What you are watching
Halden Valley is an automated, fictional battle on a 400 × 400 abstract-unit map. Granite defends the village on the north bank; Ember approaches from the south. The map is generated terrain, not a real place or a reconstruction of a historical battle. You configure a scenario and observe it; clicking units inspects them rather than issuing orders.
Start a run
- Open Parameters. Choose shared conditions under General, then a strategy and resources for each side.
- Select Apply & restart. This discards the current run and starts a fresh, paused scenario.
- Press Play. Use 1×, 2× or 4× to change playback speed. Pause to inspect a unit or compare supplies.
- Use Replay to restart the same seed and settings. Change one parameter at a time for a useful comparison.
Optimise
Optimise searches the selected side’s strategy and starting combat force mix using a genetic algorithm. It starts from applied parameters, includes the current setup, and keeps the opposing side and all other settings fixed. Initial forces can be reallocated within the original weighted game-point budget and existing parameter limits; the panel lists every weight. Population is candidate count per generation; generations is the number of breeding/evaluation rounds. Training seeds are shared across candidates. The best training candidate and original setup are compared on separate validation seeds. Wins are the primary fitness measure, with objective progress and remaining strength as tie-breakers. More samples reduce noise but take longer. No global optimum or improvement is guaranteed. Review the candidate and validation before Apply candidate & restart; closing or Stop leaves your parameters unchanged.
Repeated runs
Run until stable pauses the displayed battle and opens an accelerated batch tool. It holds the currently applied parameters fixed and varies the seed. Stability means the attacker win-rate estimate meets the selected uncertainty tolerance and recent-check consistency rule; it does not mean identical outcomes, the best strategy, or convergence of every metric. Start batch, watch the completed-run count, and Stop at any time. Closing cancels work. A run-limit result is explicitly inconclusive. The panel explains the minimum sample, check interval and uncertainty assumptions, and reports remaining strength, losses by type, supplies, crossing delays, bridge work and combat activity with means and observed ranges. Only win rate controls stopping; the other metrics need not have converged. Rendering is omitted, but physics and simulation time steps are unchanged. Parallel workers selects how many independent simulations run at once; automatic uses a conservative device limit, while up to eight can be requested for throughput. Exact sight-check caching reduces repeated work without changing the model.
How the battle ends
Attackers must reach 100% village capture with ready infantry, hold a functioning north-bank supply point, have delivered supplies there by truck, and still have stock at that point. Capturing the village alone is insufficient. Defenders win when the scenario time limit expires without those conditions, or when no living mobile ground combat attackers and no scheduled reinforcement waves remain.
To establish the north-bank point, after 40 simulation seconds attackers need at least two ready mobile ground combat units, including infantry, within 33 map units of its designated location. No ready defender mobile ground combat unit may be within 38 units. This must last for the configured security time; interruption resets the timer. The new point starts empty. The previous forward point becomes a south-bank relay.
Two clocks, different meanings
Elapsed / limit is simulation time in minutes:seconds. The default limit is 7:00 (420 seconds). Day N · HH:MM is a separate, accelerated daylight clock, starting at 07:00. By default, 360 simulation seconds represents a full 24-hour day: one simulation second advances this clock by four minutes. Thus a seven-minute scenario spans 28 displayed hours. These are game scales, not real-world durations. Both clocks pause together and advance faster at 2× or 4×.
Drawing on the rendered image
Select Draw line and drag across the battlefield. Lines remain visible on the image; Undo and Clear lines remove them. Done drawing or Escape restores camera interaction. These are screen annotations, not movement orders or terrain-anchored routes. Save image exports the battlefield with the lines as a PNG.
