Design Spike — Does Mail Fix the Dead Zone?
Date: 2026-09-15
Type: Solver spike, no playtest. ~30 minutes.
Question: The independent playtest found a dead zone — when population
falls, the spear cap falls with it, and iron has nowhere to go. Does adding
mail as a second military sink fix it without creating a dominant strategy?
Answer: **Partially. It fixes the dead zone. It has not yet been tuned to
avoid dominance.** Do not write the v3 ruleset yet.
What was tested
Five configurations of mail, each run 600–800 times against two opposing policies (buy spears first vs. buy mail first). Two things were measured:
- Survival — is the difficulty sane?
- Dead turns — turns where the player has iron and no useful sink. This is the metric the playtest actually generated; it is what "its kinda frustrating, i want more diversity in weapons and tools" feels like in numbers.
The detector was validated before it was trusted
A control run reproduced v2's exact conditions (mail priced out of reach) and reported 42.4% dead turns — roughly two turns in five with nothing worth buying, matching the tester's experience. Without that control, a low dead-turn reading would have been meaningless. An earlier version of the detector reported 1.5% on the same conditions and was wrong; it was counting legality rather than usefulness.
Results
| Mail cost | Absorb rule | Mail cap | Spear-first | Mail-first | Gap | Dead turns | Verdict |
|---|---|---|---|---|---|---|---|
| 3 | 1 loss each | population−2 | 47.2% | 75.2% | +28.0% | 4.9% | Mail dominates |
| 1 | 1 per 2 mail | tied to spear count | 28.0% | 32.7% | +4.7% | 37.0% | Dead zone returns |
| 2 | 1 per 2 mail | tied to spear count | 19.5% | 27.2% | +7.7% | 26.9% | Dead zone returns |
| 1 | 1 per 2 mail | 2×(population−2) | 60.3% | 86.1% | +25.9% | 11.2% | Too easy, dominant |
| 2 | 1 per 2 mail | 2×(population−2) | 41.7% | 49.3% | +7.6% | 2.5% | Closest |
| 3 | 1 per 2 mail | 2×(population−2) | 22.9% | 14.4% | +8.4% | 3.4% | Too hard, dominant |
Findings
1. Mail fixes the dead zone decisively — when its cap is independent of spears.
Dead turns drop from 42.4% (no mail) to 2.5% (mail at cost 2, cap 2×carriers). That is the playtest's complaint solved. Iron always has somewhere to go.
2. But tying mail's cap to spear count re-creates the problem.
Configurations where mail required a spearman to wear it returned dead turns of 26.9% and 37.0% — barely better than no mail at all. The reason is structural: when the population collapses, both caps collapse together, which is exactly the failure being fixed. Mail's cap must not depend on the spear cap.
3. No configuration yet avoids a dominant strategy.
Every one of the five has a gap of 4.7 points or more between buy-spears-first and buy-mail-first. The best case (cost 2, cap 2× carriers) still favours mail by 7.6 points. A second sink that has an obviously correct purchase order is a second sink, not a second decision.
4. The dominance flips direction with cost, which is the useful clue.
At cost 1 and 2, mail-first wins. At cost 3, spear-first wins by 8.4 points. The crossover is somewhere between 2 and 3 — but a cost of 2.5 is not expressible in whole iron, which suggests the lever is wrong. The fix is probably not price.
What to try next, in order
- Make mail situational rather than strictly better. Right now mail always absorbs a loss. If instead it only helped against certain raid types (a raid of archers vs. a raid of raiders), neither purchase would be unconditionally correct, and the player would be reading the threat's character as well as its size. This fits the concept doc's Pillar 3 — telegraphed, readable threat — and would give the Threat Forecasting UI something more interesting to show than one number.
- Give mail a different lead time. Spears take 3 turns; if mail took 1, it becomes the late reaction and spears the early commitment. Two sinks on different clocks are harder to collapse into one rule.
- Only then sweep cost again. Price is the last lever, not the first.
Method note
This spike cost 30 minutes and prevented writing a v3 ruleset that would have failed its own adversarial pass. The alternative — writing the rules first and discovering dominance in playtest — is what happened to v1 and cost a full session plus a withdrawn verdict.
Solver: mail-spike.py in this directory.