Garrisons

How a garrison is put on the ground

Who defends an objective and how dangerous they are comes from the garrison profile — see Configuration. This page is about the other half: where those men end up standing, what they do while nobody is attacking them, and how they behave when somebody starts.

Groups are scattered around the objective rather than stacked on it. Each one is given a spot on clear ground, kept a minimum distance from the groups already placed, so a position reads as several covering each other rather than one large knot. The spacing is a preference rather than a rule: where the ground inside the spread cannot satisfy it, the group is placed anyway. Fielding the garrison matters more than spacing it.

Two checks stop a man being placed somewhere he cannot fight from. One tests that the spot is clear; the other tests through the middle of it, because a ring of clear ground can perfectly well have a rock or a wall standing in the centre of it. That second check is what stops the occasional soldier spawning inside scenery.

Settling in

The engine runs AI at reduced fidelity when no player is near them, which is what makes a large map affordable. The side effect is that a garrison built somewhere quiet does not move: it is still standing in the formation it spawned in when somebody drives up on it twenty minutes later.

So a new garrison is held at full simulation for a short window after the objective is built, long enough for its groups to walk onto the ground they were given, and then handed back to the engine. Approach during that window and you will find men in position; approach during it in a Game Master camera and you have woken them yourself, which is worth knowing before you conclude the setting does nothing.

Setting the window to zero switches this off entirely — nothing is ever held awake, which is the engine’s own behaviour and costs the least.

It is a real server cost while it runs, paid per objective, so pushing it far past a minute buys very little. If a garrison still looks bunched when you arrive, the answer is usually that the objective was built while you were already close enough to keep it awake anyway.

Holding, and patrolling

Not every group digs in. A share of a garrison’s groups walk a circuit around the objective instead of holding a position, which is what makes a place look occupied rather than staged.

Three rules bound that, and they are deliberate:

• At least one group always holds. However the share is set, somebody is on the position.
• No group patrols unless the garrison has two or more. A single-group garrison is a sentry post; a sentry post that wanders is just an empty objective.
• The share rounds down, so a small garrison digs in whatever the setting says.

A group that is holding is penned to a radius around its position — it will use that ground and take cover in it, but it will not chase somebody over the hill. A patrolling group is on a longer leash, because breaking off to fight is a reasonable thing for a patrol to do.

Being drawn off, and going firm

A patrol that chases a contact away from its objective is a group that has been removed from the defence for nothing, and a garrison that can be peeled apart one group at a time is a garrison that can be beaten by leading each of its patrols away in turn.

So a patrol drawn further than a set distance from its objective goes firm: it stops where it stands, holds that ground, and calls for help. Three things can happen next, and they are all normal:

• Help is sent, arrives, and the patrol goes back to its circuit.
• No new men may be fielded — because of the reserve or the concurrent ceiling — so a party already out in the field is re-tasked to it instead. That costs nothing against either limit, since those men are already fielded.
• Nothing at all can be sent. The patrol holds where it is until a timeout expires and then resumes its circuit on its own, so a group that called for help nobody could send does not hold the same patch of ground for the rest of the objective’s life.

A group already in a firefight is left alone rather than being ordered about mid-contact.

The distance that triggers this sits between the circuit a patrol walks and the leash it is kept on, so ordinary patrolling never trips it. Set it to zero and a patrol may chase as far as its leash allows.

Hearing shooting

A garrison does not have to lose a man to know it is being shot at. Its soldiers hear gunfire — rounds that miss, rounds that go past, rounds that hit nothing at all — and that is enough to put the objective in contact and start it answering back. Before this, only a confirmed kill registered, so a position that was being shot at inaccurately sat there and took it.

What the men can actually hear is the engine’s own model, not something this mod invents: an unsuppressed shot carries a few hundred metres, a suppressed one carries a small fraction of that. Firing on a garrison with a suppressor from a distance really does go unnoticed.

When shooting is heard from within investigating range, one patrol group is sent to look. Not the whole garrison, and never a group that is holding a position — the men on the objective stay on the objective. The patrol walks out to where the firing came from, holds there and observes for a while, then returns to its circuit. If the shooting is still going on when it gets back it will be sent out again shortly afterwards.

A garrison with only one group therefore investigates nothing, because it has no patrol. That is intended. The alternative is a sentry post that walks away from itself the first time somebody shoots near it.

Two things this deliberately does not do. It does not send a detachment — walking men off the position to search is a much bigger commitment, and that is reserved for a confirmed shooter, which Reinforcements covers. And it does not read a shooter who is firing from inside a vehicle, because the engine files the shot against the vehicle rather than the person in it.

Settings

These live alongside the rest of the mod’s settings — see Configuration for how to override them, including the note that booleans in the JSON file are written true and false.

A complete example

A garrison that spreads widely, keeps most of its groups on patrol, and investigates shooting out to half a kilometre:

{
  "m_fGarrisonSpreadM": 100,
  "m_fGarrisonSlotSeparationM": 30,
  "m_fGarrisonVehicleSeparationM": 30,
  "m_fGarrisonDefendRadiusM": 60,
  "m_iGarrisonSettleSec": 60,
  "m_fGarrisonPatrolFraction": 0.6,
  "m_fPatrolFirmRangeM": 150,
  "m_fPatrolReliefRangeM": 100,
  "m_iPatrolFirmTimeoutSec": 300,
  "m_bResponseHearsShots": true,
  "m_fPatrolInvestigateRangeM": 500,
  "m_iPatrolInvestigateSec": 120
}

Tuning it

If garrisons look like they spawned in a heap, the settling window is the first thing to check, not the spread. A garrison that was never held awake has not moved at all, and widening the spread only scatters the heap further out.

If nobody ever investigates, check the garrison actually has two or more groups at the tier you are playing. One group means one holding group and no patrol, and nothing on this page applies.

If patrols keep going firm during ordinary patrolling, the firm range has been set at or below the spread. It needs to sit above the circuit they walk, or they trip it on their own rounds and spend the objective’s life stationary.

Investigating range is not a difficulty dial so much as a distance one. Raising it does not make the garrison better at finding you — the patrol still walks to where you fired from and looks — it only means they will make the walk from further away, and be away from the objective for longer while they do it.