Tactics That Matter

Flanking, morale and routing, the unit triangle, terrain, and disengaging.

Numbers aren’t everything. Sparta’s battles are decided by where units are and which way they face — flanking, morale, and the rock-paper-scissors of unit types turn an even fight lopsided.

Flanking

Hitting a unit from the side or rear deals far more damage and morale loss than a frontal clash:

Table 1: Damage and morale multipliers by attack angle.

Attack angle	Multiplier
Front	×1.0
Side (flank)	×1.5
Rear	×2.0

Maneuver behind the enemy line. A regiment pinned frontally by a screen while your cavalry hits its rear will melt — and start a rout that spreads.

The enemy watches its own flanks too. An AI-controlled unit that finds a live threat closing on its side or rear turns to meet it — it won’t stand fighting head-on while your cavalry lines up an unanswered flank strike. Catching the enemy still costs real setup: bring the second threat into contact before it reacts, or pin the front so the flanking unit has nowhere better to turn.

Morale & routing

Units don’t fight to the last soldier. When a unit loses enough soldiers or takes flank/rear hits, it routs — it flees (its flag fading) toward its own back edge and stops fighting for now. A routing unit can flee past the field’s visible edge into a margin beyond it before it’s gone for good, staying on the field — and still fightable — for as long as it’s plausibly still in view. A unit breaks with its ranks as they stand: each man runs from his own place in the block, whichever way the block had turned before it broke. A unit that breaks while still in contact with the enemy, facing within about 75 degrees of its way out, swings round onto its flight no faster than its outermost men can follow, as it does on any such turn in its flight, so its men keep their places and run at once, not waiting for the turn. A turn of more than about 75 degrees is made at once, each man turning where he stands.

A routing unit runs faster than it marches. Unless it was already falling back, it starts its flight from a standstill (whatever broke it: a melee, missile fire, a neighbour’s rout), picks up speed over its first few seconds, and then flees at about 1.3 times its own top speed. Once clear of its pursuers it pulls away from troops of its own type; a pursuer that stays in contact with it holds it to that pursuer’s own pace. When a router slips in and out of a pursuer’s reach, the point its formation is laid out from moves once, from its front ranks to its whole block, rather than jumping between the two each time.

Figure 1: A broken cavalry squadron (bottom, faded) flees down the field while a fresh squadron of the same type sprints after it. The router picks up speed over its first few seconds, then pulls away at full flight, its riders running together as a block.

• Routs are contagious. A nearby ally watching a unit break is more likely to break too. One collapse can cascade down a line.
• Local force ratio, combat status & named morale ladder. Morale reflects local tactical context: being locally outnumbered, under unanswered missile fire (ambient incoming fire erosion halts at the shaken suppression floor at 50.0, shaking the formation without breaking it unaided, though separate casualty morale shock remains unfloored, and erosion halts entirely if the target can reply with its own ranged weapons), or losing local melee balance drains morale over time, while local numerical superiority steadies nerve. Unit combat status (Winning decisively, Winning, Balanced, Losing, Losing decisively) is derived from engaged physics and displayed on the Combat: HUD readout, alongside the 7-rung morale ladder (routing…berserk).
• This makes morale a weapon: you can win by breaking the enemy’s will before you’ve killed even half their soldiers.
• A routing unit is still fair game. Routing doesn’t grant immunity — a fleeing unit can still be struck and run down, chased as relentlessly as your pursuer can manage. So chase a broken enemy down to finish it, rather than assuming it’s out of the fight.
• A rout that runs through your ranks doesn’t spin them round. When a broken enemy flees straight through one of your blocks, your men keep fighting it as it passes, but the block keeps its facing and the block’s ranks keep their places instead of swinging round after the fugitives, so once they have gone each man steps back to his own place. A block riding a rout down from behind still turns to keep its front on it.
• A “broken” unit can rally if nothing catches it. As it flees, its nerve steadies — morale climbs back toward a baseline, fast at first and levelling off. If it breaks contact (no enemy close enough to keep hitting it) and still fields enough men, its recovering morale carries it over the rally threshold and it returns to your control, ready to fight again. The block fled with its ranks as they stood when it broke, so they are often turned against its new heading, and a half-turn leaves the short rank leading. It re-squares its ranks once it has pulled up (below), which for horse rallying at full gallop takes several seconds (about 5.5 seconds in the demo for this change). First the men turn where they stand, at the pace of an ordered about-face, by at most a quarter-turn, until the ranks lie square across the way they face, facing either way. So a rallied unit can end up facing across its line of flight: turning the ranks themselves onto that line would march a block that is not square into a different shape, and a unit only rallies clear of the enemy, so keeping its men in place matters more than which way it faces. An enemy can still close in while the unit pulls up. If one comes within reach before the ranks have re-squared, the re-square is called off and the unit turns to face the enemy with its ranks as they stand. Your next order squares them first, turning the men where they stand, and if you give none, the re-square resumes once every enemy has stayed a quarter beyond that reach for half a second. When the ranks do re-square, they do so from the places the men held. If its ranks face the wrong way, each man keeps his file, and only the under-strength files step up one rank to close the front, whichever the layout: File-major reform (the default), Row-major, or square (orbis or schiltron). As with an ordered about-face, the men don’t counter-march through the block to do it, so once it has pulled up a rallied unit is back in fighting shape quickly rather than spending seconds shuffling through itself. A unit that rallies at a run pulls up before it turns and re-forms, where it has the room and no enemy is near it or its path (and it stops short if one comes near while it slows): it brakes at its own rate with its men keeping step, over about 3 to 5 metres for foot running flat out and about 30 for horse, so the block halts as one instead of stopping dead and leaving its men to run on past their places and walk back.
• Keep pressing it and it “shatters” instead — still in contact when its time runs out, or gutted past reforming. A shattered unit does NOT vanish: it stays on the field, still routing, but can never again recover morale or rally. From there it either escapes (flees clear of the map, gone but got away) or is run down and annihilated (soldiers reduced to zero) before it can.
• A routing unit paths around impassable terrain too — fleeing is no longer a straight line toward the back edge; a router detours around a hill in its way the same as an ordered move does. It swings wide enough for its whole block to pass, turns onto the detour gradually, and its men veer off the rock rather than running into it, so even a deep block keeps together round the hill and regains its full flight pace once past it. (Currently, a block that breaks just short of a hill can still leave some of its leading men behind.) If every escape direction is blocked, it gives up fleeing and stands to fight instead of uselessly running in place against the obstacle.
• The battle isn’t over while a side is only routing. A team is beaten only once its last unit is truly gone (escaped, annihilated, or otherwise removed), not while its final unit is merely routing. Even a lone survivor that breaks can flee, rally, and fight on — so press a rout home rather than assuming a broken last unit is a win.

Figure 2: A broken unit recovering with nothing to interfere: staged already routing, with no enemy anywhere on the field, so pursuit and further casualties can’t factor in. Its nerve steadies as it flees, morale climbs back over the rally threshold, and it returns to your control. (The regression guard for this behavior is test_morale_recovery_demo_scenario.gd.)

Figure 3: The same recovery for a Row-major regiment. The flight leaves the grid flipped, with the 5-man short rank leading. A row-major block carries no per-soldier file to step up, so the rally deals the slots again file by file: each man keeps his file and his place in its order. The men of the five full-depth files stand still, and the men of the six short files each step one rank forward to close the front, so the short rank ends at the rear.

Figure 4: The same recovery once more, for a regiment holding square. A square keeps its own assignment of men to cells, which neither the file-major step-up nor the row-major pairing reaches, so undoing the flip has to name that third assignment directly. The rally deals the square’s own cells file by file, as for row-major: the men of the four full-depth files stand still, and the men of the four short files each step one rank forward, so the block re-squares where it stands instead of marching itself inside out.

Figure 5: Rout and pursuit: a low-morale infantry block is charged by cavalry and breaks — it flees, its flag fading, toward its own edge. The faster cavalry ride it down partway up the field, and the casualties from being run down keep its morale too low to ever recover: it’s annihilated rather than rallying — contrast the isolated recovery above (Figure 2), where nothing interferes.

Figure 6: A team’s last unit routs. A lone player infantry is broken and flees, its flag fading; the enemy keeps after it, since a routing unit is still fair game, and stays on its heels as it goes. The point: the battle stays live instead of ending the instant the last unit broke, for as long as it’s still on the field.

Figure 7: Routing beside terrain: a brittle cavalry troop starts just west of the impassable hill, every man clear of it, and meets a charge driven straight in from its western flank. Opening melee is per-soldier, so the troop fights for a while before it breaks; it then routs and flees north, drifting slightly east, past the hill’s west face, every man clear of it, with the charge’s files still among its own and riding north with it, its ranks slanting across its flight as it swings slowly round onto it, while its nerve steadies, rallying near the end of the clip well past the hill’s northern end. The clip does not show a detour around the hill. A router with no terrain escape route left at all still stands and fights instead of fleeing in place (see Terrain) — the open field this clip runs in never actually boxes a unit in, so that specific case isn’t shown here.

Rank cycling (trained units)

A well-trained melee unit periodically rotates its ranks — the tired front line steps back and a fresh rank takes its place, signalled by a whistle. Cycling slows fatigue buildup (by up to half at full training) and lets the unit recover morale even while still fighting. The better trained the unit, the more often it cycles. Ranged units don’t cycle.

Fatigue builds at a realistic pace — sustained melee wears a unit down over minutes, not seconds, and resting in the rear restores it just as gradually. So relief is a deliberate decision for a prolonged fight, not a frantic every-few-seconds chore.

Training therefore pays off in a long grind: a drilled regiment wears down a green one of equal numbers, because the green unit tires and wavers first. Rank cycling buys TIME, not immunity: the recovery it grants fades as the regiment thins out (fewer fresh files left to rotate to the front) and stops once it’s badly bloodied. So a veteran unit facing the same sustained pressure still routs eventually — it just takes noticeably longer and costs the attacker more losses to break it than a green unit under the same attack.

Rank relief (discipline toggle). Rank cycling is on by default for every unit, but you can switch it off with I (see How to play for the control). A unit with rank relief off holds its ranks fixed — no fatigue relief, no in-fight morale recovery from cycling — which trades away that endurance edge for a line that never opens the small gaps a whistled rotation briefly creates. Leave it on for a grinding attritional fight; consider switching it off only when you’d rather keep the front rank locked in place than let it step back mid-melee.

Line relief

Rank cycling rotates soldiers within a regiment; line relief rotates whole regiments. Right-click an engaged friendly with a fresh unit selected: the fresh unit advances through its tired ally and takes over the melee, and the relieved regiment falls back toward its own rear — where fatigue drains away and morale recovers (see How to play for the controls).

• The swap is a regiment pass-through, with a gated file lane. For its duration the two friendly regiments may interpenetrate, so the line never opens a company-sized hole for the enemy to pour into while the fresh unit works its way forward. Tight files stay packed: 0.45 m synaspismos on foot and 1 m knee-to-knee on horse leave a derived side gap of ~0, so a partner cannot march between files — blocked relief through a locked Tight interval is intended. A block that already has open file clearance (0.9 m Normal or 1.8 m Loose on foot; 2 × 6 m or 4 × 12 m cavalry) widens a lane — and only in the ranks the partner is actually marching through: ranks it has not reached yet stay closed, the lane opening a couple of ranks ahead of the pass rather than fanning the whole block the moment the order arms.
• The retreat is an order like any other. The relieved unit falls back under a normal move order (hold Space to see it), and you can redirect it — to a safer rest spot, or straight to another fight if you must.
• Interrupting the reliever cancels the swap. The swap lives on the reliever’s relief order, so a fresh order mid-pass ends it where it stands and the pair stops ignoring each other’s bodies. Avoid changing your mind while the two regiments still overlap.
• The relieved regiment backs out facing the enemy. It withdraws as a fighting body — shields still toward the line it just left — instead of turning round and marching off, so the ranks stay square the whole way to the rear and are ready to turn about and go back in (Figure 8).
• Expect to bleed a little on the way out. The enemy keeps swinging while the retiring ranks are in reach, so pulling out of contact costs a few men — far fewer than staying until the regiment breaks. Backing out fronted on the enemy is what keeps that toll small: the men taking those last blows are the front rank, behind their shields, not a turned back.

Figure 8: The relieved regiment’s fighting withdrawal: it falls back to its own rear with its heading held on the enemy, ranks square, rather than pivoting away.

Rotate before the break, not after: a tired regiment fights measurably worse (see Stamina) and routs sooner, and pulling it out while it still holds banks a rested unit for the next push.

Passage of lines

Line relief swaps one pair of regiments at a time by default — every fresh unit you order onto an engaged friendly swaps with (or, past the first, piles onto the fight of) that ONE unit you clicked. When a whole tired line needs relieving at once, select every fresh unit in your reserve, switch the group-order mode to Distributed (X, or the Spread button — see How to play for the control), then right-click any one engaged unit in the tired line. Each fresh unit pairs with its OWN nearby engaged friendly instead of all of them swapping with the single unit clicked, so the whole reserve passes through the whole tired line in one order — Polybius’s triplex acies, a fresh line relieving a spent one through gaps opened in the line behind it.

• The candidate pool is engaged friendlies only. A reliever never swaps with an idle or already-routing unit just to keep busy — if fewer of your tired line are actually fighting than you have fresh units, the extras fall back to piling onto a fight instead of manufacturing a swap.
• Proximity picks which tired units are relieved; position along the line picks who relieves whom. The engaged friendlies nearest the unit you clicked are relieved first, one per fresh unit. Within that set, both lines are ordered along the tired line’s own length and paired flank to flank, so each fresh unit relieves the unit in front of it and no two pairings cross. If you have more fresh units than tired ones, the extras double up on the neighbouring tired unit rather than crossing the line.
• Every pairing is a full line relief, with the same pass-through exemption, the same real retreat order, and the same “expect to bleed a little on the way out” cost described above — distributed relief only changes how the targets are assigned, not how any one swap behaves.

