diff --git a/docs/scenarios/CLEAN_GROUND_PLAYTEST_7.md b/docs/scenarios/CLEAN_GROUND_PLAYTEST_7.md new file mode 100644 index 0000000..1958c11 --- /dev/null +++ b/docs/scenarios/CLEAN_GROUND_PLAYTEST_7.md @@ -0,0 +1,257 @@ +# CLEAN GROUND — desk playtest 7 (an honest session against v0.16) + +**Scenario version:** v0.16, at commit aa3ae24 +**Date:** 13 September 2026 +**Method:** **full session, GM chair, six players — the declared cast.** Seeded roller, +**seed 7742**, banding from `rules.mjs`. **Nothing was forced.** Pass 6 forced its triggers +on purpose and said so; this one takes what the dice give, which is the only way to find out +what an ordinary evening looks like against the current text. Earlier passes: 4271, 9314, +5580/7731, 6142, 8815, 3319. + +Ratings read through `expandFromRegister`, not recalled. + +**Why this pass.** v0.16 applied seven fixes off a *forced* pass. Forced passes prove that a +branch works; they cannot tell you how often a GM meets one. And the list pass 6 closed with +named two things an honest session might reach — a fight the players choose, and Act Four +with Ashcroft having accepted. It reached neither. **What it found instead is the thing both +of those are instances of.** + +**Verdict: the document is excellent at the two outcomes a table sees most and nearly silent +at the three it sees the rest of the time. Seventeen failure cases are written across the +acts — the house rule, honoured. Four beats state a fumble. Three state a special or a +critical. Nine sessions in ten will contain a special or better on a beat the document says +nothing about.** + +--- + +## The rolls, in order + +``` +seed 7742 — no forcing + +ACT ONE + Bhattacharya Research — Corrigan's paperwork 36 vs 63 success + Braithwaite Medicine — the body 99 vs 53 FUMBLE + Renshaw Persuade — Swinburne 1 vs 63 CRITICAL + Pollard Spot — the stump 100 vs 53 FUMBLE + Ashcroft Listen — the valley, crossing 81 vs 63 failure + +ACT TWO + Renshaw Persuade — the column 19 vs 63 success + Pollard Spot — the same six at the back 26 vs 53 success + Ashcroft Insight — it is offered, not remembered 42 vs 63 success + → ASHCROFT REFUSES THE OFFER + +ACT THREE + Bhattacharya Anomaly Lore — the sky has not moved 39 vs 63 success + Bhattacharya Research — RESERVE ENTRY (NORTHERN) 84 vs 63 failure + Ashcroft Anomaly Lore — what a peg is 21 vs 63 success + Pollard Listen — the lorry 45 vs 53 success + hollow man swap 75 rolled 98 failure vs Okonkwo 55 rolled 52 success + → OKONKWO HOLDS. Second pass running the swap has failed. + +ACT FOUR + Braithwaite Xenology (Baseline) — the understudy 6 vs 53 SPECIAL + Renshaw Insight — it is word-perfect 56 vs 63 success + Renshaw Psychology — it thinks it is the file 62 vs 40 failure + Bhattacharya Borrowed Authority — pull the record 92 vs 63 failure +``` + +**A generous seed — and it still landed on four bands the document does not cover.** A +fumbled Medicine, a fumbled Spot, a critical Persuade and a special Xenology, in thirteen +rolls. Three of those four are *good* rolls. This is not a pass about things going wrong. + +⚠ **Two fumbles in thirteen rolls is a 4.4% event and is not evidence of a rate.** It is an +illustration and nothing more; the rates below are enumerated from the rule, not counted from +this seed. The project has been caught reading a single stream as a distribution more than +once and this pass is not adding to the tally. + +--- + +## ⚠ FINDING 1 — the document covers two bands out of five + +This is the pass's result, and it arrived by accident: the seed simply kept landing outside +success and failure. + +| Band | Beats that state an outcome | Of 22 rollable beats | +|---|---|---| +| Failure | **17 written** | the house rule, honoured properly | +| Fumble | **4** | Swinburne, the peg, the Xenology (new in v0.16), one in the Close | +| Special or critical | **3** | the column's Persuade, the child's Insight, Borrowed Authority | + +Enumerated through `gradeRoll` across all 100 results at each rating, against the cast's best +rating for each of the twenty-two beats the acts name: + +| | Per session | +|---|---| +| At least one **fumble** | **42.1%** | +| At least one fumble **on a beat with no written case** | **36.0%** | +| At least one **special or better** | **93.9%** | +| At least one special or better **on a beat with no written case** | **90.8%** | + +**Nine sessions in ten contain a good roll the document has no answer for.** Specials are +not rare — 13% at a rating of 63, 11% at 53 — and the scenario asks for twenty-two rolls. +The three places that *do* write one are the three best beats in the case: *give something +portable, never more conversation*; *she has been told to smile, and she is doing it +carefully*; *the file has been read recently, from inside the cordon*. The document already +knows how much a special is worth. It writes one about one time in seven. + +- **This pass's own two.