Concept Prototype Report: The Muster (displaced-cost allocation)
Date: 2026-09-14
Prototype Path: Paper
Concept File: design/gdd/game-concept.md
Verdict: PIVOT (corrected 2026-09-14 — was PROCEED)
⚠ CORRECTION — 2026-09-14
**This report originally recorded PROCEED with the hypothesis CONFIRMED. That verdict was wrong and has been withdrawn.** An adversarial economy review found two defects that invalidate the evidence. Both were findable before the playtest ran; neither was caught by the person who designed the rules, ran the session, and wrote the verdict (the same person, in the same sitting — see Evidential Quality below).
Defect 1 — a dominant strategy makes the allocation decision trivial.
| Quantity | Value |
|---|---|
| Total iron available | 3/turn × 10 turns = 30 |
| Maximum useful spears | 15 (1 per settler; rules state the surplus is wasted) |
| Maximum possible raid strength | 12 (2d6) |
| Turn 15 spears is reachable | turn 5 |
| Earliest possible raid | turn 7 |
15 defenders beats any raid the rules can generate, and the player can bank 15 spears two turns before the earliest raid. Optimal play is therefore: **all iron to spears until 15, then all iron to ploughs, forever.** That is a step function with one computable knee, not a tradeoff curve. The tension recorded in the play log came from the player under-investing relative to the dominant strategy — not from a decision that was genuinely hard.
**Defect 2 — the "ploughs were never regretted" finding came from an instrument that could not produce regret.**
Harvest Score has no consumption, no threshold, no failure state, and no other system reads it. A player cannot regret building ploughs when a low Harvest Score costs them nothing. The original report treated the absence of a negative signal as the presence of a positive one, from a test structurally incapable of emitting the negative signal. This is the "meter that reads zero unconditionally" failure — the original text claimed this disproved the plough-as-trap risk. **It did not test that risk at all.**
Corrected verdict: PIVOT. The mechanic is untested, not disproven. The displaced-cost idea may well work; this prototype could not have found out. See PIVOT-NOTE.md for what to keep and what to change.
Hypothesis
If the player is forced to choose, weeks before any threat appears, whether the smith's iron becomes ploughs or spearheads — and that choice is the only cost of military readiness, with the muster itself being free — then when the horn sounds later, the player will feel the weight of their earlier choice.
Evidence sought: a player who under-armed reports feeling responsible for the shortfall (not merely annoyed at a game system), and a player who over-armed reports that it cost them something real (not "a decision with no downside").
Riskiest Assumption Tested
That the plough is a genuine, felt sacrifice rather than an obviously-inferior trap option that any sensible player ignores in hindsight. If ploughs carry no real value, there is no tradeoff — only "always build spears" — and the concept's central claim that the economy is the militia collapses into decoration.
This is the same risk the systems index flags against the Dual-Purpose Output System, and the same shape as the Manor Lords coupling trap recorded in the concept doc's Design Risks: a resource gate in front of a separate combat system still reads as bolted on.
~~Result: the assumption held.~~ WITHDRAWN — the assumption was never exercised. Harvest Score could not generate regret, so 'is the plough a trap?' had no way to answer yes. See Correction at the top.
Approach
A 10-turn paper abstraction of one season, played through once in full with the raid turn and raid strength rolled in advance and hidden from the player-side reasoning until they triggered.
Path chosen: Paper Reason for path: The hypothesis concerns whether an allocation tradeoff feels weighty, not whether an animation or input feels good. Paper resolves that question in minutes rather than engine-rebuild cycles, and the numbers (iron per turn, gear per defender, raid strength) can be retuned instantly.
Shortcuts taken (intentional):
- Single resource (iron); no wood, charcoal, or food chain
- Single gear type (spears); no mail, shields, or tiers
- Settlers are an undifferentiated count, not named individuals
- Ploughs abstract to a bare "Harvest Score" number
- One raid, one direction, one season
- No UI, art, or spatial layout of any kind
Result
~~The hypothesis was judged CONFIRMED.~~ WITHDRAWN — see Correction above. The observations below are recorded as they happened, but they do not support the original verdict: the decision the player agonised over had a computable right answer they had not worked out.
The moment it worked was turn 7 — the look-back. The raid landed one turn after the player had finally committed hard to spears (turns 5 and 6). The recorded reaction: *"If I'd known it was turn 7 I would have gone all-in two turns ago. This actually stings — I can see exactly where I under-invested."* The eight ploughs built in turns 1 and 2 became retroactively visible as a choice with a price, which is precisely the beat the hypothesis predicted.
The worst part was the opening. Turns 1 and 2 were consequence-free. "The horn will sound before winter" is too vague to generate pressure that early, so the first fifth of the season passed without the player feeling they were deciding anything. Felt pressure only began around turn 5 — roughly halfway into the raid's actual 7–10 window.