Reinforcement insertion

Three ways to combine two regiments, and they answer different questions. A merge (M) pools two selected regiments into one on the spot: the extra men drop onto the tail of the primary’s grid, instantly. Line relief and passage of lines (above) swap a fresh regiment for a tired one: the headcount in the line never changes, only who is holding it. Reinforcement insertion is the missing case — the line’s headcount grows and the newcomers arrive on foot: a reserve marches up behind a friendly host and its men file into the host’s ranks, so the host’s own men keep their places relative to one another while the block doubles. Asclepiodotus called it doubling by number, as against the file-doubling drills (explicatio and duplicatio), which double a regiment’s own frontage by place without adding a man.

Figure 9: A 40-man reserve files into a 40-man host by files: select the reserve, press Shift + M, right-click the host. The reserve marches to a rendezvous just behind the host, then its files step forward into gaps the host’s own files open as they spread, and the host goes from 9 files to 18 with its front rank holding its ground.

• Select the reserve, arm with Shift + M (files) or Ctrl + Shift + M (ranks), right-click the host. The arm is one-shot, like a stance arm: it applies to the next right-click on a friendly outside the selection and clears on Esc. The host can be idle or already fighting; it keeps whatever order it holds and its fight goes on.
• Two axes. Insertion by files (Shift + M) widens the line: the host’s frontage grows by the inserted files, up to double, and the front rank does not move forward. Use it on an engaged line that needs to overlap the enemy. Insertion by ranks (Ctrl + Shift + M) deepens the block: frontage holds, each inserted rank steps in between two existing ranks, and the front rank holds its ground as the anchor steps rearward. Use it on a column before contact for the push.
• The order is refused, with the reason flashed on the HUD, when the two are on different teams, either is routing or dead, the reserve is itself in contact (a body in a fight cannot file off), the loadouts differ, the host is squared or too far off to hold files, the host reflows row-major (cavalry and undisciplined foot have no files to interleave into), or the host is drilled in file groups (subunit-aware insertion is not yet wired).
• It costs cohesion. The host starts at the same strangers debuff a merge pays and recovers at the usual rate; Asclepiodotus notes that some condemned doubling with the enemy near, and preferred to fake the widening with light troops on the wings rather than disturb the phalanx. A host reshaped while moving faster than a walk also pays the moving-reshape penalty.

Acies triplex

The Roman acies triplex deployed infantry in three lines — hastati, principes, triarii — with gaps between maniples in one line roughly matching their own frontage width, and the line behind offset to cover those gaps: a checkerboard (quincunx) pattern. Passage of lines (above) builds the tactical half — swapping a whole tired line for a fresh one through gaps. Checkerboard form-up builds the spatial half: a fifth multi-unit form-up split mode (off by default; enable it in ☰ Menu → Form-up — Checkerboard (quincunx), or cycle to it with Y once enabled) that lays your selection out as two staggered rows instead of one straight line. See How to play: Controls (the “Deploy several units at once” entry) for the drag mechanics.

A unit dragged into a checkerboard arrangement remembers which line it landed on, the way it remembers its stance or formation mode: the front row is Line 1 and the offset row behind it is Line 2. The deploy order carries that assignment, so a saved replay reproduces it on playback. Moving the unit’s card to another tray row afterwards re-tags it with the new row – live only, since a card move isn’t an order and so isn’t recorded in the replay. Nothing reads the tag yet, though, so the layout still behaves as a one-shot arrangement in play — and nothing yet models the historical system’s chief weakness either: a gapped, multi-line formation trades frontage flexibility for flank exposure, so a checkerboard-formed army isn’t currently more vulnerable to flanking than a solid line. See docs/acies-triplex-design.md in the project repository for the full design note and what’s tracked as follow-up work.

Tray formation

The unit card tray can author the same kind of gapped, multi-line layout by hand: rows are ranks (Line 1 at the top of the panel is the front rank), columns are files left-to-right, and an empty cell is a literal gap. Check Tray formation placement in the tray header, select the units, and right-drag a flank line – the drag sets the front line’s position and facing the same way every other form-up does, and the tray grid (gaps included) places every unit relative to that line.

This is not the Checkerboard (quincunx) Form-up split. Checkerboard ignores the tray and alternates the selection by index into two computed rows. Tray formation uses the grid you actually laid out, so a three-line hastati / principes / triarii stack with offset gaps is expressible, and a one-line filled tray still only reorders (then splits by whatever Form-up mode is live). See How to play: Deploy from the unit card tray for the gesture.

A tray deploy carries the same line tag a checkerboard one does: each unit remembers the row it was placed from (the top row is Line 1), recorded on the deploy order so a replay reproduces it. The column is not carried, so a unit doesn’t remember its exact cell – and, as with checkerboard, nothing reads the tag yet, so the arrangement itself is still one-shot in play.

Figure 10: A tray-authored quincunx: two units on the front rank with a gap, one unit behind covering that gap – the same spatial idea as checkerboard form-up, but the gap is a cell you left empty, not an alternating-index rule.

Echelon (oblique order)

Every form-up split covered so far keeps the whole selection on one straight rank. Echelon breaks that on purpose: it staggers your line into a diagonal, one flank stepped forward and the other held back. Caesar deployed this way at times in the Gallic War, and Frederick the Great refined it into his signature oblique order at Leuthen — weight one flank, refuse the other, and let the enemy fight an uneven battle instead of a fair one.

Two more multi-unit form-up split modes carry it, both off by default (enable one in ☰ Menu → Form-up, or cycle to it with Y once enabled):

• Echelon (right leads) — the unit nearest your drag’s right flank advances furthest, the left-flank unit trails furthest back, and everyone in between steps evenly across the gap. Concentrate the leading (right) flank on a weak point in the enemy line while the trailing (left) flank stays out of easy reach.
• Echelon (left leads) — the mirror image: the left flank leads, the right flank refuses.

See How to play: Controls (the “Deploy several units at once” entry) for the drag mechanics — echelon reuses the exact same box-select-then-drag gesture every other split mode does, just with a staggered depth instead of one flat line.

Unlike checkerboard and tray formation, echelon assigns no line at all, leaving each unit’s existing line tag alone. The stagger itself is a one-shot layout: a unit dragged into an echelon doesn’t remember which flank it led from, so a later reform or reshape doesn’t preserve it. And nothing currently models the tactical risk a real refused flank carries either — the trailing units aren’t rendered any more exposed to being turned than a straight line would be, so the payoff (concentrating force where you choose) isn’t yet balanced by the historical downside (a flank that isn’t in the fight yet is a flank an enemy could exploit before it arrives).

Stamina

Every soldier carries a stamina pool that drains through combat and refills at rest. Each blow thrown costs stamina; throwing blows costs more than absorbing them. Rising from the ground after being knocked down costs a burst extra. Low stamina lowers the soldier’s condition factor, which scales both attack output and defence — a tired soldier hits softer and is easier to hit.

So a unit that has been grinding at the front for a while is measurably weaker than one that just arrived fresh. Rank cycling (above) helps by rotating tired soldiers to the rear to recover; simply pulling a unit out of melee and letting it rest works too.

The pace of the march spends stamina too. Every soldier’s pool follows the regiment’s live pace, not just its blows: a halted regiment regenerates, a walk is neutral, a jog drains slowly, and a sprint drains fast. So a regiment that jogs its whole approach arrives sooner than one that walks, but arrives with less in the tank, and one that sprints the whole way arrives spent — its condition factor already down before the first blow lands. That is the trade-off behind choosing a gait by click count and behind AUTO pace’s default profile: the unthreatened walk costs nothing, so the men reach charge distance fresh and spend the pool only on the terminal sprint (or on an earlier jog if under missile fire or pursuing). Read the pool off the selected unit’s stat sheet (the Stamina per man line, mean and spread over the living men) to see how much of it a march has already cost.

Figure 11: A lone Infantry regiment jogs a long leg south, then is re-ordered to sprint mid-jog. The stat sheet’s Stamina per man readout bleeds slowly through the jog (about one point a second) and steeply through the sprint (about five a second), with no rest in between to mask the difference.

Prone / knockdown

A blow that delivers enough force can knock a soldier off their feet. A prone soldier has no active defence — only their armour absorbs incoming hits — and must spend both time (about two seconds) and stamina to get back up.

Visually, prone soldiers stand out:

• At normal zoom (mark LOD): a prone soldier’s mark is flattened into a wide horizontal sliver so it reads clearly as “lying flat” among the standing marks.
• Zoomed in (figure LOD): the figure silhouette is rotated 90° to show a person lying on their side.

Engaged lines resist knockdown. A unit that is actively locked in melee (and not in Skirmish stance) braces automatically, raising the impulse threshold needed to knock a soldier flat. A skirmishing or unengaged unit does not brace and is easier to topple. A deeper file also helps — more soldiers absorbing the same front-rank pressure means fewer individual soldiers pushed past the knockdown threshold.

Motion weakens footing. A soldier caught moving or running has kinetic footing rather than a settled static stance, lowering their anchoring capacity by 20% (\(\kappa = 0.8\)). A stationary, settled rank resists being bowled over far more reliably than troops struck mid-stride.

A cluster of prone soldiers in the front rank is a warning sign: your line is tired or outmatched there, and each fallen body briefly opens a gap in your defence before they can rise.

Collision damage

A hard, fast impact — not just a landed blow — can wound. When two soldiers collide at real speed (a charge slamming into a line, not just two ranks pressed together), the force of the collision itself deals damage on top of whatever a weapon lands. A gentle, slow-closing press between two locked lines causes no extra harm this way — it takes genuine speed behind the hit.

The lighter or less-braced side of a hard collision takes the worse damage, because damage tracks the same velocity jolt as the shove itself: a braced, heavy line is already shoved less by the same impact, so it also takes less of a jolt to the body — and less jolt means less damage. This is part of why riding a charge into a set line is so costly for the attacker if the charge doesn’t break through cleanly — the impact itself hurts both sides, and the braced defender takes less of it.

Line width (frontage)

You set how wide each regiment fights — [ and ] step the width by a file symmetrically, or drag a selected unit’s flank grips (the drag anchors the far flank, so the line extends toward the grip you pull — handy for stretching a line toward the flank you need to cover without giving ground on the other side). The same soldiers spread across a different shape:

• Wide and shallow puts more men in the front rank, so more of them are swinging at once — good for overlapping an enemy’s flank or covering ground so a foe can’t slip around your line.
• Narrow and deep stacks more ranks behind a thinner front. Fewer fight at any moment, but the column has more bodies to feed forward as the front falls — it holds a chokepoint or punches through one spot rather than spreading thin.

A wider line is easier to outlast in depth and quicker to crumble once it bends; a deeper one trades frontage for staying power. Heavy infantry types with a declared subunit structure (such as Spearmen with 16-man files) derive their default full-strength width directly from that subunit size (e.g. 9 files for 140 men) with HUD readouts indicating file count and file capacity (Figure 12). A width you’ve set by hand (the bracket keys or a flank drag) holds exactly as you left it — the game never overrides your explicit choice.

Figure 12: A 140-man Spearmen regiment formed up into 9 files (16 ranks deep) in drill mode. The default frontage derives directly from declared subunit size (16 men per file, the historical lochos) rather than the generic roster-wide aspect ratio.

Cavalry regiments field a square/oblong squadron block rather than line-infantry files (Figure 13). An 80-man cavalry regiment fields 9 columns (aspect ~1.0) operating as a cohesive structureless squadron block with row-major reflow instead of line-infantry file tracking.

When such a block changes its width, whether you widen or narrow it or it closes its ranks by itself after heavy losses, each rider steps into the new place nearest where he stands, so nobody crosses the block to get there. When a rider falls, the riders behind him in his file step up, and the rear rank fills the back of that file, so the squadron’s slots keep their shape at once, though the riders can take a while to walk up to them. The same goes for any foot regiment set to reform row-major.

Figure 13: An 80-horse Cavalry regiment formed up into a 9-column squadron block in drill mode. Frontage derives from squadron geometry rather than line-infantry files, operating structureless with cohesive row-major reflow.

When a file-grouped unit reforms to a narrower or wider frontage, the reform operates at the subunit level rather than as a whole-block relabeling (Figure 14). Retained files keep their existing soldiers and internal ranks without criss-crossing the formation; only the flank file peels off to march into the rear ranks of the remaining files.

Figure 14: An 80-man Spearmen regiment (5 files of 16 men) narrows frontage by one file to 4 files of 20 men in drill mode. The 64 men in files 0–3 stay in their exact files and ranks without displacement, while the 16 men from the folded flank file march into the rear ranks.

Close the ranks. Left on auto, though, a regiment that’s lost half its strength narrows its own frontage one notch, trading width for depth so the survivors reform into a smaller, still-solid block rather than stretching a nearly-spent line across its original width. It widens back out again if the regiment claws back above about two-thirds strength (reinforcement, a relief, a lull in the fighting) — but not the instant it edges past halfway, so a unit trading casualties right around the line doesn’t flicker between shapes fight to fight.

File-doubling for a fast reshape. The [ / ] steps are fine for fine-tuning, but two drill maneuvers swap the whole shape in one press. B (explicatio) doubles the frontage and halves the depth — a column unfolds into a wide line to overlap a flank or plug a gap. N (duplicatio) does the reverse: it halves the frontage and doubles the depth, folding a broad line into a deep column so you can squeeze down a lane or mass depth to punch through one point. The men walk to their new slots rather than snapping, so there’s a beat of disorder while the block reshapes — don’t start the maneuver in contact, and give it room before the enemy arrives. A natural rhythm is duplicatio to march a column through a chokepoint, then explicatio to deploy back into line on the far side.

