braithwaite = lore-diggers · Okonkwo = the one who asks what the plan is and then does it.

Verdict up front: not runnable. Three clue routes cannot be rolled by the cast, the one working test in Act Four has no failure case, and the scenario is about five hours long in a three-hour-forty-five budget.

---

Pre-flight audit (before any dice)

Every skill the acts name, against the six sheets actually cast:

Called for	Where	Who in the cast can roll it	Base
Track	Act One, sets from the stump	nobody	10
Science (Botany)	Act One, the ring count	nobody	1
Navigate	Act Three, the pole star	nobody	10
Repair (Mechanical)	Act Three, the lorry	Okonkwo 35	25
Psychology	Act Four	Renshaw 40	5
Xenology (Baseline)	Act Four, the tell	Braithwaite 53	1

⚠ Three of the scenario's clue routes are unrollable by the party it casts. This is the same defect class as the Scent line that was fixed in Act Four, and Science (Botany) is literally identical to it: base 1, so the 1% floor is the whole of it.

⚠ check-scenarios cannot catch this. It resolves skill names against the catalogue; it has no opinion on whether anybody present can roll one. The guard that would have caught this does not exist.

---

Play log

Act One — Registry and the Sike Head · planned 0:45, actual 0:55 ⚠

Bhattacharya Research — body unfiled       76 vs 63   fail
Bhattacharya Knowledge — why a Geiger      18 vs 63   success
Pollard      Track — sets from the stump   10 vs 10   success
Braithwaite  Science (Botany) — ring count 99 vs  1   fail
Braithwaite  Medicine — the bones          25 vs 53   success
Pollard      Spot — the notch              63 vs 53   fail

• Research failed, and the handout saved the act. Swinburne's photocopy of CG-H01 is written as the failure case and it worked exactly as intended. Keep that pattern.
• Medicine landed, so the body clue came through twice over. Braithwaite read the nine resolings aloud and the talker asked who resoles boots nine times. That is the scene working.
• The peg clue survived on a 10% roll coming up exactly 10. Track succeeded on the nose. Had it failed — and it fails nine times in ten — the act's stated failure case hands over a boot belonging to nobody in Britain.
• ⚠ The failure case backstops the wrong clue. A strange boot says something came through. It does not say this stump is a survey peg, which is one of the three the act says must land. With Track at 10, Botany at 1 and Spot failed, this run reached the seam with the peg understood only by luck.
• Timing: the Registry briefing ran 15 minutes because six players each asked one sensible question. Stores 10, farm and stump 15, body 10, seam 5.

Act Two — Over · planned 1:05, actual 1:00 ✓

Ashcroft Insight — that memory is new      74 vs 63   fail
Renshaw  Persuade — the column             48 vs 63   success
Pollard  Spot — the same six at the back   89 vs 53   fail

• The best act in the scenario, and the only one that came in under time.
• The crossing read-aloud did its job. Nobody rolled anything, the silence held, and the tactician broke it by asking for the Geiger — which is the Act One Knowledge success paying off an act later. Good wiring.
• Insight failed, so the Coherence drain went unnoticed, which costs nothing yet and is arguably better: the record should be winning quietly at this stage.
• Spot failed and it did not matter, because the read-aloud already says six at the very back are not quite keeping up. The information was in the room without the roll. This is the design pattern the rest of the scenario should copy.
• Renshaw spent 25 minutes with Ivy and would have spent forty. The talker's verdict: the single best scene in the document.

Act Three — There is no north · planned 0:50, actual 1:05 ⚠

Ashcroft     Navigate — pole star           93 vs 10   fail
Pollard      Spot — own boot prints         72 vs 53   fail
Okonkwo      Repair (Mech) — the lorry      11 vs 35   success
Bhattacharya Research — RESERVE ENTRY        7 vs 63   special
Ashcroft     Anomaly Lore — what a peg is  100 vs 63   FUMBLE

