A grognard exercise, not a scientific model. What that means

Roadmap / 0.1.3.5

Province pacification set from research

Version 0.1.3.5, 17 Nov 2026. tag v0.1.3.5, commit 382117c. File docs/RESULTS_v0.1.3.5_pacification_research.md.
Written at this version and not revised since. Later versions may change these numbers; see the roadmap.

Switch: spatial.enabled.
New option value: pacification_calibration: research (now the default; median and window from v0.1.3.4 remain selectable).
New options: hard_provinces (on by default), classes (fix a province as hard, contested or accommodating), accommodate_prob (chance a province accommodates; off by default).
Runs: 3,000 paired runs per variant, every other module on. 30 tests pass; all harness checks pass (26, 26 and 22), including the check that the uniform split still equals the single-area model.
Scope: provinces as aggregates. No sites, routes or tactics.

Headline

  1. The open question from v0.1.3.4 is closed, and the answer is "slow". That version showed the result swinging by up to 10 points on how fast one province goes quiet. The research puts the speed near the slow end, and rules out the fast calibration. Across the whole range the research allows, US withdrawal in a full annexation moves by about half a point either way.
  2. Full annexation barely changes: 71.4% US withdrawal (71.7% in v0.1.3.4). Alberta detachment (61.6%) and coercion (34.4%) do not change at all.
  3. The reason is timing. No province can go quiet in its first 12 to 18 months, and the average war in the model is decided in 23. Ontario goes quiet at any point in only 5% of runs.
  4. What does matter is whether a province's leaders accommodate the occupier. In the Alberta scenario, every 10 points of chance that Alberta accommodates takes about 6 points off US withdrawal. This is an excursion, not the default: 75% of Albertans voted to remain in October.

What changed in the model

All figures are from the research report of 17 November 2026 (an analytical estimate from historical occupations, not a measurement).

Element v0.1.3.4 v0.1.3.5
Time for a contested province to go quiet, held at full strength median about 5 years (scale fitted to RAND's national figure) median 24 months, range 12 to 48, counted after the delay
Delay before any province can go quiet none 12 to 18 months from the first occupier arriving
Provinces that hold out none 20 to 35% of provinces are "hard": median 60 to 96 months
Provinces still fighting after 6.5 years same rate as before rate cut to a tenth
Effect of control on the rate fixed by the national formula exponent 0.5 to 2 (the evidence that force ratio matters is weak)
Resistance left in a pacified province 5 to 30% 8 to 20% (central 15%)
Relapse when resistance climbs past half its old level a monthly chance: 0.1 to 0.3% with the garrison kept, plus 2 to 4% scaled by the garrison shortfall, plus up to 1% with strong outside support
Accommodating provinces not modelled start quiet and do not relapse; off unless a scenario turns it on

A war confined to one province keeps the national rate, so the Alberta baseline is untouched.

Results: full annexation

Change is against v0.1.3.4's default. Intervals are 95%.

Variant US withdraws Change US deaths Gasoline rise Ontario ever quiet
v0.1.3.4 default (median calibration) 71.7% — 1,635 30¢ 5.0%
Research calibration (new default) 71.4% −0.3 (−0.6 to 0.0) 1,638 30¢ 5.3%
National pacification (no province effect) 70.4% −1.2 (−1.7 to −0.8) 1,649 30¢ —
Research, population first 72.9% +1.2 (+0.8 to +1.7) 1,206 45¢ 12.5%
Research, resources first 72.4% +0.7 (+0.3 to +1.1) 1,292 20¢ 6.4%

The allocation findings from v0.1.3.4 hold: concentrating troops cuts US deaths by about a quarter, resources first cuts the gasoline shock by a third, population first makes it half as bad again.

How much each research number matters

Change is against the new default (71.4%).

Variation US withdraws Change
Median 12 months (fast end) 70.9% −0.5 (−0.8 to −0.1)
Median 48 months (slow end) 71.6% +0.2 (0.0 to +0.5)
No delay 71.2% −0.2 (−0.5 to +0.1)
No hard provinces 71.2% −0.2 (−0.4 to 0.0)
Control matters more (exponent 2) 71.5% +0.1 (−0.1 to +0.3)
Control matters less (exponent 0.5) 71.1% −0.2 (−0.5 to 0.0)
Quebec fixed as hard 71.5% +0.1 (0.0 to +0.3)
Ontario and Quebec fixed as hard 71.6% +0.2 (0.0 to +0.5)
Alberta accommodates 71.6% +0.2 (−0.1 to +0.5)
Alberta and Saskatchewan accommodate 71.6% +0.2 (−0.1 to +0.6)

Nothing here moves the result by more than half a point. An accommodating Alberta does not help the US win the whole country (the troops it frees are 12% of the population's worth), but it cuts US deaths by 6% (1,539) and by 8% with Saskatchewan (1,508).

Results: Alberta detachment, if Alberta's leaders might accommodate

The baseline is unchanged. The table is an excursion: a chance that Alberta is quiet from the start.

Chance Alberta accommodates US withdraws Change Annexation consolidated US deaths
0 (baseline) 61.6% — 9.2% 308
10% 56.5% −5.1 (−5.9 to −4.3) 17.2% 287
20% 49.8% −11.8 (−13.0 to −10.7) 26.7% 262
50% 31.9% −29.7 (−31.4 to −28.1) 52.6% 192
100% 1.9% −59.7 (−61.5 to −58.0) 96.6% 72

This is close to arithmetic: an accommodating province is defined as quiet, so the war ends when it is taken. The table says how much rides on the assumption, not how likely it is. It is the largest lever the provincial work has found.

What it means

  • For the model: the province-level speed is no longer a free parameter with a 10-point swing. It is anchored, and the anchored range does not move the answer.
  • For the US: occupying province by province offers no shortcut on the timescale the war is decided on. The fast "ink spot" result in v0.1.3.4 needed provinces to go quiet in about nine months; the research says none do in the first year.
  • For Canada: what protects the country in this part of the model is not garrison arithmetic. It is that no provincial government treats with the occupier. That is a political question the model does not yet represent.

Limits

  • The research figures are analytical estimates from other countries' occupations. No case resembles a rich, urban, federal democracy occupied by its neighbour.
  • Resistance in the model starts at its level and declines or holds. The research says the commoner pattern in rich urban societies is a quiet start with violence growing later. Not modelled.
  • The research says violence concentrates in two to four regions. The model spreads resistance more evenly than that.
  • Provincial governments are not actors. Accommodation is a switch set by hand.
  • Hard provinces are drawn at random. Which provinces would be hard is not modelled, beyond fixing them by hand.

Next

Candidates, in the order I would take them: provincial governments as actors (so accommodation is an outcome, not a switch); resistance that grows from a quiet start; partial annexation as a war aim.