Deploy from column with one press: turn and unfold. Shift+Q / Shift+E chain a quarter-turn and an explicatio into a single drill order. A column marching by files quarter-turns to bring its long side to the front, and the instant the turn completes the block doubles the frontage it now presents — the explicatio is measured against the new facing, so a 5-wide, 8-deep column turned left becomes a 16-wide, 3-deep line facing the flank it turned toward. Reach for it when a threat appears off the side of a column on the march: one order both faces it and puts the whole regiment on a broad front, without the two-keypress gap in which a half-turned column is at its most awkward. Pressing Q and then B by hand is not the same drill — the plain explicatio doubles the grid’s old files, which after a quarter-turn run along the block’s depth.

Everyone shifts to the file nearest him. Changing the width renumbers every file, so in principle each man could end up anywhere in the new grid. He doesn’t: the block keeps its left-to-right order through the reshape, so a man on the right flank takes a file on the right, and he holds his place in his own column’s depth as well. The block widens or narrows by opening and closing between the files rather than filing men across itself, which is both quicker and far tidier to watch. Measured on a lone regiment drilling with no enemy near it, that cuts the ground covered in a reshape by about a third, and the men walk to their new places without threading through each other on the way (Figure 15).

One caveat worth knowing if you are watching closely: this tidier dealing is chosen at the moment the frontage changes. A regiment that then takes casualties in a melee keeps the places it was dealt, but a fresh reshape is what re-chooses them — the block does not continuously re-sort itself under fire, which would be its own kind of churn.

Figure 15: A lone Spearmen block widening by three files, with no enemy on the field. The frontage change renumbers every file, so each man could in principle be sent anywhere in the new grid. Instead the block keeps its left-to-right order and each man takes the file nearest him, so the formation opens outward and settles rather than filing men across itself.

Anchored explicatio for a pinned flank. The plain explicatio widens about the unit’s own centre, so both flanks drift when the frontage doubles — fine in the open, but a problem when one flank is braced against a hill, a river, or a neighbouring unit that can’t shift with it. Shift+B widens while holding the right flank fixed; Ctrl+B holds the left. The whole widen lands on the free flank instead, so the pinned side never moves and the line still gets the extra frontage where there’s room for it.

How a line reforms after losses. By default (File-major, the unit settings panel’s File-major reform button), every soldier holds a persistent file (column) assignment: when a man falls, only the survivors behind him in his OWN file step forward to close that one gap, and every other file is untouched — nobody switches columns just because a neighbouring file took a hit. A mauled unit’s rear reads as jagged, not uniform: a hard-hit file can run noticeably shallower than one that’s taken no losses at all, side by side in the same rank of survivors. The front rank still keeps its full width for as long as any file has a man left to hold it, and the block never bulges wider than the frontage you set — only the depth, file by file, tells you where the losses landed.

Click the button through to Row-major to keep every rank full instead. When a man falls, the men behind him in his file step forward to close the gap, and the short rear rank sends a man forward to fill the back of that file, so the depth shrinks rank by rank, uniformly. The rear rank’s survivors close up toward the centre — the wings shorten first while the middle files stay deepest — rather than leaving a ragged, gap-toothed back rank. Every other man keeps his place or steps one place over, and nobody crosses the block to fill a gap.

A third click cycles to Auto, which resolves to File-major or Row-major per unit from its own disciplined flag rather than a fixed choice — a disciplined, professional regiment reflows File-major (it holds its ranks under loss), an undisciplined mob reflows Row-major (it just closes up whatever gap the loss opens). Useful for a mixed army: set the whole selection to Auto once, and each unit reforms the way its own training would suggest, rather than picking File-major or Row-major by hand per unit.

Deploy a whole battle line at once. Select several regiments and drag one flank line (right-click and drag): the line splits among them. By default it splits to equal depth (space), so every regiment forms up to the same physical depth — a uniform front where bigger (or denser) units just take a wider slice. That keeps your whole line’s staying power even, with no thin spot for the enemy to punch through. Press Y (or Shift + Y to cycle the other way) to switch modes — equal width when you’d rather every unit cover the same frontage (small units stack deeper), or one of the count-based variants when you want density differences ignored entirely (see Deploy several units at once in How to Play for what each of the four modes holds equal). They sort left-to-right by where they already stand, so nobody crosses the field to reach a slot — a fast way to dress a ragged group into one clean front before the enemy closes. Hold Shift to place them in selection order instead when you want a specific regiment on a specific flank.

Use ☰ Menu → Form-up to configure the split: the upper radio buttons set the mode each battle starts in; the lower checkboxes (“Y-key cycles through…”) control which modes Y steps between mid-battle. Uncheck a mode to skip it entirely from the cycle.

To deploy a multi-line layout you drew in the unit card tray (ranks, files, and empty-cell gaps), check Tray formation placement instead – see Tray formation. That path uses the tray grid, not the Y-key split.

The unit triangle (rock-paper-scissors)

Three roles, each strong against one and weak to another:

🗡️ Infantry

The reliable line. No special bonus — your anvil to pin the enemy in place while something else does the killing.

🐎 Cavalry

Fast — they charge at a hard canter (~8.5 m/s, far quicker than any foot unit’s advance) — with a charge bonus that grows with closing speed. Devastating into a flank or rear, but blunted when it charges into braced spears.

🛡️ Spearmen

Anti-cavalry (marked with a spear). They blunt the cavalry charge, so use them to screen your line against enemy horse. Their long spears also out-reach shorter weapons, landing the first blows as lines close (see below).

The loop: cavalry beats exposed flanks → spearmen beat cavalry → infantry pin so your cavalry can get around to those flanks. Read the enemy’s screen before you commit your horse.

Formation stances

Press T to cycle a unit’s density — named on the control bar by the live slot-center pitch in metres (the interval includes the man or horse). On foot that is 0.9 m (pyknosis, the default fighting order), 0.45 m locked (Tight: synaspismos with better missile and charge defence), 1.8 m (true open order), and 0.45 m Square (the anti-cavalry ring, covered below). Shift + T cycles the other way, handy for stepping back one density without looping all the way round. Two heavier shielded close-order stances pack below the 0.45 m Tight floor and are set from the control-bar Formation menu or their own direct keys. Cavalry uses the same identity with a different Tight preset (1 m files × 3 m ranks; Normal 2x, Loose 4x).

Density also decides how deep an enemy can press into the line once melee is joined. Tight, Square, Schiltron, Shield Wall, and Testudo hold the enemy at the front rank with essentially no depth-wise mixing — shields lock with the men either side, so one soldier going down doesn’t open a seam. Normal order holds a couple of ranks of the line before it starts to give; Loose order has no such discipline, so soldiers there can end up more deeply mixed with the enemy.

Two infantry blocks that meet squarely front to front fight along one line. They come into contact as their front ranks close to within weapon reach, and the press stops there, so each block’s ranks stay behind its own front rank instead of being drawn in among the enemy’s. Blocks that meet only partly face to face, offset to one side, still press some way in. A square faces every way, so a line attacking it also stops at its outer rank. An infantry block that attacks another’s flank or rear stops the same way, at the defender’s end file or rear rank, and fights it from outside the block. A cavalry charge and a fight with missile troops are still measured centre to centre, so there the attackers can end up among the defenders’ ranks.

Figure 16: An infantry block attacks another block from behind. It comes into contact as its front rank closes to within reach of the defender’s rear rank, and the press stops there: it fights the defender’s rear rank from outside the block, and neither block’s men end up inside the other.

When a cavalry troop rides so far into a foot block that the troop’s centre is inside the block’s ranks, the block keeps its ranks where its men stand while it fights the troop. Its men still turn to face the riders, but the formation does not wheel round its centre after them, so no man is sent across the block to a new place in the middle of the fight. This does not apply to a troop on the cycle-charge stance, which rides in and out again by design, or to riders whose centre is still outside the block. Once the block is out of contact, it turns its ranks back onto its heading by the amount it turned toward the riders, no faster than the men can follow, without waiting for an order. Only that part is undone: ranks it had already turned in place on another enemy stay as they are, so a block whose ranks stand near a quarter-turn off is not swung round by a half-turn to straighten a few degrees. The ranks only ever turn toward square with the heading, never away from it: if turning toward the riders had partly undone a turn made earlier on another enemy, undoing it would turn the ranks further off, so they stay where they stand, and ranks that come square before the whole turn is undone stop there. Ranks you had drilled a quarter-turn round count as square at that quarter-turn, whether you drilled them before or after the riders came in, so they turn back onto it, and once back on it they count as drilled again, so riders inside the ranks in a later fight are undone onto the same quarter-turn. Ranks that cannot be brought nearer that quarter-turn stay where they stand, and no longer count as drilled: the next order squares them as it would any other turn. If it comes back into contact before it has finished, it stops and finishes once clear again. If it turns its ranks in place onto another enemy before it has finished, or a sudden large turn is folded into its ranks, the rest is left as it stands rather than undone on top of that turn, and only an order squares it. An order given in the meantime squares all of its ranks to its new heading once the block has been clear of the fight for half a second. Ranks turned more than a quarter-turn off are first reflected in place by a half-turn, and the rest of the turn is made no faster than the men can follow, except that ranks left exactly a quarter-turn off are dealt out in one step, and so, at the quarter-turn, are ranks within 10 degrees of it when the order comes to a block that is not marching (described with re-facing to fight, below). A cavalry troop with foot or other horsemen among it still wheels its formation to follow its enemy.

Figure 17: A brittle cavalry troop charges an infantry block standing on Hold and pushes into its ranks. The infantry turns to face the riders, but its ranks stay where its men stand rather than wheeling round after them.

Figure 18: An infantry block on Hold first snaps round toward a cavalry charge, then turns back as the riders come inside its ranks, until its ranks are square again. Once the riders have broken and gone, its ranks stay square rather than wheeling back off its heading.

Stopping at the edge makes a flank or rear attack cheaper for the attacker, since the defender’s men facing away from it cannot strike back. In a test of 120 men against 120 in tight order, over the first 19 seconds, an attack on the rear cost the attacker no men and an attack on a flank about 3, where pushing in among the defender’s ranks had cost it about 11 and 18. The defender lost a little fewer than before (about 39 instead of 46 in the rear, 18 instead of 20 on a flank), because only the attacker’s front rank now reaches it.

A block that has broken and is fleeing is the exception: troops that catch it from behind or on a flank still push in among its ranks, and keep cutting it down. In a test where a block of 120 stood between two enemy blocks, one in front and one coming on behind, then broke and fled past the one behind, it had lost about 22 men 15 seconds into the battle, against about 14 before this change; had the pursuers stopped at its rear rank, it would have lost about 9. The extra 8 fall in the pursuit, but what costs them comes before it: the block behind now meets its rear rank sooner, without walking in among its ranks first, so its men lose heart sooner and it breaks half a second earlier, before losing a man, while it is still inside the trap. With that earlier contact taken away, and the pursuit left as it is, it would have lost about as many as before the change.

Because only the front ranks reach each other, a frontal fight between two such blocks is slower to kill than it used to be. In a fight of 120 men against 120, each in tight order and neither breaking, about a fifth fewer men fell in the first 19 seconds than when the blocks pressed into each other. The block standing its ground lost about half as many as before, while the block that advanced lost about the same. Units therefore take longer to break in a straight-up frontal fight.

An enemy can still end up inside a block, past the ranks that are fighting: a lone man or a small party cut off in the press, two blocks that start a battle on top of each other, or a block overrun by a cavalry charge or a pursuit. He is not safe there. Only a block’s front few ranks normally strike, but once the centre of an enemy unit’s men has got inside a block, past its fighting ranks, any man of the block with one of its men within reach strikes him too, whatever rank he stands in, so a few men who have broken into the middle of a block are cut down quickly. An enemy unit fighting a block from outside, or from among its fighting ranks, is not inside it, even where its leading men are pushed in past those ranks. Front to front, the ranks in contact are never quite level, and in Loose order the two lines mix, but the fighting ranks deal with the men pushed in among them, and the ranks behind do not join in. An attack on a flank or the rear stops at the block’s edge, so there too the men it presses on fight it as before. So none of this makes a straight frontal fight, or an attack on a flank or the rear, any deadlier: two blocks of 120 meeting front to front, in Tight, Normal or Loose order, or one attacking the other’s flank or rear in Tight or Normal order, fight exactly as they did without it. Routing troops are left out of this. A routing enemy caught inside a block is struck only by its fighting ranks, and the men of a routing block never turn on an enemy inside it. But a block that breaks, flees and rallies with its pursuer pressed in among its ranks has that pursuer inside it, and strikes back: in a test where a block of 120 broke in front of another block of 120 and rallied, the pursuer lost about 30 men in the first 40 seconds where it lost about 25 without this.

Only the men actually in bodily contact get shoved about by the fighting. A spear reaches several ranks deep, so rank-four and rank-five spearmen strike over the heads of the men in front — but nothing is touching them, so they go on dressing their ranks normally while the contact rank absorbs the shoving. That is what lets a formation stay a formation through a long grind instead of dissolving into a mob: the ranks behind the fighting hold the shape the front rank recovers back into. A block packed tightly enough that its ranks stand closer than a man is wide — Shield Wall and Testudo — genuinely puts a second rank in contact, and gives ground accordingly.

Figure 19: Two spear phalanxes grind head-on. Only the contact rank is shoved about, so the ranks behind it keep their files and the blocks stay recognisable formations rather than dissolving into a mob.

Figure 20: Formation density cycling with the Shift-reverse: T steps 0.9 m to 0.45 m locked to 1.8 m, then Shift + T steps back — the same parallelism as Shift+Y and Shift+X reversing the form-up-split and group-attack-mode cycles.

• Shield Wall (L) — shields locked edge-to-edge, a static line. It sharply cuts the missile and melee damage a unit takes from the front, but only from the front: a shot or blow into the flank or rear ignores the wall. The wall barely moves, so it is a holding stance. Anchor a flank or a chokepoint with it and let the enemy break on the front — but keep its flanks screened, because that is where it is soft.
• Testudo (U) — shields locked front, sides, and overhead: the strongest missile cover in the game, and it works from every direction. The cost is that the men fight weakly (packed head-down under their shields) and move slowest of all. It is an arrow-crossing stance, not a fighting one: walk a unit through a beaten zone of archery in testudo, then drop to Tight or Normal before it reaches melee.