** Renshaw rolled **1** on Swinburne — a critical on the scene the + document has taken the most care over, and there is nothing there. A GM improvises "she is + very helpful", which is the flat outcome desk pass 4 complained about *for the fumble* and + which v0.11 fixed by writing the dog. The critical deserves the same treatment and has not + had it. Braithwaite rolled **6** on Xenology, a special, and the text gives a success — a + quiet word in the GM's ear — with nothing above it. +- **Fumble bands, by contrast, are now stated** in GM ESSENTIALS as of v0.16, and this pass + is what that clause was for: **99 against Medicine 53 is a fumble.** Under the wording v0.15 + carried it reads as a plain failure and Braithwaite simply learns nothing from the body. + The clause earned itself in the very next pass. +- **Fix direction, and it is small.** Twenty-two beats do not need twenty-two specials. The + house rule for failure is *every clue states its failure case*; the parallel rule should be + **every beat that can be rolled more than once, or that a player will care about winning, + states what a special buys** — and the existing three show the shape: something portable, + something private, or something early. + +--- + +## ⚠ FINDING 2 — v0.16 wrote the trigger into EXPOSURE and never wrote EXPOSURE into the trigger + +Self-inflicted, one version old, and found by trying to run the thing. + +The mixed-fight block added in v0.16 is explicitly about one moment: *a hollow man tries the +swap in Act Three, fails in front of everybody, and the party fires into the column it is +standing in*. **That beat, at line 768, contains no pointer to it.** It reads: + +> **A hollow man tries the swap** — opposed POW×5 (GM ESSENTIALS has the ruling), 2 +> Coherence. Survivable, and it is the table's first sight of the mechanism… + +"Survivable" is the last word a GM reads before the most expensive decision in the case. + +- **Act Two does it correctly** and has since before this pass: *"If violence against the + column is even possible at your table, read EXPOSURE in Act Four first."* Act Three, where + the trigger now demonstrably is, says nothing. +- **Distance matters here.** The swap is at line 768; the mixed table is at line 910; the + cost-in-minutes lesson at 1008; the Pacing Note instruction at 182; the opposed-roll ruling + at 86. **Five locations across 1,100 lines to adjudicate one beat**, in a document whose own + prime directive is that a GM reads it in thirty-second glances. +- EXPOSURE's heading still says *"Read before Act Two, not before Act Four"* — the document + knows it is filed in the wrong act and asks the reader to compensate. That is a reasonable + trade for a long reference block. It is not a reason for the beats to be silent. +- **Fix direction:** one line under the swap beat, in the shape Act Two already uses. + +--- + +## ⚠ FINDING 3 — three of the four fumble branches do not print their band + +v0.16 put the bands in GM ESSENTIALS, which was right, and then left the branches reading +"On a fumble" with no number. A GM at line 398 with a 99 on the table has to know to go back +to line 80. + +| Fumble branch | Prints its band? | +|---|---| +| Persuade — Swinburne (63) | no — should read **99–00** | +| Anomaly Lore — the peg (63) | no — the Act Three warning has it, the branch does not | +| Xenology — the understudy (53) | **yes, 98–00 at his 53** | +| The Close (line 1173) | no — and it names **no skill at all**, so there is no band to look up | + +The one that prints it is the one written in v0.16, which is the tell: the fix was applied +where it was being thought about and not where the same need already existed. **Four words +per branch.** + +--- + +## Act-by-act + +**Act One — ran ~50 against 45, and the overrun is the finding.** Research landed early and +Corrigan's unfiled casual labour came out, the best opening this scenario has. Then both +fumbles arrived in the same act with nothing written for either, and **improvising two +fumble outcomes cold cost about ten minutes** — not because improvising is slow, but because +a GM who has just read a beautifully specific fumble branch for Swinburne assumes the others +exist and spends time looking for them. + +- **The critical on Swinburne was the one that hurt.