• Two of the three loop routes were unavailable or failed; the third carried it. Okonkwo recognised the wrecked lorry on its third appearance with his own chalk on it, and the tactician worked out the loop from that alone. Then CG-H03a and CG-H03b landed side by side and the table did the rest out loud, exactly as staged. The handout is doing the work the three skill routes were supposed to do.
• Research special put RESERVE ENTRY in Bhattacharya's hands, and the A.E.W. initials landed — the note-taker went quiet and then asked whose slide rule he was carrying. Best single moment of the run.
• ⚠ Anomaly Lore fumbled on 100 and the scenario has no answer. What does a fumbled "what a peg is" do? Nothing is written. I invented a ruling at the desk (he concludes the peg can be re-cut, which is wrong and wastes ten minutes) and an invented ruling is a hole in the text by definition.
• Telling Ivy ran 15 minutes and was worth every one. No roll, as written. Correct.
• Timing: the act has five beats and needs four. It ran 15 over.

Act Four — One of us · planned 0:45, actual 0:50, and shapeless ⚠⚠

Renshaw     Insight — the copy              70 vs 63   fail
Renshaw     Psychology — it thinks it is the file  8 vs 40   special
Braithwaite Xenology (Baseline) — the tell  77 vs 53   fail

• Psychology's special was the reveal of the night: it is not pretending to be human, it thinks it is the file. The talker repeated that line twice. Keep it.
• ⚠⚠ THE CRITICAL DEFECT. Xenology (Baseline) failed and the scenario has no failure case for it. It is named as "the only test that works", it exists above the floor on exactly one sheet, and when that sheet rolls 77 the act has no route left. I sat at the desk with nothing written to fall back on. This is the concentration risk that was logged as an open question, arriving as an actual dead end on the first run.
• Insight failing costs nothing, because the written result of succeeding is "it is word-perfect" — i.e. no information either way. Fine, but note that a player who rolls well gets nothing, which feels like a bug from the player's chair even though it is the point.
• ⚠ The act has no spine. The exposure model correctly says there is no mechanical threat, which is right — but the act is then 45 minutes of doubt with no written beat structure and no clock of its own. It sprawled. With a less patient table it would instead have collapsed in 20 minutes into "we shoot it", and the text has no answer for how to fill the other 25.

Close — Cancellation · planned 0:25, actual 0:25 ✓

• Ivy's read-aloud worked reported rather than quoted, as intended. She asks whether there is a north stopped the table dead.
• The party tried bring them through, as the text predicts. Got four across, did the arithmetic on the other thirty-seven, and chose to leave the entry filed. The costing held up and needed no improvisation.
• No roll, no NPC advice, decision made in about nine minutes of real argument. This act is finished and should not be touched.

---

Timing summary

Act	Planned	Actual	Drift
One	0:45	0:55	+10
Two	1:05	1:00	−5
break	0:10	0:10	—
Three	0:50	1:05	+15
Four	0:45	0:50	+5
Close	0:25	0:25	—
Total	3:55	4:25	+30

⚠ And that is a generous run: a 10% Track landed, a 35% Repair landed, and a 63% Research came up special. A table that misses those spends longer, not shorter, because the GM starts improvising bridges. Assume 4:45 to 5:00 for a real table of strangers in a 4:00 slot. There is no slack anywhere and the protocol is explicit that on-the-minute is already a fail.

---

Player-hat verdicts

Archetype	Verdict	Score
Methodical note-taker (Bhattacharya)	Superbly served. RESERVE ENTRY and the A.E.W. initials are the best material in the scenario for this player.	9/10
Roleplay-first talker (Renshaw)	Two outstanding scenes, Ivy and the copy. Would happily have played only those.	9/10
Quiet tactician (Pollard/Okonkwo)	Got the loop and enjoyed it, but three of his natural routes were skills nobody had, and Act Four gave him nothing to do once violence turned out to be pointless.	5/10
Lore-digger (Ashcroft/Braithwaite)	Braithwaite owns two clues and that is good — until the one that matters failed and the text abandoned him. Ashcroft's Coherence material never came up mechanically at all.	4/10

Ashcroft is the weakest sheet in play. His private seam is the best-written one in the document and nothing in five acts forced it. He spent the session as a second Insight.

---

Fix list

Nothing below is applied. Each states the problem, the change, and where it goes.