Both shielded stances still brace against a cavalry charge the way Tight does, so neither is helpless if horse reaches them — but only spearmen actually turn a charge back. Because both are slow, commit them deliberately: a testudo caught in the open by melee, or a shield wall taken in the flank, is in trouble.

Figure 21: Testudo under fire: an infantry unit locked in testudo, ringed by archers firing from all four sides, weathers the volleys that its overhead shields blunt from every direction.

Both stances still assume a soldier’s shield stays locked with the men either side of it, which only holds while the threat comes from one direction. A soldier who is actually contacted by enemies on more than one side at once — attacked from the front and the rear together, or enveloped around a flank — can’t be defended by that assumption any more, so it individually breaks: it fights that cadence as an unmodified soldier (no wall bonus, no testudo penalty) and falls back toward the gap in the ring of attackers around it, rejoining the moment an opening actually appears. This is a per-soldier reaction, not a stance change — the rest of the unit keeps holding Shield Wall or Testudo throughout, and a soldier that breaks rejoins on its own the instant the encirclement clears, with no order needed. Square and schiltron never do this: their whole design is to face attackers from every side, so being surrounded is the plan, not a breakdown.

Figure 22: A Shield Wall infantry unit charged from front and rear at once: the tinted (red) soldiers have individually broken from the wall and are falling back toward the gap between their two attackers, then rejoin the formation once the rear attacker is beaten back and only one side remains. A few soldiers still flicker broken again later in the clip — ordinary flank pressure from the one remaining attacker, not a second encirclement.

Reshaping and changing density on the move

Changing formation density (T), assuming holding stances (Shield Wall, Testudo, Square), or adjusting frontage while a unit is actively moving faster than walking speed forces the regiment to brake and dress ranks. Close-order infantry cannot compress or expand spacing mid-sprint without disrupting cadence:

• Pace drops to a walk: The unit drops its speed to walk_speed, and any active sprint gait resets to a walking pace until the men settle into their newly assigned slots.
• Cohesion penalty: Rushing to dress ranks mid-move disrupts file alignment, inflicting a transient cohesion penalty that recovers back to full over time.
• Fatigue cost: Scrambling ranks mid-stride expends physical effort, adding an immediate fatigue cost.

Order formation density shifts and stance transitions while stationary or at a measured walking pace to preserve speed, charge momentum, and full combat cohesion.

Figure 23: A 120-man infantry regiment sprints at full pace, dropping speed to dress ranks and suffering a transient cohesion and fatigue penalty on a mid-sprint density change.

Form square against cavalry

When horse threatens a unit’s flank or rear and you have no spear screen to spare, form a hollow square at once: O (for orbis) jumps straight to the Square stance, and T also reaches it by cycling through Tight and Loose. The soldiers face outward on every side, so the unit presents no weak facing: an attack from any direction lands as if it were frontal (the usual ×1.5 flank / ×2 rear bonus is gone), and a cavalry charge into the ring is braced and backfires whether it hits the front, a flank, or the rear. There is no open side for the horse to exploit. The soldiers’ own outward-facing marks around the block’s perimeter make the stance easy to spot on the field — no separate ring overlay drawn on top.

The brace is physical, not a flat damage penalty. Watch a charging cavalry regiment hit a set square: its riders’ momentum is genuinely absorbed by the individual bodies it collides with, not waved through by a rule, so the charge visibly slows and settles at the ring instead of riding on through it.

Square is a defensive commitment, not a free upgrade. While it holds, the unit crawls (it can’t charge out) and hits softer (reduced offensive output). So:

The AI squares up on its own. An AI-controlled spear unit forms Square the moment enemy cavalry closes on it, with no player prompting it — the same reaction a disciplined human commander would take. Charging AI spearmen in the open only works if the cavalry arrives before the square finishes forming.

• Square up early — before the charge lands, not as it connects. The mobility penalty means a late square may not finish forming in time.
• Drop back to Normal the moment the horse is beaten off or pulls away, or the offence and mobility penalties will cost you against enemy foot.
• It’s a fallback, not a replacement for spearmen. A spear screen still blunts the charge and keeps your line mobile and hitting hard; square is what you reach for when a unit is caught in the open with cavalry on a bad angle.

The block folds inward, not across itself. A square is narrower and deeper than the line it came from — a 120-man regiment goes from fifteen files to eleven — so forming one always costs some ground. Each soldier takes the nearest sensible place in the new grid, keeping his own file and his own side of the block, instead of filing across the formation to a slot on the far flank. Measured on a lone regiment drilling with no enemy near it, that 120-man square costs about a third of the distance walked, and the ring is standing and still appreciably sooner. A real square forms under pressure rather than on the drill ground, so treat those figures as the shape of the saving and not a number to time a charge against.

Orbis vs schiltron. The hollow square comes in two variants with different trade-offs, not one generic stance:

• Orbis (O, or the control bar’s 0.45 m Square button) is the last-stand ring. Its charge brace and offence sit at the square’s baseline, but it holds its nerve better than any other stance — casualties erode its morale more slowly, so a surrounded unit fighting to the end is less likely to break outright. Reach for it when a unit is cut off and simply needs to survive as long as possible.
• Schiltron (Shift + O, or the control bar’s 0.45 m Schiltron button) is the cavalry specialist. Its denser hedge of levelled spears braces a charge harder than the orbis does — cavalry riding into a schiltron comes off worse than riding into an orbis — at the cost of hitting even softer in return, and it carries no morale bonus. Reach for it when the priority is breaking the charge itself, not outlasting a siege.

Both variants share the same hollow-square footprint and the same mobility penalty; only the charge-bracing strength, offence, and morale resilience differ.

Figure 24: Anti-cavalry square: an infantry unit is charged in the rear by two cavalry, then cycled to Square (T). The hollow ring faces outward on every side, so the rear charge finds no weak facing and is braced from behind — the unit holds where a rear-charged line would break.

Figure 25: Schiltron: the same rear cavalry charge as above, but the unit forms Schiltron (Shift+O) instead of Orbis. The denser hedge braces the charge harder than orbis does, at a deeper offence cost.

Figure 26: Two AI unit-leader decisions in one clash: the AI spearmen on the left form the anti-cavalry square on their own as player cavalry closes, with no player key press arming it. The AI infantry on the right advances and attacks, then re-targets and turns to face a second player unit that closes on its flank instead of fighting on with its back exposed.

Enemy coordination

The enemy’s regiments don’t fight in isolation — a subcommander watches the whole group and hands each unit leader a standing directive on top of its own battlefield instincts. Three behaviours come out of this:

• Line integrity — a unit that races far ahead of its neighbors gets held back near the group’s own depth instead of closing on the enemy alone.
• Mutual support — an unengaged neighbor gets sent to guard an ally that’s already fighting, rather than standing idle nearby.
• Flank coverage — a unit repositions to cover a group flank an enemy is bearing down on, before casualties or a maneuver leave it open.

A directive never overrides a unit’s own immediate reactions — turning to face a flank threat, forming anti-cavalry square, and calling for relief all still come first. It only shapes what an otherwise-idle unit does next, so the enemy line holds together without losing the individual judgement it already had.

Figure 27: Mutual support: one enemy unit stays locked in a long fight while its neighbor’s own foe dies fast, freeing it up. The subcommander sends the freed unit to guard its still-fighting neighbor, and it closes the distance to join the melee directly.

The general

Above the subcommander sits a general, driven by a doctrine profile — data, not code, so a non-programmer can author a new one. Each profile sets:

• Which army plans it knows — an advance-in-line or an envelop split into two wings, picked each AI tick against how the battle actually looks (an envelop needs a real numeric edge over the enemy, not just an even fight).
• A reserve fraction — how many units start held back instead of joining the active line.
• A reserve-commit threshold — once the active line’s own weighted-average morale drops below this, the reserves fold in.
• Whether to pursue routers — an aggressive doctrine chases a broken enemy down; a cautious one holds the line instead.

Two doctrines reading the identical battle can commit to different plans. A held-back reserve doesn’t just sit inert, either — it follows behind the advancing active line at a trailing reserve distance, and still turns to face a flank threat, forms anti-cavalry square, or answers a nearby ally’s call for relief, exactly like any other unit.

Figure 28: A cautious general holds two full-strength units in reserve while three weaker units alone meet the enemy. As the front line takes losses its morale crosses the doctrine’s threshold, and the general folds the reserves into the fight — they advance from their trailing reserve positions, march up, and join the melee.

Player delegation

You can hand a group of your own units over to the same subcommander machinery the enemy’s AI already runs: select the units and press Ctrl+Shift+a number key (0–9) to delegate them to AI subcommander group N. Press the same combination again on the same selection to take them back. A delegated group marches, forms square, calls for relief, and supports its neighbors on its own — exactly like an enemy group, right down to the orders it issues showing up in the transcript the same way a hand-issued order would.

The delegated group’s subcommander carries a period-flavored rank title drawn from your own doctrine profile (an aggressive doctrine’s subcommanders are Tribunes, a cautious doctrine’s are Legates) — shown in the selected-unit info panel while a unit is delegated.

You always outrank the AI: giving any explicit order to a delegated unit — a move, an attack, a stance, even just re-arming Hold — takes it back under your direct control immediately. There’s no suppression window and nothing to undo first; the next order you issue to that unit is exactly that, an order, not a suggestion competing with a subcommander’s own directive.

Figure 29: Two units delegated to subcommander group 5 (a Tribune, under the aggressive doctrine) march on and engage the lone enemy with no further orders; partway through, reclaiming manual control with Hold (Ctrl+H) instantly clears the delegation.

Weapon reach

Each weapon has an effective reach — how close two units must get before they trade blows. A longer weapon reaches that contact distance first, so it lands the opening strikes of a clash while a shorter-weapon enemy is still closing the gap:

• Spears reach furthest (~2.4 m) — spearmen strike first as the lines meet, on top of blunting cavalry charges.
• Swords (infantry, cavalry) are the mid-length baseline (~1.3–1.5 m).
• Archers carry only a short sidearm (~0.6 m); they’re meant to shoot, not duel in the press.

So a spear line meeting a sword line head-on gets free hits during the approach — another reason to lead with spearmen and let the shorter weapons pin and flank. Combat resolves per soldier by weapon reach: a spearman strikes any enemy within its longer reach, while a swordsman beyond its own (shorter) reach cannot strike back. That is also how a clash starts: the opening blow is the same per-soldier contest, not a regiment-sized lump of casualties fired the moment the two centres cross a circle of weapon-range-plus-radii. A small unit therefore cannot vanish in one tick at a distance its men have not closed, and a rear or flank contact hits the soldiers who are actually next to the blow rather than a facing-front rank on the far side of the block.

Figure 30: A 10-horse cavalry troop attacks another 10-horse troop head-on. Opening melee is per-soldier: neither roster vanishes in one tick at regiment-circle range, and both sides trade wounds once the men are actually in reach.

That opening-strike edge now holds up once the lines are locked, not just on the approach. A spearman doesn’t back away once a swordsman closes in — it holds its ground and keeps fighting, and its reach advantage keeps landing hits the same way it did on the approach. Landed strikes already shove the struck soldier back (see “Knockback now pushes both ways” below), so a longer-reach line pushing a shorter-reach one back is a real consequence of winning the exchange, not a separate rule bolted on top of it. The swordsman, for its part, keeps pressing forward until it closes inside its own reach — standing just outside the spear’s edge gains it nothing, so it closes the gap rather than trading blows it can’t answer. A spear line facing a sword line therefore holds its position and keeps landing free hits while the swordsmen work to close the last stretch, instead of both sides settling to a common distance.

A regiment’s front line isn’t the only rank that can strike, either: how many ranks deep a unit’s engaged tier runs scales with its own weapon’s reach and how tightly its ranks are packed — a long-reach spear regiment can field several ranks able to strike an enemy within their own weapon’s reach, while a widely-spaced cavalry regiment (a mounted soldier takes far more ground nose to tail) only ever fields its front rank.

Figure 31: A Spearmen regiment holds its ground once locked in melee with an Infantry/sword regiment — it doesn’t retreat to hold reach distance, and its reach advantage keeps landing hits the same way it did on the approach. The sword presses forward to close the gap, but keeps taking losses faster than the spear does.

Knockback now pushes both ways. Every blow obeys Newton’s third law: the struck soldier is shoved back as before, but the striking soldier also recoils along the same line, split between the two by effective mass (bracing counts as extra mass). A light, unbraced attacker rocks backward hard on contact; a heavy or set formation barely moves. That’s why a cavalry charge into a braced spear wall now visibly backfires on the horses themselves, not just on the spearmen holding their ground — the heavier, set defenders absorb most of the shove and hand a proportionally bigger kick back to the lighter attacker.

Drawing the other weapon

Some troops carry two weapons and deploy holding one of them: legionary Infantry march with the gladius drawn and a pilum stowed; Spearmen deploy holding the long spear with a secondary sidearm dagger/blade; Cavalry ride holding the spatha longsword with a heavy lance sidearm. Press Shift+I and the whole selection draws the other weapon, with a second press stowing it again. You can see the change on the men themselves once you’re zoomed in far enough for figure silhouettes: Infantry pick up the long shaft when drawing the pilum and restore the shield-forward stance when stowing it; Spearmen drop the spear shaft and draw their sidearm short blade; Cavalry switch between the spatha and the long lance shaft.

Switching is an in-place order — the block doesn’t march, wheel or reform to do it, so it costs you nothing positionally and can be done in contact.

What it does cost is throughput, because the two weapons sit at opposite ends of the reach-versus-rate trade the section above describes:

Table 2: Carried weapon combinations across the roster, and the tactical trades between them.

