diff --git a/docs/scenarios/CLEAN_GROUND_PLAYTEST_6.md b/docs/scenarios/CLEAN_GROUND_PLAYTEST_6.md new file mode 100644 index 0000000..7320a4e --- /dev/null +++ b/docs/scenarios/CLEAN_GROUND_PLAYTEST_6.md @@ -0,0 +1,330 @@ +# CLEAN GROUND — desk playtest 6 (the four untested branches) + +**Scenario version:** v0.15, at commit 8705b00 +**Date:** 13 September 2026 +**Method:** **branch pass, not a session pass.** Seeded roller, **seed 3319**, banding from +`rules.mjs`. Earlier passes were 4271, 9314, 5580/7731, 6142, 8815. + +**Why this pass.** Five passes closed with the same four lines, unchanged each time: + +> *Still untested after five passes: the Swinburne fumble branch, the re-cut-the-peg +> fumble, a failed Xenology in Act Four, and any fight at all.* + +Three of those are rare — a 2% band, a 2% band, a 47% miss nobody happened to roll — and +waiting for a sixth honest session to walk into them is waiting for something that may never +arrive. So this pass **forces the trigger and then plays honestly from there**: for each +branch the roller draws from the seed's stream until the roll lands in the target band, and +every roll downstream of it is taken as it falls. The draw counts are printed so the forcing +is visible. **Nothing after a forced trigger was chosen.** + +**Every rating was read out of `roster.mjs` through `expandFromRegister`, not recalled** — +the standing consequence of pass 4 typing six of sixteen from memory and pass 5 overstating +one of its own. The sheet is printed below so the next pass can check it rather than trust it. + +``` +Ashcroft anomaly_lore 63 · insight 63 · listen 63 · dodge 63 · spot 40 · research 40 POWx5 85 NO WEAPON +Bhattacharya research 63 · anomaly_lore 63 · insight 40 · spot 40 · dodge 43 POWx5 75 pistol_light +Renshaw insight 63 · persuade 63 · listen 63 · psychology 40 · spot 40 · dodge 43 POWx5 70 pistol_light +Braithwaite medicine 53 · xenology(baseline) 53 · insight 53 · research 53 · spot 35 POWx5 60 dagger +Pollard listen 53 · spot 53 · dodge 53 · brawl 53 · anomaly_lore 35 POWx5 65 cudgel +Okonkwo firearm(rifle) 53 · brawl 53 · dodge 53 · spot 35 · anomaly_lore 33 POWx5 55 rifle_med, prybar +``` + +**Verdict: all four branches work as written. What this pass found is not in the branches — +it is that the scenario measures a fight nobody will have, and prices no fight in minutes.** + +--- + +## The rolls, in order + +``` +seed 3319 + +ACT ONE — the Swinburne fumble, FORCED + Renshaw Persuade — Swinburne 100 vs 63 FUMBLE [27 draws, band 99-100] + Braithwaite Medicine — the body 63 vs 53 failure + Bhattacharya Research — Corrigan's paperwork 85 vs 63 failure + Ashcroft Spot — the stump, walked to in silence 85 vs 40 failure + +ACT TWO — the column + Renshaw Persuade — the column 78 vs 63 failure + Pollard Spot — the same six, and the same six yesterday 93 vs 53 failure + Ashcroft Insight — it is offered, not remembered 27 vs 63 SUCCESS + +ACT THREE — the re-cut peg, FORCED + Bhattacharya Research — RESERVE ENTRY (NORTHERN) 27 vs 63 success + Ashcroft Anomaly Lore — what a peg is 100 vs 63 FUMBLE [103 draws, band 99-100] + hollow man swap 75 rolled 58 success vs Okonkwo 55 rolled 31 success + → equal band, goes to the resisting side. OKONKWO HOLDS. + +ACT FOUR — the Xenology that fails, FORCED + Braithwaite Xenology (Baseline) — the understudy 91 vs 53 FAILURE [2 draws] + Renshaw Insight — it is word-perfect 39 vs 63 success + Renshaw Psychology — it thinks it is the file 42 vs 40 failure + + COUNTDOWN STEP 5 — Ashcroft REFUSED, so it takes the lowest POW + understudy 75 rolled 34 success + Okonkwo 55 rolled 59 failure + → winner ACTIVE · settled TRUE · OKONKWO IS TAKEN +``` + +**A hostile seed, and deliberately not re-rolled.** Act One returned nothing at all: the +body, the paperwork and the stump all missed, on top of a fumbled Swinburne. That is the +worst opening any pass has produced and it is the right conditions for a branch test. + +**Two things happened here for the first time in six passes.** Ashcroft **refused** THE +OFFER — passes 4 and 5 both had him accept, and every published figure about him assumed +that. And **a player character was taken**, which is the one Act Four outcome the text +writes instructions for and no pass had reached. + +--- + +## The four branches, one at a time + +### 1. The Swinburne fumble — WORKS, and the text is right to like it + +She produced the dog. The walk to the stump in silence is a better Act One than the drawer +and the document's claim that *"the GM should not regret the fumble"* survives contact. + +- ⚠ **But this seed found the compounding case the branch does not consider.