1. CRITICAL — write a failure case for Xenology (Baseline). Text hole. Act Four. Recommended: make the physical tell automatic on close examination and let the roll buy it early and privately instead. A failed roll then costs time and exposure rather than the whole route. Alternative: Countdown step 4→5 forces the tell into the open whether anybody rolled or not.
2. CRITICAL — move the three unrollable clues onto skills the cast holds. Text hole. Acts One and Three. Track → Spot (Pollard 53). Science (Botany) → Knowledge (Bhattacharya 63) or automatic on inspection. Navigate → Spot or Anomaly Lore. Recasting instead — adding Sandoval, Lindqvist or Holloway — fixes the symptom and leaves the trap set for the next GM who uses the recommended six.
3. Rewrite Act One's failure case so it backstops the clue it claims to. Text hole. A strange boot proves arrival, not a peg. The stump needs a route that cannot be missed — the notch itself, visible to anyone who looks, with the roll buying the 1962 date.
4. Write the fumble result for Anomaly Lore. Text hole. Act Three. Also audit every other roll in the document for a written failure, since this is the second time this class of bug has appeared in a Custodians scenario.
5. Cut 30–45 minutes, pre-written, not improvised. Timing. Act Three loses a beat (the northern seam walk is redundant once the lorry and the almanac exist). Act One's briefing needs a hard 10-minute cap written into it. Put both in a Pacing Note.
6. Give Act Four a spine. Staging. It needs its own short clock — three or four named beats with an order — so it neither sprawls nor collapses. The exposure model says the tension cannot come from threat, so the text must supply the structure instead.
7. Force Ashcroft's seam. Missing option. Add one beat where the record's offer to him is explicit and costed, so the best private box in the document actually fires.
8. Note that a successful Act Four Insight yields nothing and say so in the text, so the GM can tell the player that the absence of a tell is the information.

What is already working and must not be lost

• Handouts carrying the case when rolls fail. CG-H01's photocopy, CG-H02 in Bhattacharya's hand, and the H03a/H03b pair all did their jobs. The almanac trick worked exactly as staged.
• Seeding a clue in read-aloud so a failed roll is harmless — the six at the back. This is the pattern fixes 2 and 3 should imitate.
• The Close. Finished. Do not touch it.

---

---

FIXES APPLIED — v0.6, verified against this same seed

All eight applied as a separate pass. Verification was a re-run of seed 4271 against the rewritten acts, so the identical dice that broke the first run were put through the fixed text. One beat was added (THE OFFER), so the stream diverges after Act Two's first Insight.

The roll that broke it	Then	Now	Outcome
Ring count	Botany 99 vs 1 — clue dead	Knowledge 99 vs 63 — still fails	The notch is automatic. Peg lands anyway; only its 1962 meaning defers to CG-H02
Tracks from the stump	Track 10 vs 10 — survived on the nose	Spot 10 vs 53 — special	Found handsomely instead of by luck
Pole star / sky	Navigate 93 vs 10 — dead	Anomaly Lore 93 vs 63 — still fails	Almanac pair carries it; loop still found via Repair 11
What a peg is	FUMBLE 100 — no text, I invented a ruling	FUMBLE 100 — same roll	Written now: the re-cut delusion, tried once, fails in silence
The tell	Xenology 77 vs 53 — act had no route	Xenology 77 vs 53 — same fail	Tell is ungated. Roll now buys it early and privately; failing costs timing, not the act

Every roll in the document now has a written miss. Act One's three must-land clues all landed on this run despite two failures, one of them automatically.

Timing is improved but not solved. The cuts recover about 20–25 minutes against the 4:25 actual, putting a re-run near 4:00–4:05 on a 3:55 budget. That is still no slack, and this protocol is explicit that on-the-minute is a fail for strangers. A third cut is likely needed — the honest candidate is compressing beat 2 of Act Four, since the tell no longer depends on getting through all three tests.

Not verified by this pass, and not verifiable from a desk: whether Act Four's spine actually holds 45 minutes at a real table, and whether THE OFFER lands as a decision rather than a formality. Both need the player chair.

Next pass: stress re-run of Act Four (weakest act, most changed by the fixes), then a player-chair pass playing contrarian — refuse the column, shoot early, skip the depot.