Troop	Deployed weapon	Reach	Sidearm	Reach	Gameplay trade
Infantry	Gladius (1.3 m)	1.3 m	Pilum (2.0 m)	2.0 m	Pilum trades rate for reach and piercing against armour; gladius maximizes sustained DPS
Spearmen	Spear (2.4 m)	2.4 m	Sidearm (0.6 m)	0.6 m	Spear stops cavalry charges and reaches forward; sidearm speeds close-quarters infighting
Cavalry	Spatha (1.5 m)	1.5 m	Lance (3.0 m)	3.0 m	Lance extends reach and delivers prone knockdown; spatha cuts in sustained melee

The pilum reaches most of the way to a spear’s 2.4 m and penetrates 25% of defender armour as a piercing weapon, so a regiment holding it lands the opening strikes of a clash and can field more ranks in its engaged tier — but it swings slower, deals less baseline damage, and imparts minimal whole-body knockback shove, so a line that keeps holding it once the fight is locked kills at roughly half the rate it would with the blade. The historical use follows from that: draw the pilum as an enemy closes, then switch to the gladius when the lines actually meet. Leave it too late and you’ve spent the approach with the wrong weapon; switch too early and you throw away the reach you were carrying it for.

Figure 32: An Infantry regiment draws its pilum and then stows it again. Every soldier’s silhouette picks up the shaft and then drops back to the shield, and the block holds its ground throughout — re-equipping never moves or reforms it.

Figure 33: A Spearmen regiment switches to its sidearm blade via Shift+I and engages in close melee. Every soldier puts away the spear shaft to wield their sidearm blade and scutum shield.

Movement speed

Units move at realistic real-world speeds, measured in metres per second, with three independent paces per type — walk, jog, and sprint (see Pace modes below) — scaled to how much each type’s panoply weighs:

Table 3: Movement paces by unit type, in metres per second.

Type	Walk	Jog	Sprint
Spearmen (heaviest kit)	1.1 m/s	1.8 m/s	2.8 m/s
Infantry	1.3 m/s	2.5 m/s	4.0 m/s
Archers (lightest kit)	1.5 m/s	3.0 m/s	4.5 m/s
Cavalry	1.7 m/s (walk gait)	3.5 m/s (trot)	8.5 m/s (hard canter)

Heavier armor and weapons cost proportionally more at a run than at a walk — walking pace barely changes under load, but running form degrades fast, so spearmen’s walk sits close to the lighter types’ while their sprint lags well behind. Cavalry’s walk and jog map onto the horse’s own real walk/trot gaits rather than a rider-weight-scaled fraction, since a rider’s kit is a small fraction of what the horse itself carries. Its charge pace is deliberately toned down from a true historical gallop (which would close distance faster than the map/camera scale reads well) toward a fast canter — see docs/SPEED_VERIFICATION.md for the historical comparison.

That sprint gap is a lever: cavalry cross the field and reach an exposed flank far sooner than infantry can re-form to face them, so commit horse decisively and time your foot to be where the charge lands. (The faster the gallop at the moment of contact, the bigger the charge bonus — so a clean run-up hits harder than a standing start.)

Pace modes

By default, foot units advance on AUTO pace: they walk on the approach, jog when a ranged enemy is firing at them (to spend less time under missiles), and sprint at full speed once within about 10 m of the target — so a charge arrives at full momentum while the walk-and-jog approach keeps the stamina bill to that final stretch rather than the whole march.

Check Walk advance in the unit settings panel to lock the selection to walk pace throughout. This is required for formed stances like shield wall or pike phalanx: those formations hold their shape only at a walk, and an auto-sprint will break them. Keep Walk advance on whenever you are using a tight formation that would crumble if the unit broke into a run — spearmen start with it on by default; every other type starts with it off.

The AUTO sprint-override also matters for charge timing: a unit that enters sprint range will accelerate through any missile fire it still faces — gaining maximum speed at contact rather than jogging through the kill zone. Use that to your advantage when committing cavalry or an assault infantry charge.

Figure 34: Pace modes: a regiment attack-marches on AUTO pace — walking on approach, jogging through the beaten zone once enemy archery reaches it, then sprinting at full charge speed inside charge distance of its target — all three phases as distinct legs of one advance. Check Walk advance in the unit settings panel to lock the selection to walk pace throughout.

Choosing a gait by click count

AUTO pace is the default, but you can lock a move order to a specific gait instead: right-click a destination once for Walk, twice for Jog, three times for Run, and four or more times for Sprint — click the same spot again within the multi-click window to add another click to the count. A click on a new destination (or one far enough from the last click, or one that arrives after the window has passed) restarts the count at Walk.

There are only three real paces per unit type (for cavalry, walk/trot/gallop — see the speed table above). Run has no pace of its own: it holds jog speed while still far from the destination and switches to sprint speed once inside the same close-range threshold AUTO uses, so it’s a distance-gated mix of the two rather than a fourth gait.

This overrides AUTO for that order: an explicit gait holds for the whole move, ignoring the fire/distance triggers AUTO would otherwise use. Reach for a chosen gait when you want a unit to hold a slow, orderly walk past a threat it would otherwise jog or sprint through (screening a flank, keeping a formation intact under fire without also toggling Walk advance), or to force an early sprint before AUTO’s close-range threshold would normally trigger it.

A chosen gait is also a chosen stamina bill: a walk is free, a jog drains the men slowly for as long as it lasts, and a sprint drains them fast, so a regiment sprinted across the whole field arrives sooner and arrives tired. Spend the faster gaits on the legs that need them.

Figure 35: Four identical Cavalry regiments sent on the same move, one click apart: Walk, Jog, Run, and Sprint pull apart into four distinct paces instead of moving in lockstep.

Side-stepping the line

A small lateral move order is carried out as a side-step: the unit shuffles sideways while holding its facing, instead of pivoting to face the move and pivoting back. That matters when you are dressing a line just before contact. Pivoting a regiment turns it away from the enemy for the length of the turn — its front momentarily points sideways, and the ranks lose their dressing while the soldiers reshuffle. A side-step keeps every soldier looking at the enemy and keeps the line aligned, so you can nudge a unit a few paces to close a gap or overlap an enemy flank without ever presenting your own flank in the process.

The cut-off is roughly a unit’s own width: a move that heads forward, wider than that or not, turns and marches normally. A wider lateral shift is a different maneuver again — see the lateral pivot below — rather than a plain centre pivot. So to reposition laterally in good order, make the adjustment small — a short step keeps the line intact, a long one commits the unit to a pivot and a march.

The lateral pivot

A move too far to the side to side-step, but not far enough off-facing to count as a move to the rear, is a file march: the regiment quarter-turns toward the destination’s side in place, marches straight there keeping its pre-turn footprint — the block never reforms into a line facing the new heading, so it stays exactly the shape it started as, moving sideways relative to its own printed grid — and then quarter-turns back to its original facing once it arrives. The unit ends up facing precisely the way it started; the reorientation only ever served the march itself.

Relative to the new heading, this reads as ranks and files trading places for the length of the march: what was the frontage (spread across the line) becomes the marching column’s depth, and what was the depth becomes its narrow frontage — soldiers advance in file, the way a column on the road does, not shoulder to shoulder. From the outside the block’s world silhouette doesn’t visibly rotate at all while it travels — only the individual soldiers’ own facing, and the direction they’re headed, change. Because nothing ever reforms, a lateral pivot has no reform delay the way a move to the rear’s reform can — a depleted unit executes it exactly as fast as a fresh one.

Figure 36: A Spearmen line ordered far to its own right: it quarter-turns toward the destination in place, marches straight there without ever reforming (the wide line stays wide, sliding sideways relative to its own shape), then quarter-turns back to its original facing on arrival.

Backing up costs time

An in-place maneuver that makes men back up takes longer than one where they step forward. When a formation reshapes without the block marching — a frontage change, a centre pivot on the spot — some soldiers have to move backward relative to the way they face to reach their new slots. Troops shuffle backward at only a fraction of their forward (jog) pace, so those men lag, and the whole maneuver settles more slowly than it would if everyone were stepping forward. That fraction varies by type — a heavily-kitted Spearman shuffles backward at barely a third of his jog pace, while a lighter Archer is closer to half — so a reshape’s settling time depends on which type is doing the backing up, not just how far it must move. The practical upshot: a reshape that pushes the rear ranks backward (deepening the line) is the slowest to dress, and slowest of all for the heaviest-kitted troops. Order it with a margin of time before contact, not at the moment of impact.

Figure 37: Per-type backward-walk speed: a Spearmen and an Archers regiment, forced to the same formation density and given the same rear-move order at the same tick. The lighter-kitted Archers settle their reform and step off first; the heavier-kitted Spearmen are still holding position when the clip ends.

The about-face (conversio) and quarter-turn are not reshapes, so this penalty doesn’t touch them: every soldier turns in place and holds their exact position for the whole turn — nobody steps forward or backward at all, only facing changes. Their duration depends solely on the drill turn rate, not the backward-speed cap.

The reform after an about-face dresses the line before or after the march. Flipping a unit that has taken losses puts its short, under-strength rank — previously the safe rear rank — at the new front, so a rear move reforms after the turn: under file-major reform, the men in full-depth files hold their ground while the short files step forward one rank pitch to square the front, avoiding a countermarch through the block. With Reform before move on, that brief step is taken before the unit steps off; with it off, the unit marches at once and reforms on arrival, buying speed at the price of leading the whole march with its thinnest rank.

This full file-countermarch mechanism is available as its own standalone drill — see How to play: Controls for the Countermarch (exelismos) hotkeys. A rear-move order reforms by holding ground; the standalone drill runs the full countermarch deliberately, always bringing a full rank forward, and offers a choice of where the whole regiment ends up afterward: the same ground (Choral/Persian), advanced (Macedonian), or withdrawn (Laconian). Use it instead of a rear-move order when you want the full-rank reform without also committing to a specific destination.

The centre pivot

A larger turn is a centre pivot: the unit rotates about its own centre onto its new heading in good order rather than snapping round. When the turn is sharp the unit pivots before it advances — it nearly halts, faces its destination, then marches — so it never slides sideways or backwards at speed. A destination to the rear therefore triggers a full 180° pivot: the whole regiment turns around to face where it is going before it steps off. (Don’t confuse that with the conversio about-face on V, where each soldier turns in place and the block does not rotate.)

The cost is time: a sharp reorientation spends a beat pivoting before the unit makes ground, and a 180° pivot spends the most. Plan turns ahead of contact — a unit told to reverse at the moment of impact will spend its pivot facing the wrong way. Note that a chasing or charging unit turns instantly instead, so an attack order never pays the pivot tax on the approach; the orderly centre pivot is reserved for your own move orders, where keeping the ranks dressed is worth the beat. Only a block that is much deeper than its own file spacing pays a noticeable pacing cost during this pivot. The ramp is continuous, so an ordinary line keeps near-full jog pace through the turn and a stock cavalry block pays the most.

A levy, or any unit ordered to run, skips the pivot: its heading turns straight onto its path. When you send it to a place (a move order, or a step back out of a fight), its men still do not run to new places as it turns: the block’s ranks come round after the new heading no faster than the men on its far corners can jog, so a deep block setting off a few tens of degrees off its front keeps its men in their places instead of sending them across the formation. A march on an enemy (an attack, a chase or a caracole) still swings its ranks round at once.

Re-facing to fight has a cost of its own. When a unit engages an enemy that is roughly a quarter-turn or more off its current front — attacked in the flank, or sent to hit a foe off to one side — it does not snap its facing round. Instead the whole line turns in place to bring its front to bear, the men holding their ground as the block pivots, and it withholds its blows until it has come about. A small correction still snaps and fights at once, so a unit already roughly fronting its target loses no time. The upshot: catching a unit on the flank buys you a genuine window — it must turn to answer you before it can strike back — so a flank attack rewards you both with the flank casualty bonus and the beat the enemy spends coming about. Once the turn settles, the unit by default just fights with whatever edge it now presents; the engine also supports reshaping the frontage right after the turn — back to the width it had before turning, or to match the frontage of the enemy it turned to face — reusing the same width-resize easing as a manual frontage drag, so the reshape never teleports the ranks. There’s no in-game control to pick a non-default reshape yet, so every unit still fights this out today.

A turn in place leaves the ranks lying where the men stood when it began, so a unit that has come about to fight is no longer square to the way it faces. A fresh order squares it again, but never while it is still in contact: an order given to a unit that is fighting, such as a new attack on the same enemy (the AI re-issues its orders every few seconds), leaves the ranks where the men stand, and the unit squares up only once it is out of the fight. A fight lets go of a unit for a moment between bouts, so the unit counts as out of it only after half a second clear of the enemy; a unit chasing its enemy squares up on the move. When it squares up, on leaving the fight or for an order given out of contact, the ranks come round onto the heading no faster than the men on its far corners can jog. Until they do, the unit holds its heading for its ranks: a chase or march that would turn it further away from them is slowed until the ranks catch up, and the unit is not left marching ever further crabwise. A drill ordered while a unit still waits to square up cancels the squaring, so the ranks keep the turn the drill gave them. Two cases still square at once: a unit standing a quarter-turn off its ranks, whose men each step to the nearest place in the squared block, and a distant unit drawn as a grid mark, which has no men to move. A unit whose ranks stand within 10 degrees of a quarter-turn off, given an order out of contact while it is not marching, is dealt out the same way at the quarter-turn, and only the last few degrees come round at the jogging pace. Measured on blocks squaring while they hold their ground, up to about 2.8 times as wide as they are deep (or deeper), wheeling the ranks nearly a quarter-turn round the centre walked the men further, and sent more of them across the block, than letting each step to his nearest place; past 10 degrees from the quarter-turn the wheel was the shorter walk for some blocks, so it is kept there. A wider line walked further when dealt out, so near a quarter-turn it wheels instead, as does a unit already marching when the order comes; either is still dealt out at exactly a quarter-turn, as above.

Figure 38: An infantry block on Hold turns in place to face an attack from the east, which leaves its ranks 85 degrees off the way it faces. Ordered on along its new heading once the attackers have routed, its men each step to the nearest place in the squared block, rather than wheeling the ranks the whole 85 degrees round its centre.

