PROTOTYPE v2 - NOT FOR PRODUCTION

Question: With no computable safe threshold and a real cost on both sides,

does the iron allocation stay a live decision across a season?

Date: 2026-09-14

Predecessor: prototypes/muster-mechanic-concept/ (verdict PIVOT)

The Muster v2 — Paper Prototype Rules

The Question

The first prototype failed because 15 spears beat any raid and the plough payoff could not be regretted. Both defects are closed here. The question now is whether the allocation stays live — a decision the player keeps making, not one they solve once.

This ruleset passed an adversarial pass before it was written up. See adversarial-pass.py and the Solver Findings section. That is the step the first prototype skipped.

Setup

• 3 seasons of 8 turns (24 turns)
• 12 settlers to start
• Iron income: 0.5 per working settler per turn, rounded down at spend time. Mustered settlers and dead settlers produce nothing.
• The reserve: 2 settlers never muster and always work.
• Failure floor: below 3 settlers the settlement is abandoned. Run ends.

The Allocation (every turn)

Iron is spent on spears (2 iron) or ploughs (2 iron). Damaged spears can be repaired for 1 iron and repair takes priority over forging new.

Unspent iron carries over and does nothing on its own.

Lead Time — the rule that makes this a decision

A spear ordered this turn is ready 3 turns later. Iron committed to the forge is gone; the spear arrives when it arrives.

This is the load-bearing rule. Without it the player can wait to see the raid and buy spears on the spot, and a fixed ratio beats reading the threat. With it, being armed requires having decided three turns ago — and the solver confirms reading the threat then beats any constant policy by 8 points.

Threat — visible scale, hidden timing

From turn 2 of each season the player can see the season's **expected raid strength**:

expected = 2 + 1.5 × (season number, 0-based) + 0.4 × current population

The raid lands on a turn drawn secretly between turn 4 and turn 8. Actual strength is the expected value with noise (roll 2d6, treat 7 as "as expected", each point above or below shifts strength by about 5%; or in a digital version, a normal draw with 22% deviation).

Scale is knowable. Timing is not. The player can compute what they need but not when they need it, so the lead time bites.

The Muster

Free to call, as before. **Defenders = ready spears, capped at population minus the 2-settler reserve.** Mustered settlers stop producing for that turn.

Resolution

• Shortfall = Raid Strength − Defenders (minimum 0)
• If shortfall ≥ population − 3, the settlement is overrun. Run ends.
• Otherwise the settlement loses shortfall settlers.
• Wear: every raid damages max(1, defenders ÷ 2) spears, plus one per lost settler. Damaged spears return to the armoury and need 1 iron each to repair.

Season End — where ploughs bite

Each plough feeds 1 settler for the season. If ploughs are fewer than the population, the difference starves:

• If starvation ≥ population − 3, the settlement is abandoned. Run ends.
• Otherwise lose that many settlers.

A quarter of the ploughs carry into the next season (rounded down). The rest are consumed.

This is what makes a skipped plough regrettable. In v1 Harvest Score was a number nothing read; here it is the difference between a fed settlement and a funeral.

Solver Findings — read before playing

From adversarial-pass.py, 600 runs per policy:

No dominant constant ratio. Reading the visible threat estimate and arming to meet it beats every fixed policy. That is the decision the design wants.

A constraint the solver found: this only holds while the game is hard enough. Soften raid scaling and the adaptive edge collapses under 2 points — at 88% survival the choice stops mattering. The tuning band is narrow and the design should know it. Configurations that work: raid scaling k between 1.5 and 2.5 with a population coefficient of 0.4, or k=1.0–2.0 at 0.5.

What to Watch in Play

• Does the player look at the expected-strength number before committing iron?
• Is there a turn where they visibly regret a plough — or a spear?
• Do they ever hoard iron rather than commit? (Hoarding should feel bad.)
• At season end, can they say what the allocation cost them in both directions without being prompted?

Explicitly NOT Tested

Multiple raid types, gear tiers, named settlers, resources beyond iron, spatial layout, UI, the production-halt's moment-to-moment feel (paper cannot test feel).