** Medicine + failed, Research failed, the Spot at the stump failed. **The party leaves Act One with a + hostile witness, no coroner's note, no Corrigan, and nothing from the stump** — and the + fumble branch's compensation, *"H01 still reaches them, later and from the coroner"*, is + then the **only** thread out of the act. The branch is written as though it is the one + thing that went wrong. It is the branch most likely to be surrounded by other failures, + because a table that rolls badly rolls badly all evening. +- Not a defect in the branch. A missing sentence: **when the fumble lands, the GM should + pull H01's arrival forward to the break rather than leaving it floating in Act Two.** + +### 2. The re-cut peg — WORKS, and it costs exactly what the text says + +Fumbled Anomaly Lore, the agent concluded with complete confidence the peg could be re-cut, +and the table spent twenty minutes on it. The instruction *"let them try it exactly once"* +is the right instruction and the silence at the end of it does teach what the text claims. + +- ⚠ **The warning at the top of Act Three names the wrong roll.** It says the bomb fires *"if + anybody fumbles **Anomaly Lore on the depot entry**"*. The depot entry is a **Research** + roll — `[CUS: Research — RESERVE ENTRY (NORTHERN)]`. The twenty-minute branch hangs off a + different beat entirely: `[CUS: Anomaly Lore — what a peg is, and what losing one does]`. + **A GM who reads the warning and watches the depot-entry roll is watching the wrong die.** +- This is not new. **Pass 4 inherited the same confusion** — its post-pass says it re-checked + "98 against Anomaly Lore 63" *on the depot entry*, which is a roll that does not exist. The + correction was right about the band and wrong about the beat, and nobody noticed because + the warning it was checking against had the same error in it. + +### 3. A failed Xenology in Act Four — WORKS, and the v0.14 fix is CONFIRMED + +Braithwaite rolled 91 against 53. The tell arrived at beat 3 as written, un-gated, in front +of the whole cordon, and **the act had a route**. This is the exact failure that killed desk +pass 1, and the fix that was written for it has now been tested rather than reasoned about. + +- Renshaw's Insight succeeded and the GM said the sentence out loud — *you read him + carefully, and there is nothing wrong with him*. From the chair it lands as the thesis and + not as a bug, which is what desk pass 1 doubted. +- Psychology failed, so *"it thinks it is the file"* surfaced in beat 3 unprompted, per the + missed-case. **Both of Act Four's gated lines failed in the same run and the act still + worked.** That is the strongest thing this pass has to say. +- ⚠ **Xenology has no fumble case and its fumble band is wider than the GM thinks** — see + finding 4 below. At 53 it fumbles on 98–00, and the document says nothing about what a + fumbled Xenology on the understudy produces. It should: the obvious wrong answer — + *"this is a baseline human, I am certain"* — is available, useful and unwritten. + +### 4. A fight — PLAYED, and it is the wrong fight + +See the two findings below. This is where the pass earned its seed. + +--- + +## ⚠ FINDING 1 — the fight the table produces is not the fight the document measures + +The honest trigger arrived on its own: **a hollow man tried the swap on Okonkwo in Act Three +and failed in front of everybody.** The column is present in that act. The six at the back +walk with it. A party that has just watched something try to climb into a colleague is a +party that shoots, and what they shoot into is forty-one refugees. + +**EXPOSURE measures "six hollow men" and "six of the column" as separate encounters. +The table produces them added together, and the document never says what the column does +when somebody fires into it.** + +`tools/simulate.mjs`, 2000 runs, seed 11, the cast of six: + +| What they actually fight | Hurt | Down | Wiped | Deaths/run | Median rounds | +|---|---|---|---|---|---| +| Six hollow men, column stands off | 3.48 of 6 | 0.42 | **0.0%** | 0.04 | 13 | +| **+ two of the column** | 5.02 of 6 | 2.78 | **7.8%** | 0.98 | 21 | +| **+ three of the column** | 5.55 of 6 | 4.07 | **28.7%** | 1.56 | 24 | +| + six of the column | 5.99 of 6 | 5.88 | **92.0%** | 2.33 | 16 | + +**Two of forty-one taking a swing moves the encounter from "nobody dies" to one wipe in +thirteen. Three moves it to better than one in four.