While a unit that has come more than a quarter of the way round stays unsquared, its men facing the enemy do the fighting: the ranks that strike are the ones nearest the enemy in the direction the men face, which after such a turn are the rear ranks of the block as it stood, not the front ranks the turn has left behind them.

The pivot itself also slows as the unit picks up speed — a unit turns fastest from a stand and tightens its turn rate the faster it’s already moving (down to 40% of the stationary rate at a full sprint), reflecting the real lateral force a moving body or formation can exert without losing footing or cohesion. So a turn ordered from a standstill snaps around quickest; one ordered mid-charge takes noticeably longer to come about. A second, independent cap takes over once a unit is genuinely sprinting (past its own jog pace) — below that, the pace/footprint pacing above already keeps every unit turning within its own physical means, so this second cap has nothing to add for an ordinary walking or jogging march. This isn’t limited to a deliberate cavalry charge: every unit sprints during the last stretch of any approach (closing the final ground to a target at full pace), so the cap can engage for any unit type caught mid-turn during that final rush, not cavalry alone — it just never has anything to do outside a genuine sprint segment. At a full sprint, though, the pivot also never asks a soldier’s own body to redirect its cruising velocity faster than that soldier’s own acceleration allows — so a broad, fast block (cavalry most visibly, since it sprints fastest and farthest) sprinting into a sustained course correction turns more deliberately still than the speed taper alone would call for, trading a slower turn for keeping its ranks intact instead of smearing out of formation.

Speed itself isn’t instant either: a unit ramps up to its target pace (walk, jog, or sprint) and ramps down when slowing or stopping, rather than snapping to a new speed. Heavier-kitted troops (spearmen) build up speed slower than lighter ones (archers), and every foot type slows down faster than it speeds up — stopping takes no effort, building momentum does. Cavalry is symmetric: a galloping horse can’t be reined in any faster than it can build up speed in the first place.

Arrival works the same way: a unit closing on its ordered point brakes over the final stretch and comes to rest on the marker, instead of charging to the wire and lurching to a halt past it. Only the final stop brakes — a queued waypoint route flows through its corners at pace, and a charge never slows: a unit closing on an enemy carries full speed into contact, so braking costs you nothing on the attack.

Figure 39: Braking on arrival: a lone Infantry regiment sprints toward a move order, eases off over the last stretch, and settles exactly on the ordered point — no lurching stop, no coasting past the marker.

The same friction applies even when a unit loses its momentum without a deliberate stop — its last target dies with no new order queued, say. Its speed doesn’t just tick down to zero on the spot while the unit sits frozen: it keeps sliding forward along the direction it was actually heading, slowing under the same braking rate, until it drifts to a genuine halt. Watch a cavalry troop that cuts down the last man of a squad standing to fight it: it comes out of the melee still moving at about 2 m/s, so it visibly coasts on for about a second before coming to rest, instead of stopping dead the instant it has nothing left to fight.

Figure 40: Coasting to a stop: a shallow Cavalry troop wipes out a small Infantry squad that stands to fight; with no enemy left to chase and no new order, it keeps moving on the speed it carried out of the melee, decelerating, before finally coming to rest.

Wheeling on a flank

A wheel (Z left / C right) is the third way to change a line’s front, and the most useful in the line of battle. One end — the standing flank file — plants and holds while the rest of the block swings around it like a door on a hinge. The line ends up facing 90° across from where it started and lying along a new bearing, with every man still in his rank and file.

It sits between the two in-place turns and the centre pivot:

• The quarter-turn (Q / E) changes facing but not place — the men turn where they stand and the block stays put.
• The centre pivot (a move order) rotates the whole block about its middle, so both flanks sweep and the line stays centred on the same spot.
• The wheel pins one flank and sweeps the other, so the line hinges rather than spinning in place — the near flank barely moves while the far flank marches a wide arc.

That hinge is what makes it a battle-line maneuver. Wheel a wing forward to take an enemy line in the flank — the ×1.5 side bonus (or ×2 in the rear) is the whole point of getting around them. Wheel a wing back to refuse a flank when the enemy laps around your end, presenting a fresh front instead of an exposed side. Because the far flank has to march an arc, a wheel takes time and covers ground: order it before contact, not in the grinding melee, and give the swinging flank room to travel. The swing is paced by the outer file — the corner man farthest from the hinge jogs his arc, and every inner file slows proportionally to hold the line straight — so a wide or deep block wheels noticeably slower than a compact one. A fighting unit ignores the order — you can’t hinge a line that’s locked in combat.

The same hinge swing also fires automatically as part of a large oblique move to the rear: if a rear-sector destination is oblique enough that the about-face’s own 180° reversal alone would still leave a sizeable turn to close, the regiment about-faces, THEN wheels the rest of the way onto the destination’s bearing, THEN marches — so a unit sent far to the rear on an angle steps off already lined up on its destination instead of correcting the last of the turn mid-march. You don’t order this one; it’s the drilled default for that shape of move (see how-to-play).

Cavalry never has to take that halt. Send a cavalry regiment on any sharply off-heading move — forward-oblique, to a flank, or to the rear, at any angle — and it wheels through the turn in one continuous swing while it keeps advancing, instead of stopping to about-face first. That makes a mounted wing meaningfully more responsive to redirect mid-battle: reroute it toward a fresh opening the instant one appears, without paying the halt-and-about-face cost a foot regiment given the same order still pays (see how-to-play). It’s also why an AI subcommander leading a cavalry wing can decompose its own directive into a wheel-then-march the moment the front shifts, rather than waiting for you to hand it a fresh order.

Order stances

By default — with no stance armed — a unit holds its formation and fights only what reaches its own weapon range. It won’t advance on its own to close with an enemy it has merely detected across the field, so your line stays where you put it until you order it forward: a move or attack order onto the enemy, or a March to contact / Chase stance. The AI-run enemy army still advances on its own, so battles start without you having to bait them.

Figure 41: Given no order, the player’s line (top) holds its ground for the whole approach — its position never shifts — while the AI enemy (bottom) marches the full distance in on its own and the two lock in melee on contact.

Each order can carry a stance, armed with a hotkey before you click (a coloured cursor shows which is active; Esc clears it). Stances turn the same units into different tools:

• Hold (H) — the unit stands its ground instead of chasing bait. It goes one step past the plain default: it also refuses to chase a foe it has already fought and that then pulls back, where an un-stanced unit pursues. Anchor a line, or keep a screen in place while your cavalry swings wide.
• Attack flank (F) / Attack rear (R) — the unit maneuvers for the side or back rather than the front, cashing in the ×1.5 / ×2 multipliers above. Pair it with a frontal pin.
• Skirmish (K) — a ranged unit kites: it backs off when a threat closes inside its kite distance, a fixed fraction of its own missile reach, trading ground to keep shooting rather than being caught in melee. A longer-ranged missile weapon kites proportionally further out.
• Support (G) — the unit guards a chosen ally, shadowing it and intercepting any enemy that closes on it. Good for gluing spearmen to vulnerable horse, or a guard onto your archers.
• Cycle charge (J) — a caracole for cavalry: the unit charges, lands its impact hit, then peels back to a standoff and charges again, instead of grinding in a static melee where the charge bonus is already spent. The charge bonus scales with closing speed, so each fresh run lands hard — but don’t cycle into braced spears, where a fast charge backfires. The retire leg rides at a trot and brakes onto the standoff before wheeling about, and the run back in canters until the final sprint (Figure 43). The same loop also keeps after a target that breaks and flees, instead of strolling off behind it and never re-engaging.
• Sweep routers (,) — the unit prioritizes a routing (fleeing) enemy over a closer enemy that’s still standing and fighting, even as the gap to the router narrows. Arm it on a unit you want cleaning up broken formations rather than getting drawn into (or staying stuck in) a fresh melee with whoever happens to be nearest — a natural pairing with a flanking cavalry sweep once the enemy line starts to crack. The pursuit rides at a canter beyond the sprint window, so the sweeper actually gains on a fleeing router instead of strolling after it.
• Roll the line (;) — once the current target routs or is destroyed, the unit automatically re-targets the next-closest enemy still fighting instead of standing idle or chasing a broken foe down. A single attack order can carry a unit through a whole line of weaker enemies one after another — useful once you’ve broken through a flank and want the advance to keep rolling without a fresh order for every foe.
• Pin down (.) — a stall/pin trade for a unit confronted head-on: it swings roughly half as often as a normal stance, so it deals less damage over time — but each swing’s follow-through opens a brief window where its own defense is lowered, so a foe that catches it mid-swing hits harder than it otherwise would. Use it to hold a line cheaply in casualties from its own attacks, not to win a stand-up fight outright; it’s a genuine tradeoff, not a strict upgrade over a normal stance.
• All-out attack (’) — a straight offense/defense trade: +20% effective attack, but the unit’s own defense drops 20% while it’s fighting all-out. It hits harder and dies faster at the same time, so use it to force a quick, costly decision — finishing off a wavering enemy, or a rearguard buying time — not as a strict upgrade over a normal stance.
• Chase (\) — relentless pursuit: the unit overrides the normal disengage exit a plain move order (or losing steady contact) otherwise gives it, so it keeps closing on and re-engaging the same fleeing or repositioning enemy instead of breaking off. It also presses the approach at a canter beyond the sprint window (a plain attack conserves stamina at a walk out there), so “relentless” means it actually closes, not just that it never gives up. Use it to run down a routing foe rather than let it escape to rally and rejoin the fight later (Figure 44).
• Wedge charge (-) — maximizes the approach at the cost of defense and stamina: a faster pace toward the target, a lower own defense while fighting, and faster fatigue buildup while engaged. Use it to punch a hole in the enemy line before it can brace, not as a strict upgrade over a normal stance — the unit that arrives first is also the easier one to hit.
• Knockback focus (=) — an offense-only trade: the unit hits 40% softer, but every blow shoves the enemy back much harder and is more likely to fell it outright. Hold Shift when you arm or issue it to pick the push distance: the plain hotkey shoves a struck enemy just far enough to clear the battle line, while Shift+hotkey drives it much further back. Use it to break up a packed formation or buy room to maneuver, not to win a straight fight — the softer hits mean it trades its own damage output for control of space.
• Give ground (`) — a deliberate withdrawal: the unit backs away from its nearest enemy at a steady pace while it keeps fighting, regardless of weapon reach on either side. This is the explicit, player-commanded counterpart to Weapon reach’s passive standoff above — that mechanic only ever holds a longer-reach unit’s ground or presses a shorter-reach one forward; it never backs a unit away on its own. Arm it when you want to trade space for time — peeling a spent line back toward reinforcements, or buying a moment for a flank to arrive — without breaking off the fight entirely.
• Multiple engage (F2) — for a unit pressed by two or more distinct enemy regiments in contact at once. Normally a unit keeps turning to bring its front to bear on whichever single foe it’s currently fighting; with several attackers pressing from different bearings, that turn would just whipsaw the whole line back and forth between them. Under this stance the unit instead holds its current facing once genuinely 2+ enemy regiments are in contact, and widens its own frontage toward the combined width of every attacker pressing it, so its line can actually reach all of them instead of presenting a narrower front than the fight has grown to. Both effects stay off below 2 adjacent enemies — a normal single-opponent fight under this stance looks like any other stance. Casualties are no longer this stance’s own job to spread around: every soldier in an engaged regiment now finds and strikes its own nearest enemy soldier across every adjacent enemy unit in reach, regardless of stance — so once this stance’s widened frontage brings a second (or third) enemy within reach, it starts taking casualties immediately, in the same cadence, with nothing needing to rotate onto it.
• March to contact (F3) — an advance-to-contact stance for a plain move order. Normally, a plain move order with no explicit attack target pushes straight through any hostile it crosses — the unit auto-acquires and fires or fights only under an explicit attack order, or once it has no move order left at all. Under March to contact, that same plain move order instead stops the unit to fight anything it meets along the way, then — once the fight ends — resumes toward its original destination on its own, with no separate order needed. Unlike Chase, it never actively pursues a foe: a hostile that breaks contact and retreats is let go rather than followed, and the march simply continues. Arm it before issuing a move order to advance a unit through contested ground and let it handle whatever it meets along the way, without babysitting each engagement.
• Brace (F4) — a deliberate “plant and hold” stance for receiving a charge. Bracing is graded, not a flat switch: a merely-engaged line (no explicit order) is only partly set, and one still under way is barely set at all. Ordering Brace and holding the line still for about a second sets it fully — the doc’s “you must be set before it arrives” — raising its resistance to being knocked back or knocked down, and it stops the unit from pressing forward or auto-advancing while armed, the same as Hold. A grounded spear/pike line braces harder than a shield-only one at the same posture, so set spears before contact rather than shields alone when you can choose which.
• Flanking maneuver (F6) — an envelopment stance for an engaged unit: rear ranks spread outward beyond the opposing regiment’s frontage so the outer files wrap around both flanks of the enemy line. While engaged in melee under this stance, the unit holds its forward facing instead of turning off-axis as wing soldier collisions jostle, allowing the enveloping wings to deliver flank strikes into the enemy’s exposed flank arc (Figure 47).

Figure 42: Two identical Spearmen lines face an identical cavalry charge. The right line is ordered to Brace and settles well before contact; the left is left at the merely-engaged default. The braced line holds its ground almost exactly on impact, while the unbraced line visibly presses forward into the melee.

Figure 43: A cycle-charge caracole: the horse lands its impact, peels back to a trot, brakes onto the standoff, wheels about, and canters in for another run. Opening melee is per-soldier, so the first blow no longer deletes a 70-man line; the loop still peels and returns. The same loop keeps after a target that actually breaks and flees.

Figure 44: Stern-chase pacing: two identical cavalry pursue the same withdrawing infantry from beyond the sprint window. The Chase-armed unit presses at a canter and runs the quarry down; the plain attack order keeps the stamina-conserving walk and barely gains on a quarry walking away at nearly its own pace.