** Renshaw rolled a 1 on the scene the + document cares most about, and the best available reading is "she produces the note + willingly", which is what she does on a plain success. **A critical that buys nothing is + worse than a fumble that buys a dog**, and the table noticed. +- Pollard's fumbled Spot at the stump: played as him confidently identifying the felling as + recent storm damage, which put the party off the oak for ten minutes. Serviceable, and + entirely invented. + +**Act Two — ran ~40 against 45.** Clean. Persuade and Spot both landed, the six at the back +were noticed early and flatly, and **Ashcroft refused THE OFFER** — the third honest pass in +four to go that way, which is what a 63% Insight predicts. Worth stating plainly: **the +Act Four figures the document prints most prominently are for the 37% branch.** + +**Act Three — ran ~45 against 50.** The depot Research failed and Ivy's ration tin carried +it exactly as the *Missed* line promises; that fallback is the best-written thing in the act. +The peg question was answered correctly — *it cannot be done, and why* — and it landed as +information rather than refusal, as the text intends. **The swap failed for the second pass +running**, so the mechanism has still never actually taken an NPC in front of the party. + +**Act Four — ran ~45.** On budget. Xenology special, so the tell came early and privately; +Psychology missed and surfaced in beat 3 unprompted; Borrowed Authority failed, so the +timestamp clue never appeared. With Ashcroft having refused, step 5 reached for Okonkwo at +POW×5 55 — **and the pairing pass 6 asked for (accepted *and* taken) remains unreached +after seven passes.** It needs a 37% and then a 48.7%; at roughly 18% a session, it will keep +not happening. + +**Close — ran ~25.** On budget. + +--- + +## Timing + +| Act | Budget | Ran | Δ | +|---|---|---|---| +| One | 45 | ~50 | **+5** — two unwritten fumbles | +| Two | 45 | ~40 | −5 | +| Three | 50 | ~45 | −5 | +| Four | 45 | ~45 | — | +| Close | 25 | ~25 | — | +| **Play** | **210** | **~205** | **−5** | +| **Wall clock** | **3:40** | **~3:35** | **−5** | + +**Inside budget on a generous seed, at six players, with every cut available and none taken.** +That is the first honest pass to land inside the clock without help, and it is the strongest +thing this pass has to say about v0.13's recut. + +--- + +## Player-hat verdicts + +| Archetype | Verdict | Score | +|---|---|---| +| Methodical veteran | The ration tin catching the failed depot roll is the best save in the case | 9/10 | +| Roleplay-first talker | Rolled a 1 on the most important conversation in Act One and got an ordinary outcome | 5/10 | +| Quiet tactician | Noticed the six at the back early and cleanly. Nobody shot anybody. Good evening | 8/10 | +| Lore-digger | The peg answer is the moment the case clicks, and it arrived as information | 9/10 | + +--- + +## Fix list + +1. **⚠ Write what a special buys, on the beats where it matters.** Three of twenty-two + currently do, and **90.8% of sessions contain a special or critical on a beat that does + not**. Start with the two this pass hit: a **critical Persuade on Swinburne** (she does not + just hand over the note — she comes with them to the stump, which she otherwise only does + on a fumble, and this time she talks), and a **special Xenology** (the tell, early and + private, *plus* what the hands are imitating — that it has watched somebody specific). +2. **⚠ Point the Act Three swap beat at EXPOSURE.** One line, in the shape Act Two already + uses. v0.16 named that beat as the trigger for the fight it measured and then left the beat + silent, with "survivable" as the last word before the decision. +3. **⚠ Print the fumble band at each fumble branch.** Three of four do not. `99–00` at + Swinburne and at the peg. **The Close's is worse than missing a band — it names no skill**: + *"on a fumble asking the question"* has no `[CUS: …]` tag on it, so there is nothing to + roll and nothing to grade. Either tag it or make it a GM call in terms. +4. **Write the fumbled Medicine and the fumbled Spot**, or say in one line that unwritten + fumbles are the GM's to invent and stop implying otherwise. **36% of sessions hit one.** + The document's four written fumbles are so good that their absence elsewhere reads as an + oversight rather than a licence. +5. **Say that Act Four's headline figures are for the 37% branch.** Ashcroft refuses about + five times in eight, and this pass and pass 6 both did. The 48.7% / 14.5% pair is printed + far more prominently than the 21.9% / 3.8% one that most tables will actually be in. + +**Still untested after seven passes:** a fight the players choose; Act Four with Ashcroft +having accepted *and* being taken (~18% a session — it may take several more passes and that +is not a defect); a hollow man that succeeds at the swap in Act Three, which has now failed +twice running; and the human run.