** The printed row is 0.0%, and the +printed row assumes a crowd that stands still while its neighbours are shot. + +- **Stable across seeds**, so this is not one cold stream: wiped at 11 / 4242 / 90210 / 8815 + is 7.8 / 7.5 / 7.3 / 7.5% at two, and 28.7 / 29.1 / 27.8 / 28.5% at three. +- The document already knows the shape of this — *"the curve between six and ten is almost + vertical and a GM improvising bodies will cross it without noticing"* — and then prints a + hollow-man row that invites exactly that improvisation from the other direction. +- **Fix direction:** one row and one sentence in EXPOSURE. The six at the back are never + fought alone, and the GM's real question is not *how dangerous are they* but *how many of + the column join in*. Two is the number that matters. + +--- + +## ⚠ FINDING 2 — at four players, six hollow men wipe 58.7%, and the only figure printed is 0.0% + +This one is worse, because it is a number the document prints for a table it explicitly +supports. + +| Four-player cut | Hurt | Down | Wiped | Deaths/run | Median rounds | +|---|---|---|---|---|---| +| Three hollow men | 2.80 of 4 | 0.33 | **0.3%** | 0.03 | 13 | +| **Six hollow men, column stands off** | 3.94 of 4 | 3.40 | **58.7%** | 0.42 | 30 | +| + two of the column | 4.00 of 4 | 4.00 | **99.9%** | 1.11 | 9 | + +**EXPOSURE's four-player block contains no hollow-man row at all.** It lists three of the +column and six of the column, and its whole purpose is to say *the cut is a different game*. +A GM running four players looks up the hollow men, finds **0.0% wiped**, and is reading the +six-player table. + +- Stable: 58.7 / 60.5 / 60.3 / 59.8% across the four seeds. +- **The curve from three to six is near-vertical at four players — 0.3% to 58.7%** — which is + the same cliff the document warns about for the column between six and ten, in a place it + does not warn about. +- **And the hollow men are the ones the party is most likely to shoot**, because they are the + only figures in the valley the scenario tells the players are not people. The encounter + with the lowest moral cost is the one that wipes the smaller table. +- The four-player block's own explanation applies unchanged and should be borrowed: Ashcroft + carries no weapon, so the cut is three effective attackers against six. + +--- + +## ⚠ FINDING 3 — GM ESSENTIALS understates every fumble band in the scenario + +GM ESSENTIALS says: *"96–99 never succeed; 00 always fumbles."* + +Both halves are true and together they read as *a fumble is a roll of 00*. The rule in +`rules.mjs` is `fumbleStart = 101 - ceil((101 - band) / 20)`, tested **before** the 96–99 +clause, so the fumble band widens as the skill falls: + +| Rating | Fumbles on | Chance | +|---|---|---| +| 85 | 00 | 1% | +| 63 | 99–00 | 2% | +| 60, 53 | 98–00 | 3% | +| 40, 35, 25 | 97–00 | **4%** | + +**At Spot 40 a fumble is four times as likely as the essentials line implies**, and the cast +rolls Spot 40 repeatedly through two acts. Every fumble branch this pass was written to test +is commoner than the document's own rules summary suggests — which is an argument for the +branches, not against them. + +- Verified with `gradeRoll`: 98 against 63 is a **failure**; 98 against 53 is a **fumble**; + 97 against 40 is a **fumble**; 96 against 40 is a failure. +- Pass 4's post-pass got its own case right — 98 against Anomaly Lore 63 is a plain failure — + but got it right from the 96–99 clause rather than from `fumbleStart`, which is the same + answer by luck. Had the rating been 53 the correction would have been wrong. +- **Fix direction:** one clause in GM ESSENTIALS. *A fumble is the top 1/20th of the failure + range — 00 at high skill, 97–00 at 40 or below.* + +--- + +## ⚠ FINDING 4 — EXPOSURE prices fights in bodies and never in minutes + +The whole section is built around *what a fight costs*, and the cost it never states is the +one that ends conventions. + +| | Median rounds | +|---|---| +| Six hollow men, six players *(printed row)* | **13** | +| Six hollow men + three of the column | **24** | +| Six hollow men, four players | **30** | +| Six of the column, six players *(printed row)* | 14 | + +A d100 round with six agents and nine opponents is not a fast round. At three to four minutes +each — an estimate, not a measurement, and it should be measured — **the encounter this +scenario warns about is a sixty-to-ninety minute encounter.