Figure 45: Two equal-reach Infantry regiments lock in melee. Once engaged, the player issues Give ground on their own unit — it backs steadily away from the enemy line while the two sides keep trading blows, instead of holding its ground the way the passive reach standoff would.

Figure 46: A player Infantry regiment arms Multiple engage before two smaller enemy Infantry regiments, closing from different bearings, reach contact. The first enemy engages normally; once the second also presses in, the regiment’s frontage widens from 11 to 14 files (their combined width) and its facing holds steady instead of swinging between the two of them.

Figure 47: A player Infantry regiment locked in melee with an enemy line executes a Flanking maneuver. Its frontage widens beyond the enemy’s line width, allowing its outer wings to envelop the opponent’s flanks and land flank strikes while holding forward facing.

Figure 48: A player Infantry regiment is ordered under March to contact toward a point beyond a weaker hostile sitting in its path. It halts and fights on contact instead of marching through — the info panel keeps reading its original destination order the whole time, so the queued move was never cleared — routs the weaker foe, then resumes marching and arrives at its original destination.

Push stance (/) is the aggressive, offensive counterpart to Give ground: an engaged unit continuously applies physical pressure to drive its opponent backwards. Where Give ground yields territory while fighting, Push stance advances into the enemy line, pushing opposing soldiers back while maintaining melee contact.

Figure 49: Two Infantry regiments lock in a melee standoff. The player selects their own unit and issues Push stance — its front line advances into the enemy’s, driving the opposing soldiers steadily backward while the two sides keep trading blows, instead of holding at the standoff distance the way a plain engagement would.

Disengage and step back (Ctrl + ↓) is different in kind from the stances above: a one-shot maneuver, not a standing order. Where Give ground keeps fighting while backing away, disengaging breaks the fight off entirely — the unit stops striking back and steps a short, fixed distance to the rear, holding facing. Use it to pull a unit out of a fight that isn’t worth finishing, so it can regroup or be redirected elsewhere, rather than to hold a losing line (that’s what Give ground is for). It costs something too: the enemy’s own attack order doesn’t release just because you stopped fighting back, so it keeps swinging — with the flank/rear bonus — for as long as it can still reach the retreating line.

Disengage with sacrifice (Shift + Ctrl + ↓) is a harsher variant for heavily pressed regiments. A small rearguard detachment (~10% of remaining soldiers) splits off at the point of contact as a genuine separate unit, stays on Hold, and keeps fighting for real — it is not a casualty write-off or a flat pursuit-speed penalty on the pursuer. The main body’s survivors step back and march clear immediately; the enemy is held up because it is physically blocked by the rearguard’s own bodies, not because a number slowed it down. The rearguard fights until it is destroyed or a short lifetime expires, whichever comes first.

Figure 50: A player Infantry regiment locked in melee is ordered to disengage with sacrifice. A small detachment splits off, stays put, and keeps trading blows with the enemy while the rest of the regiment marches away unimpeded — the enemy stays engaged with the rearguard instead of following the retreat.

Figure 51: Two evenly-matched Infantry regiments hold a stalemate in melee. The player disengages their own unit — it clears its target and steps back, holding facing, while the enemy (still committed to the fight) keeps pressing after it.

A volley is not instant: the arrows fly a real arc and land where they were aimed a moment later, so a fast unit can partly outrun a shot loosed at long range, and casualties arrive on the beat the arrows come down rather than the instant they leave the bow.

Each missile weapon has its own reach

A ranged regiment shoots with the missile profile of the weapon it carries, not with one reach shared by every archer on the field. A profile fixes six things at once:

• Reach — how far the regiment can loose at all. The bow reaches 8 m; the pilum, a heavy shaft thrown by hand, reaches 15 m. A longer-reaching profile extends auto-acquisition to its missile reach. The regiment acquires a target on its own past the ordinary acquisition distance.

• Cadence — how long the regiment takes between volleys. A bow looses every second; a pilum regiment throws every two.

• Flight & trajectory — whether the volley is lobbed or thrown flat. Inside half of maximum range, shots fire flat for quick delivery. Beyond that threshold, dynamic trajectory solving selects the lower launch angle (up to 45 degrees). Dedicated flat weapons like the pilum stay flat at all distances.

• Gravity — per-profile vertical acceleration governing flight hang time and arc height.

• Punch — how hard each volley hits against a melee strike of the same stats.

• Accuracy at range — how much of that punch survives to the edge of the reach. A bow loses nothing across its 8 m. A pilum volley falls off steadily with distance and lands at half its effect at the full 15 m, so the last few metres of its reach are the least worth using.

• Ammo — how many volleys the regiment carries before exhausting its ammunition. A bow regiment fires without limit (-1 / unlimited). A pilum regiment carries 2 throws by default, expending its heavy shafts in the opening clash before closing for melee or retiring. Units out of ammo halt firing and stop suppressing enemy formations.

• Overhead friendly clearance — volleys passing over friendly troops test ballistic arc height. Shots clearing the 2.0 m shoulder clearance threshold pass safely overhead without friendly fire, while low or flat shots through friendly lines are intercepted.

So a longer reach is not a free upgrade. A pilum regiment at 14 m throws at a target it can only just reach, with about 64% of the effect it would deliver at 5 m. Closing that distance improves a volley by about 56%, at the cost of being closer to whatever it is throwing at. The kite distance under Skirmish scales with the reach, so a longer-ranged skirmisher backs off further before it is caught.

Figure 52: Two player-controlled archer regiments stand on one line, both 14.5 m from an enemy regiment. The left one carries the pilum profile (15 m reach) and opens fire on its own; the right one carries the bow (8 m reach) and holds, the enemy out of both its reach and its auto-acquisition range.

Figure 53: Player-controlled archers lob a volley at 14 m over a friendly infantry line deployed 6 m ahead. Because the dynamic ballistic arc reaches approximately 47 wu of height at the friendly position, it clears the 2.0 m shoulder clearance threshold and passes safely overhead without friendly fire, striking the enemy line beyond.

Figure 54: Rhodian slingers with an authored 180 m reach engage an advancing enemy infantry block from 42.5 m (850 world units). Because the distance exceeds the 30 m demote range, both formations fight in the far tier as aggregate flock marks. The slingers’ long reach spans the gap, opening fire immediately and eroding enemy morale and cohesion well before the lines close to close-tier promotion range.

Turn your shields toward the archers

Arrows are resolved one man at a time when they land, and each one is tested against the shield the man it reaches is actually holding. A shield covers an arc in front of its bearer — a body shield like the scutum covers a wide front, a cavalry round shield much less, and archers carry no shield at all. An arrow that arrives inside that arc is usually stopped: it either glances off, or sticks fast in the shield face. An arrow that arrives from outside the arc has nothing in its way and kills.

So which way a regiment is standing now decides how much a volley costs it, before any stance bonus enters:

• Face the archers. A regiment turned toward the incoming fire puts its shields between its men and the arrows, and most of the volley is wasted on wood.
• Never let archery reach a turned flank or an exposed back. The shields point the wrong way, so the volley lands nearly in full — and the flank casualty multiplier stacks on top of that, so an enfilading volley is far worse than the sum of its parts.
• Shields do not help everyone equally. Cavalry carry a smaller shield and stop a good deal less, and your own archers carry none, so they are as exposed facing the enemy as they are running from them — keep them out of the beaten zone rather than trying to turn them into it.

Arrows that stick in a shield stay there, and their weight tells. A regiment that has stood under archery long enough is dragging a growing load of spent shafts, and its cover degrades as it does — capped, so a bombardment never strips a shield bare, but a wall that has absorbed a long exchange no longer blocks the way it did at the first volley. It takes a sustained pounding to matter, so this is a reason to rotate a battered regiment out or close the distance, not a reason to fear the opening exchange.

This is separate from the stance bonuses in Formation stances: shield wall and testudo change how much damage a volley does in the first place, while the arc above decides, arrow by arrow, whether an individual man’s shield was in the way at all. The two stack, so a testudo whose men are also facing the archers gets both.

Figure 55: Shield arc under archery: two identical 60-man Infantry blocks, each under fire from its own Archers regiment due north. The left block faces the arrows and its shields cover them, while the right block is turned east and its flank is bare. Watch the two soldier counts diverge — the facing block ends the clip at 53 of 60 with morale 75, the turned one at 25 of 60 with morale 32.

Group attack distribution

When you have several units selected and right-click a target, press X to cycle between two group-order modes. The same toggle governs both an attack on an enemy and a relief order on a friendly (see Passage of lines) — which pair of behaviours applies depends only on what you right-click:

• Focused (default) — every selected unit charges the same enemy, or (on a friendly target) every reliever swaps with the one tired unit clicked. Use it to overwhelm a single regiment quickly.
• Distributed — each unit picks a different nearby enemy, sorted by proximity to the clicked target; on a friendly target, each reliever instead relieves the nearby engaged ally in front of it (see Passage of lines). Extra attackers cycle through the list if you have more units than targets. Use it to hold an entire line rather than piling onto one regiment (or one relief swap) and leaving the rest unengaged.

The flash message in the top-centre of the screen confirms the active mode. The mode is sticky — it stays set until you press X again (Shift + X cycles the other way).

Figure 56: Group attack distribution: three regiments in distributed mode each target a different nearby enemy instead of all piling onto one.

Deployment distance

The standard battle opens with the two armies’ front ranks 18 m apart. Each block dresses its front rank on its army’s line and its depth runs back from there, so even a deep cavalry squadron starts on its own ground, clear of the hill and of the enemy’s cavalry. 18 m is close enough that every unit keeps individual-soldier fidelity from the first tick. A campaign clash can open wider. Each province in a campaign’s data (data/campaigns/*.json) may declare a deployment_gap_m, the distance in metres between the two armies’ front lines when a fight over that province is fought out in the tactical battle; an open plain deploys wider than a defile. The attacker’s line and the terrain band stay where they are, the defender’s line moves back by the difference, and the field grows with it so the defenders keep the same ground behind them. A province that declares nothing fights at the standard distance, so the campaigns you already know play exactly as before until their data says otherwise.

Beyond 30 m both armies open in the far simulation tier (grid marks instead of bodies) and promote back to individual soldiers once the gap between the blocks closes inside 20 m, so a wide deployment is a march before it is a fight. Dress the line and pick your ground before stepping off: a formation that arrives ragged after a long walk meets a line that had the same time to prepare.

Figure 57: A 60 m deployment: both lines open as far-tier marks, the player’s line is marched at the enemy while the enemy advances to meet it, and both promote back to individual soldiers as the gap closes.

Fog of war

Battles open with complete visibility across the entire map by default. Toggling fog of war (F7 or ☰ Menu -> Fog of war) limits your intelligence. When fog is enabled, you see only what your own regiments can directly observe. Every friendly regiment projects a circular sight disc around its center. An enemy unit whose center lies outside every friendly sight disc is hidden. Friendly units are always visible to you. Terrain actively shapes line of sight: impassable features (like hills) block sight lines completely, while wooded ground screens vision and attenuates observation range. The field itself carries its own persistent explored/unexplored state, normally driven by your own units’ sight discs rather than by enemy visibility. Ground your side has never covered with a sight disc renders as an opaque dark fill; ground it has explored but does not currently see is dimmed rather than hidden outright; ground currently inside a friendly sight disc renders normally. Ground under an enemy you currently see also renders normally and becomes explored, so once that enemy moves on the ground stays dimmed rather than going dark again. Once explored, ground never reverts to unexplored — marching your army away from a stretch of terrain dims it but never re-hides it. See How to play: Fog of war for the full render-state breakdown.

Table 4: Sight ranges by unit category under fog of war.

Unit	Multiplier	Sight (default field)	Role
Cavalry	1.4x	21 m	Scouting, screening, pursuit
Ranged	1.2x	18 m	Forward observation, support
Foot	1.0x	15 m	Line-of-sight across front

Sight ranges default by unit category.

Mounted troops see farthest: 21 m on the default field (a 1.4x multiplier). Ranged troops see intermediate distances: 18 m on the default field (a 1.2x multiplier). Foot troops hold the baseline: 15 m on the default field (a 1.0x multiplier). When a unit routs, panic degrades its observation: a fleeing unit suffers a 0.6x penalty to its sight radius.

Why a cavalry screen is worth more

On an open field with fog active, blind marching is dangerous. Even at the standard 18 m between the front ranks, opposing foot lines deploy out of sight. Heavy infantry marching forward without reconnaissance risks blundering into spears. This gives cavalry a decisive tactical purpose before the lines meet. With a 21 m sight radius (a 40% reach advantage over foot) and superior speed, horsemen can push ahead of the heavy line as a scouting screen. A forward screen reveals enemy deployments and flags open flanks. It illuminates approaching columns while your main formation stays hidden. Skirmishers and archers (18 m reach) can also screen the heavy line, maintaining contact while staying outside the 15 m sight of enemy foot.

Terrain occlusion and screening

Open ground allows units to observe up to their full sight radius, but terrain actively alters visibility:

• Impassable terrain blocks line of sight: Hills and solid barriers completely occlude vision (sight: "block"). An enemy directly behind a hill cannot be seen, regardless of proximity, until friendly scouts maneuver around the obstacle and establish an unobstructed line of sight.

• Foliage screens and attenuates sight lines: Forests and slow patches (sight: "screen") attenuate vision across them. Each screening patch cuts effective sight range by half (a default 0.5x multiplier), compounding if multiple patches intervene. An enemy sitting deep inside or across a wooded patch requires scouts to close well within the attenuated radius to spot them.

What ghost markers tell you — and what they do not

When a spotted enemy leaves your sight discs, it leaves a ghost marker. The marker is a fixed rectangular silhouette (3 m wide by 1.5 m deep). It sits at the exact position and facing the enemy had when last observed. Over the next ten seconds (600 ticks), the marker fades from 60% to 20% opacity. Understanding the limits of a ghost marker is critical:

• It marks where the enemy was, not where it stands now.