Controls and navigation
| Control | Meaning |
|---|
| Play / Pause; 1× / 2× / 4× | Run or pause; speed multiplies simulation time, not weapon statistics. |
| Replay / Run again | Reset time, units, stocks, bridges, wrecks and outcome using the selected seed and parameters. |
| Weather | Changes conditions and restarts the same seed, paused. |
| View | Observer shows all units. Defender/Attacker contacts restrict enemy display to current team sightings. It does not change combat rules. |
| Seed / New seed | An integer from 1 to 2,147,483,646 controls the model’s seeded randomness. New seed restarts with a different sequence. Visual effects also use unseeded randomness; compare model outcomes, not identical smoke. |
| Parameters / Cancel / Defaults | Opening pauses playback. Cancel discards unapplied edits; closing restores the previous playback state. Defaults fills the editor with baseline values; Apply is still required. Settings are stored in this browser when storage is available. |
| Camera buttons / Follow | Move the viewpoint or track an inspected moving unit; no effect on its orders. Drag to orbit, right-drag to pan, scroll to zoom; touch gestures control the camera on mobile. |
| Supply lines / Place names / Unit markers | Visibility layers only. A drawn line or radius does not guarantee a delivery or a clear route. |
| Forces | On narrow screens, opens the force and supply panels. |
| Keep viewing | Dismisses the result overlay; it does not resume an ended battle. |
Force and supply panels
| Field or number | Exact interpretation |
|---|
| Large force percentage | Living combat-unit health divided by maximum health of all combat units entered into that side’s roster, including losses. It is health-weighted strength, not percent of soldiers alive. Future reinforcements are excluded until they arrive; arrivals can change the denominator. |
| Unit counts: alive / total | Living roster units of that type / total entered so far, including destroyed units. Infantry means squads; Missiles means launchers, not remaining rounds. Rail-convoy trucks are excluded from the roster truck count. |
| Suppressed / reorganizing | Current living combat units with suppression ≥70 / cohesion ≤25. These counts can overlap. Only infantry is fully pinned at suppression ≥70. |
| Enemy contacts | Currently seen enemy units; not the enemy’s total strength. Contacts refresh about every 0.7 simulation seconds. Unseen stocks and force totals are hidden in contact views. |
| Rail reserve / next convoy | Unshipped finite supply points (SP) / countdown to the next convoy attempt. Each rail convoy truck can carry 110 SP. A depot loss stops incoming dispatch. |
| Depot / Received | Current rear stock / cumulative SP received at the depot. Received is a running total, not usable stock. |
| Loaded forward / returning | Local fleet trucks on outbound supply trips / returning trips. Loading and unloading trucks are not counted in those two numbers. |
| Forward point / Delivered | Current forward stock / cumulative SP delivered to forward facilities. Delivered is not remaining stock and can include the south-bank relay. |
| Front line: connected / total | Eligible living ground units with an accessible stocked local supply point / eligible living ground units. Aircraft, trucks and infrastructure are excluded. |
| Front line: % ammo | Unweighted average ammunition fraction across living roster combat units that carry ammunition, including eligible air assets. It is not the percentage of all rounds remaining. |
| Reinforcements / waves remaining | Undeployed tank, IFV and squad budgets / scheduled waves still pending. Not extra units already on the map. |
| South-bank relay / north-bank deliveries | Remaining relay SP / cumulative SP actually delivered to the new north-bank point. |
| Supply dot colours | Depot green ≥300 SP, amber ≥100, red below 100; forward point green ≥120, amber ≥40; average ammo green ≥60%, amber ≥30%. Red is a warning, not necessarily zero stock. |
Selected unit inspector
| Field | Meaning |
|---|
| Health | Remaining health / type maximum, shown as a percentage. Zero destroys the unit. Supply does not repair health. |
| Cohesion | 0–100 measure reduced by damage and nearby friendly losses. At ≤25 the unit stops normal movement and fire to reorganize. |
| Steadiness / Suppression | Steadiness = 100 − suppression. Suppression is pressure, not physical damage; it slows movement and firing and decays over time. ≥70 pins infantry. |
| Ammo / Fuel | Percent of that type’s capacity, not necessarily a count of shots. Fuel is displayed on a 0–100 scale. Some stationary or infantry units have no fuel field. |
| Cargo | Truck load in SP. Local capacity: 45 SP; rail-convoy capacity: 110 SP. |
| Status | Current activity or obstacle: engaging, holding, loading, returning, queueing, pinned, reorganizing, out of fuel, waiting for a crossing, or steering around an obstacle. |
| Ground interaction | Whether the hull is a fixed obstacle, a live solid footprint, or a disabled hull eligible for pushing. |
| Supply | Whether the unit currently has an accessible stocked supply point under the local distribution rules. |
| Weapon range / yellow ring | Nominal type range in abstract map units (u); not a guarantee of detection or a clear line of sight. |
| Kills | Destroyed units credited to this unit. Not a count of people. |
| Terrain / protection | Current terrain classification and 100 × (1 − cover multiplier). Protection modifies damage; it is not an armour rating or guaranteed survival chance. |
| Sensor | Drones are unarmed and share team sightings. |
| Bar colours | Above 60% is green; above 30% amber; otherwise red. Inspector steadiness is the inverse of suppression. |
Objective, map status and result
| Field | Meaning |
|---|
| Capture percentage | Progress toward village occupation. Ready attacking infantry advances it only without ready defenders in the objective. Armour alone cannot capture. Both sides present means contested; defenders alone reduce partial capture. |