The muster resolved as a clean win (10 defenders against a raid strength of 7, no losses), but the player noted the margin was uncomfortable: *"I won because the roll was 7 and not 11."*
Metrics
| Metric | Value |
|---|---|
| Path used | Paper |
| Iterations to playable | N/A (paper) |
| Prototype duration | ~1 session (rules + one full simulated play-through) |
| Playtesters | 1 internal — also the rules author and the verdict writer (see Evidential Quality) |
| Feel assessment | Tension absent turns 1–4; sharp from turn 5; peak at turn 7 look-back. Win margin (10 def. vs 7 str.) read as luck-adjacent rather than earned. |
| Hypothesis verdict | ~~CONFIRMED~~ → INCONCLUSIVE (test admitted a dominant strategy) |
Recommendation: PIVOT
The displaced-cost premise is not disproven — it is untested. Two defects in the prototype's own construction meant the experiment could not have produced a negative result on either of the things it claimed to validate. A dominant strategy (15 spears, guaranteed win, bankable before the earliest raid) made the allocation decision computable rather than felt, and a consequence-free plough payoff made "was the plough a trap?" unfalsifiable.
The turn-7 look-back was a real and promising reaction, and the free muster drew no objection — those are worth carrying forward. But they were produced by a player who had not solved the arithmetic, which is not the same as a design that resists being solved.
Evidential Quality — read before trusting any finding above
One playtester. That playtester wrote the rules, ran the session, rolled the dice, narrated the player-side reasoning, and authored the verdict, in a single sitting. Every bias runs the same direction: toward confirming a hypothesis the same party had just spent a session building a case for. The dominant strategy sat in four lines of arithmetic and went unchecked.
Treat nothing in the original report as validated. A PROCEED verdict on a design hypothesis requires, at minimum: a tester who did not write the rules, and an explicit adversarial pass asking "what is the optimal play, and does it make the decision disappear?" before the session runs.
Carry Forward (whether pivoting or re-testing)
- Core tuning values discovered:
- A 10-turn season with a raid window in the final 40% leaves the first ~40% of turns tensionless. The threat signal must begin meaningfully earlier — target pressure by turn 2–3, not turn 5.
- 3 iron/turn against 15 settlers produced a meaningful shortage: the player reached only 10 spears for 15 settlers after committing from turn 3 onward. That ratio (roughly 1 iron per settler per 5 turns) is worth carrying into the Production Chains GDD as a starting point.
- Raid strength as 2d6 (mean 7, range 2–12) is **capped below the maximum achievable defence (15)**, which is the root of the dominant strategy. Raid strength must be able to exceed any reachable defence, or must scale with settlement progress, so "enough spears" is never a fixed memorizable number.
- Assumptions NOT confirmed (corrected 2026-09-14):
- ~~Displaced cost works.~~ Unproven. The look-back reaction was real, but it came from a player who had not computed the dominant strategy. Retest against a ruleset with no guaranteed-win threshold.
- ~~The dual-purpose output has genuine two-sided tension.~~ Untested. Harvest Score could not generate regret, so the plough-as-trap risk was never exercised. This risk is reinstated in the systems index.
- Weakly supported: a free muster drew no objection from the tester. This is the one finding not undermined by either defect, but it rests on a single self-run session.
- Assumptions disproved:
- That "the horn will sound before winter" is sufficient telegraphing. It is not. Pillar 3 promises readable, escalating threat, and this prototype shows vagueness produces dead time rather than dread.
- Emergent mechanics:
- **Uncertainty about raid timing is doing real design work.** The player's escalating commitment from turn 3 onward was driven by not knowing when, not by the size of the threat. A fully known raid turn would likely flatten this into a calculable optimum. The concept doc never names this; it should, and it needs reconciling with Pillar 3's telegraphing promise — the tension is between "the player can see it coming" and "the player cannot compute the exact optimal pivot turn."
- Post-raid turns (8–10) reverted to pure plough optimisation with no tension. Acceptable in a four-season game; worth noting for MVP session shape.
Next steps:
/design-review design/gdd/game-concept.md/gate-check/design-system telegraphed-threat-system— the system this prototype has the most to say about/design-system dual-purpose-output-system— carry the confirmed two-sided tension into Tuning Knobs and Formulas
Lessons Learned
- What assumptions were broken by actually building this? That threat telegraphing is a late-game concern. It is an opening concern: without early pressure the first turns are inert, and a season that starts inert wastes a fifth of its own runtime.
- What surprised us that didn't show up in the brainstorm? That when matters more than how much. The brainstorm and concept doc focused entirely on the size of the tradeoff (how much iron, how many spears); the prototype's tension came overwhelmingly from uncertainty about timing. That is an unnamed mechanic currently doing load-bearing work.
- What would we test differently next time? Tighten or remove the raid-strength dice before the next pass. The "I won because the roll was 7" reaction means variance is competing with the allocation decision for credit, and a paper prototype is the cheapest possible place to isolate that: run the same 10 turns with a fixed raid strength and see whether the turn-7 look-back still lands as hard.
Prototype code location:
prototypes/muster-mechanic-concept/
This code is throwaway. Never refactor into production.