** Act Three is budgeted at fifty +minutes and already carries a twenty-minute bomb. + +- This pass hit both in the same act: fumbled peg **plus** the fight it triggered. With the + northern seam cut taken the act still cannot absorb them together, and nothing in the + Pacing Note contemplates a fight at all. +- EXPOSURE's five numbered lessons are all about lethality. **There should be a sixth: a + fight in this scenario costs the rest of the session, and that is a better reason not to + have one than the wipe rate.** A GM who is told the party probably survives has been told + the less important half. + +--- + +## Minor — `simulate.mjs` mislabels a mixed fight + +`--creature a,a,a,b,b` builds the right force and prints the wrong name for it. Six hollow +men and six of the column print as **"12 x Hollow man"**, taking the first spec's name and +the total count. The composition is correct — 12 hollow men wipe 3.5% of parties and the +mixed twelve wipe 92.0%, so the two are not being confused internally — but **anything +pasted out of that header into a document is false**, and every figure in findings 1 and 2 +came out of a run whose header lied about what was fought. + +Caught only because the header was read. It is the same shape as R-278's unread parameter +and R-264's summed armour: the tool was right and its report was not. + +--- + +## Timing + +**This is not a session timing pass** — passes 4 and 5 measured the clock and nothing here +supersedes them. These are branch costs, measured against v0.15's budgets: + +| | Budget | This pass | Note | +|---|---|---|---| +| Act One, everything failed + the dog | 45 | ~40 | Fewer clues means fewer conversations, not more | +| Act Three, fumbled peg, seam cut taken | 50 | ~55 | Exactly as the warning predicts | +| Act Three **with the fight** | 50 | **~115** | The act cannot hold it. Nothing says so | +| Act Four, both gated lines failed | 45 | ~45 | On budget, un-gated tell carried it | + +--- + +## Player-hat verdicts + +| Archetype | Verdict | Score | +|---|---|---| +| Methodical veteran | Act One returned nothing and the dog was still the best scene of the night | 8/10 | +| Roleplay-first talker | Walking to the stump behind a woman who will not speak to you — best silence in the case | 9/10 | +| Quiet tactician | Shot the six at the back, and would do it again. Nobody at the table knew what that cost | 4/10 | +| Lore-digger | Spent twenty minutes re-cutting a peg and learned more from it failing than from the depot | 8/10 | + +--- + +## Fix list + +1. **⚠ Measure the fight the table will actually have.** *(EXPOSURE)* Add the mixed rows — + six hollow men with two and three of the column joining, **0.0% → 7.8% → 28.7%** — and + say plainly that the six at the back are never fought alone and the GM's live question is + how many of the column join in. + +2. **⚠ Give the four-player block a hollow-man row.** *(EXPOSURE)* **58.7% wiped** against + the 0.0% currently printed, with the 0.3%-at-three row beside it so the cliff is visible. + The encounter with the lowest moral cost is the one that wipes the cut. + +3. **⚠ Fix the Act Three warning's roll name.** The twenty-minute bomb is on **Anomaly Lore — + what a peg is**, not on the depot entry, which is a **Research** roll. Correct pass 4's + post-pass in the same breath; it inherited the error. + +4. **⚠ State the real fumble band in GM ESSENTIALS.** "00 always fumbles" reads as 1% and is + 4% at Spot 40. One clause: *the top twentieth of the failure range, so 97–00 at 40 or + below.* + +5. **Add the sixth EXPOSURE lesson: a fight costs the session.** Median 13 rounds on the + printed row, 24 mixed, 30 at four players. Price it in minutes beside the wipe rate, and + put a line in the Pacing Note about what to cut when one starts. + +6. **Write the fumbled Xenology.** At 53 it fumbles on 98–00 and the document is silent. + *"This is a baseline human, and I am certain"* is the obvious wrong answer and it is worth + more than a shrug. + +7. **Minor: label mixed fights in `simulate.mjs`.** A run of `a,a,b` prints as "3 x A". + +**Still untested after six passes** — the list is finally shorter, and honest about why: +a fight that the *players choose* rather than one this pass reasoned them into; Act Four with +a PC taken **and** Ashcroft having accepted, which is the pairing the countdown was written +for; and the first human run, which remains the only thing that can settle whether any of +this survives a table.