• It preserves the last-seen facing, but hides subsequent maneuvers.

• Its opacity indicates age, fading as the sighting grows stale.

• Right-clicking a ghost marker falls through to a ground move order.

Attack orders require a visible target, so an order against a ghost cannot lock on. If the enemy has marched away, your troops will march into empty ground.

Scouting and pursuit decisions

Fog of war fundamentally reshapes pursuit and battlefield control. Fog of war does not alter the retreat boundary. The retreat margin remains fixed at 9.5 m regardless of whether fog is enabled. However, if your pursuers lose sight of a routing enemy, that enemy enters the dark. A broken unit that breaks contact and gains distance can steady its nerve and rally. Dropping pursuit in the fog risks allowing shattered remnants to reform. Fast cavalry must stay in contact, shadowing routers until they shatter or break completely.

Fog of war binds both sides — it holds your own troops back too

Fog of war is not a player-only handicap, and it is not only about the enemy general’s own plans. Every AI side — the enemy general down through its unit leaders, and a delegated group under your own command — reads exactly the same fogged view fog gives you: a unit it can’t currently perceive is a unit it can’t newly engage or newly react to. There is no separate, easier ruleset behind the enemy’s decisions; the same sight discs, the same terrain occlusion and screening, the same “own units always known” rule apply on both sides. This cuts both ways tactically. An unseen flank march genuinely blindsides the enemy general — it commits to nothing against a threat none of its own units has sighted, and only reacts once contact is made or a unit’s own disc reaches the mover. The reverse runs on the same rules, but is not a guarantee: an enemy force you cannot see is judged by the identical perception gate you are, so a screen that denies it sight of your own dispositions still denies it a basis for a plan, not just denies you information about its — yet because sight radii differ by unit type (cavalry sees roughly 40% farther than foot) and terrain and position differ per unit, the two sides’ actual information can be asymmetric: an enemy watched only by short-sighted foot, or aided by better ground, can sometimes see you while staying unseen itself.

The same rule reaches all the way down to individual units, on both sides, not just command-level decisions. An idle ranged unit — yours or the enemy’s — will not loose a volley at a target its own side has not sighted, even one already within its bow’s reach; a unit on Support stance will not peel off to fire on, or chase down, a threat near its ward until its own side actually perceives that threat. Only ONE reflex stays enemy-side only: an idle unit auto-marching toward a merely-detected foe is exclusively how the AI closes distance on its own (your own units always hold formation and wait for an order, fog or no fog, so this specific reflex never applied to you either way). Everything else — who freshly fires, who freshly gives chase, who enters combat unprompted — plays by one identical rule on both sides: nothing on either side of the field can newly react to what its own side’s perception has not yet sighted, though because that perception itself differs by sight radius and terrain, an equal rule does not guarantee equal information. One disclosed exception, on both sides: this gate only ever stops a FRESH pick. Once a unit has already committed to an explicit attack order, it keeps closing on and fighting that same target’s current live position even after your side’s perception of it lapses — an attack already under way is not re-checked, tracked as #1626 for a future last-known-contact model.

Terrain

Two types of terrain patch appear on the battle field:

Table 5: Terrain patches and their movement effects.

Patch	Colour	Effect
Forest	Dark green	Passable - units can enter, but movement is slowed to 60% of normal speed
Hill	Tan	Impassable - units route around it via A* pathfinding

Units automatically path around impassable terrain when ordered to move. A formed regiment takes the detour in good order: it steers by the corridor’s long-run direction and pivots onto it no faster than its corner man can run, so the block holds rank-and-file around the turn instead of smearing into a crowd. The forest is a different trade-off: going around it costs distance; going through it costs time. Generally it is faster to push cavalry or fast-moving infantry through rather than around a narrow forest.

A move order (a click, an arrow-key side-step, or an AI order) whose formation footprint at the destination would overlap impassable terrain is pulled back along the move to the nearest spot where the whole formation fits, laid out in the facing it will hold there (a quarter-turned block is first squared to that facing, as the order itself squares it, each man taking the new place nearest where he stands rather than marching across the block to reach it). A deep column side-stepping toward a hill stops with its rear ranks at the hill’s edge rather than among the rocks. If no spot along the move is clear, the unit stays where it is. Either way a unit under your command says so: a short HUD message names it and says it is halting short of impassable ground, at the edge of the field, or holding. A block that already overlaps the terrain where it stands, such as a deep cavalry block deployed against a hill, can still move, as long as the order would not push it deeper into the terrain than it already is. The edge of the field works the same way: an order is pulled back until the whole formation fits on the field, so a deep column sent to the edge halts with its front rank on it rather than half off the map. Individual soldiers are kept off a hill too, whatever moves them toward it, not only orders. A man pushed, swept or marched onto the hill is held at its edge. A rider on the outside of a column brushing past a hill stops at its edge rather than riding over the rocks. (A unit deployed partly on a hill keeps the men already there where they stand, free to walk off it.)

Figure 58: An order into the hill, amended: the block is sent to the middle of the hill, its march is cut back short of the hill’s edge, and the HUD names the unit and says it is halting short of impassable ground.

Figure 59: A deep column (three files, twenty ranks) facing north beside the hill is side-stepped east six times. The first two steps go the full distance; the third stops with the right file’s rear ranks at the hill’s edge; the last three leave the block where it stands. No soldier ends up on the hill.

The obstacle a unit routes around is the terrain you see, plus the unit’s own footprint — and that footprint margin is two different sizes depending on whether the leg is straight or turning. A straight, unturning leg keeps a margin equal to the width the block actually sweeps sideways as its centre follows that exact leg: its half-frontage when marching straight ahead (files to either side of the line of march), widening toward its fuller half-depth the more that leg cuts across its own facing instead of along it — a side-step or a formed turn still under way sweeps close to the block’s full depth even though the block itself never turns to face that direction. A small squad slips through a lane that a wide battle line marching straight ahead must round with a broad swing, and a deep column (many ranks, few files) needs no more room than its own narrow frontage on a straight march, regardless of how long it is — but the same column stepping sideways needs its full depth of room instead. Every detour leg — rounding an obstacle’s actual corner, or the coarser waypoint the route falls back to when no corner is in clear view — is where the route’s direction changes, and so can the block’s orientation relative to it, so it keeps the fuller margin wherever the ground leaves room for it, regardless of travel direction: the corner man’s reach from the centre, folding in the block’s depth as well as its width. A lane that looks clear of the hill is clear — for any unit narrow enough to fit through it on its own straight-leg margin. The walked detour hugs the corner’s own margin the whole way around a bend: straight legs run along the obstacle’s own grown boundary rather than a wider routing-grid lane, so two blocks of different widths round the same hill on two visibly different tracks (Figure 61). When a leg starts or ends closer to the hill than the straight-leg margin, that margin is cut, for the whole leg, to the smaller room its two ends leave; at the destination end that room is measured square to the hill’s edges. So a block sent to a spot just off one of the hill’s corners marches straight there and halts facing along its line of march, instead of swinging wide around a corner it was never going to touch — as long as its line of march clears the hill by that smaller room; one that would cross it still routes around. On such a leg the block’s flank can pass the hill’s edge closer than its full margin; individual soldiers are still held off the hill itself. A block left standing just off one of the hill’s corners, inside its margin, walks straight off on an order that takes it away from both of the corner’s edges, or away from one and parallel to the other, rather than first edging round the corner. Any other leg measures its start’s room in a straight line from where the block stands, still cut to the destination’s room as described above. That covers a leg running down a face, one turning back toward the hill, and one that moves away from one edge while tilting even slightly back toward the other. Unless the destination’s own room cuts it lower still, such a leg is judged at the start’s straight-line room, which puts the block’s own position inside the grown hill, so it is judged blocked and routes round the corner.

Figure 60: A 30-man squad ordered straight down the lane just west of the hill holds its lane the whole way — its flank passes less than a metre off the drawn hillside. The routed obstacle is the drawn slope plus the squad’s own footprint margin, so the narrow lane fits it exactly; a full-width line given the same order would swing around instead.

Figure 61: Two widths, two lanes: a wide battle line and a narrow squad round the same hillside, and each walks the drawn terrain plus exactly its own footprint margin — the line swinging out around the shoulder, the squad tucking in tight along the face — instead of both converging onto one wide routing-grid lane.

Figure 62: A deep, narrow column (three files, many ranks) marches a straight leg that passes well clear of the hill – clear of the width it actually sweeps, even though the hill sits inside the wider margin a direction-blind reading of the block’s footprint would have used. The column holds a dead-straight line the whole way instead of swinging wide to round terrain that was never really in its path.

Figure 63: A wide single-rank line turns once onto its bearing near the hill and holds it — no back-and-forth snapping between headings while the regiment settles onto its new course.

A routing unit paths around a hill the same way while fleeing — it’s no longer a straight sprint toward the back edge. If a router is boxed in with every escape direction blocked, it gives up fleeing and stands to fight rather than running in place against the obstacle.

TipForests slow cavalry charges

A cavalry unit that contacts the enemy while still inside a forest hits with a reduced charge bonus — it can’t build full speed in the trees. Stage your charges from open ground whenever possible.

Order-response delay

Units don’t react instantly to orders. There is a brief reaction delay before they execute — about 0.5 s for infantry and spearmen, 0.3 s for cavalry. A unit already in melee is not slowed by the delay.

Drill commands (the wheel, about-face, quarter-turn, and countermarch) carry their own, shorter drill response beat — about 0.2 s by default between the command and the men starting the evolution. A drill is a rehearsed, single-evolution response to a direct order, so the beat is quicker than planning a fresh march. It is a per-unit characteristic: an elite regiment can be staged with a snappier beat and a raw levy with a sluggish one, so the same command visibly ripples through a mixed line.

This models the time for orders to travel down the chain of command. In practice:

• Commit cavalry early. If you wait until the last moment to redirect horse, they may not reach the flank in time.
• Feints are risky. A unit you pull away from melee takes rear hits during the disengage and takes a moment to start moving — don’t disengage into a second threat without enough margin.

Disengaging — a risky lever

Right-click the ground (a move order) to pull a unit out of melee. It’s how you retreat a battered regiment or redeploy cavalry to a fresh flank.

WarningDisengaging exposes your back

While a unit marches away it shows its rear to whatever it left — so that enemy gets free rear hits (×2) until your unit is clear. Pull out at the right moment, or pull out into a disaster.

Stop the unit and it re-engages anything nearby. Timing the disengage — late enough that the enemy is committed, early enough that you survive the exit — is one of the game’s sharper decisions.

A move order that would walk a regiment through an enemy it is not fighting still arrests at contact. The kinematic march yields rather than compressing the rear ranks into a front that is already held; go around, or wait for a gap.

Figure 64: A player Infantry regiment is ordered straight through a held enemy line. The front arrests at contact and the rear ranks stay dressed instead of compressing into the held men.

A move order given while the unit is still fighting takes hold right away — the unit starts marching off toward the new destination immediately, it doesn’t hold the line and keep trading blows first. If no real fight happened under this order, the march just continues to its destination as normal. If the unit was engaged, the destination gets one more check once the fight has genuinely died down: if the ground you were marching toward is now held by a living enemy, the order cancels outright instead of marching the unit into occupied ground, so the commander isn’t left dragging it toward a destination the battle has since invalidated. If the destination is still clear, the march continues as normal even though a fight happened along the way. Reissue the order once the ground clears (or pick a different destination) to redirect the unit for real.

Figure 65: Redirecting an engaged unit: a mid-fight move order marches the unit off right away, but once its engagement with the enemy dies down the destination is checked one last time — if the enemy still holds that ground, the order cancels outright and the unit falls back into the fight instead of marching into occupied territory.

When that mid-fight destination lies behind the regiment — a genuine pull-back out of the fight — a well-drilled unit doesn’t pivot in place under pressure. It backs straight out instead, shields still toward the line it just left, ranks square the whole way, and only once it is clear of the grapple does it drill its about-face and march the rest of the way (Figure 66). The same drill the relieved regiment uses during a line relief, available on demand: order any disciplined, non-hasting unit rearward while it is engaged and it withdraws as a fighting body rather than turning tail. Undisciplined units, and any move ordered at the run, still break off the old way — pace over poise.

Figure 66: A fighting regiment ordered to a destination behind it backs out facing-held first — no mid-press pivot — then turns about and marches on once clear of the fight.

Facing a skirmisher screen

Not in play yet. Screening is a property of a general’s doctrine profile, and nothing in the game selects a profile yet — every battle runs the aggressive doctrine, which does not screen. The behaviour below is built, tested, and shown in the clip; what it is waiting on is a scenario or campaign that can hand a general a different profile. It is described here so the tactics are on the page when that lands.

A screening general opens the battle by pushing its light troops out in front of the heavy line rather than keeping them in it. The archers move out to a contest line about eight metres ahead of that line — one ranged reach, so their fire covers the ground immediately in front of it — hold there in the skirmish stance, and shoot at whatever comes into range. Their own heavy blocks do not break formation to rescue them. Once the opposing line closes to within about six metres of the screen, the same subcommander pulls it back — the archers break off the exchange, turn about, and fall back on a diagonal aimed at a gap between two of their own blocks rather than straight back through one (Figure 67).

Two things will follow for you. Chasing the screen with your own heavy line pulls that line forward out of dressing and hands the enemy blocks a ragged target, so it is usually better to walk up steadily and let the screen retire. And the moment the screen turns is the moment it is most exposed: cavalry sent wide, or a unit already on that flank, can catch it in the open before it reaches the interval, because its own line will not come out to save it.

Figure 67: An enemy archer unit walks out ahead of its own heavy line, holds the ground between the armies as the player line advances, then breaks off and falls back toward the interval between two blocks while the line behind it holds its ground.

---

Source: README.md · combat math lives in scripts/Unit.gd (_strike(), _flank_multiplier()).