| Phase text | A descriptive label based on time and unit positions; not an order or a separate simulated command phase. |
| Bridge percentage / reason | Structural health on a 0–100 scale, plus policy, decision, work or traffic status. Zero closes the crossing. Construction/restoration percentages measure work progress, not health. |
| Engineering resources | Unused demolition allocations, repair kits and temporary kits. A build deducts its cost at initiation; paused work retains progress. |
| Ford status | End fords admit ground vehicles and infantry; Ferris Ford admits infantry only. All close in rain. |
| Visible wrecks | Destroyed movable vehicle hulls visible under the selected viewpoint. Not all casualties or destroyed structures. |
| Suppression events | Cumulative occasions a unit crosses from below 70 suppression to ≥70; one unit can generate multiple events. |
| Battle log timestamps | Elapsed simulation time, not the daylight clock. The display shows recent events and filters side-specific entries in contact views. |
| Supplies moved (result) | Cumulative delivered forward SP, not stock remaining or exclusively north-bank deliveries. |
| Vehicle queue time: unit-s | Sum of unit waiting time when crossing permission is refused. Ten units waiting one second add ten unit-seconds. This can include crossing closures and is not total battle duration. |
| Guided launches / lost guidance | Missiles launched / guidance-loss events. Neither is a hit or destruction count. |
| Result strength / objective / battle time | The same health-weighted strength and capture measures described above, evaluated at the end; battle time is elapsed simulation time. |
The simulation loop
The model advances in fixed 1/30-second steps. It updates sightings, scheduled arrivals and orders, unit movement and fire, supplies, projectiles, bridges, capture and end conditions. Units follow authored routes, a crossing graph and local avoidance rules. Sustained lack of progress triggers a bounded obstacle search or an alternative crossing; blocked units wait and replan. They do not use a planning AI, learn from previous runs, or calculate a complete battle plan.
Observation and combat
Terrain and buildings obstruct direct sight. Range, weather, concealment, the detection multiplier and daylight affect observation. Artillery and guided launchers require reported contacts; observer mode does not give combat units extra knowledge. Targets, firing intervals, hit chances, cover, damage and suppression use authored type statistics and seeded random variation. Some area effects can harm friendly units. The same minimum-range, maximum-range and target-eligibility checks apply when acquiring, tracking and firing.
Drones patrol and report sightings. Aircraft use fuel and ammunition and return to a surviving rear depot for servicing; a rearm cycle costs 35 SP. Anti-air units target airborne contacts. Guided missiles losing their contact report for 1.5 seconds continue toward the last aim point. These are simplified game behaviours, not equipment specifications.
Movement, collisions and wrecks
Slopes, woodland, rain, suppression, targets and fuel affect movement. Ground hulls cannot pass through buildings, trunks, infrastructure, parapets or other hulls. Rotation and movement are checked in small increments. Local avoidance can still jam in congested areas; the bounded planner may still find no feasible route in dense congestion, in which case it waits and retries. Destroyed vehicles remain obstacles. Powered tanks can slowly push suitable disabled hulls; IFVs can push lighter ones. Push mass limits, obstacles and fuel can stop a push. Infantry represents a squad footprint. Barrels, foliage and low earthworks are not full collision bodies.
Crossings
Mill and Station bridges serialize vehicle traffic; infantry is exempt from that vehicle queue. Damaged decks below 60 health reduce movement to 55% of the otherwise calculated speed. Zero health closes a crossing and removes ground units caught on its span. Destroyed bridges trigger alternative-route selection. West and East End Fords admit all ground types at 45% movement speed near the water; Ferris Ford admits infantry only. Rain closes every ford.
Deliberate defender demolition requires an allocation, a continuously detected nearby threat, a policy permitting denial, and clearance of friendly traffic. The cautious policy keeps another bridge open from deliberate demolition; combat damage can still destroy it. Headquarters loss triples the demolition delay. Defender strategies can override the policy or halve the delay.
Permanent restoration costs one repair kit and 60 rear-depot SP. Nearby ready ground combat support and absence of nearby enemy ground combat units are required for progress. Defenders under a denial policy do not restore crossings. Attacker temporary construction is preferred when eligible: one temporary kit and 80 rear-depot SP buys a green replacement deck at an existing destroyed bridge site. Ready approach-bank support is required; close enemies on that bank pause construction, but far-bank defenders do not automatically veto it. Work retains progress while paused and costs are not refunded. Temporary decks carry all ground units at 55% speed, can take combat damage, and cannot be remotely demolished by the defender policy. They are not freely placed new bridges.
Supply and reinforcement
Finite rail reserves feed rear depots through convoy trucks. Local trucks move stock to forward points. Local distribution checks bank, radius and enemy ground interference along the connection; it is abstract and does not trace a physically navigable delivery path around every obstacle. Being inside a supply ring alone does not guarantee access. Ammunition and fuel consume SP; health is never replenished. At the baseline distribution rate, each half-second update can add up to 2.5% of ammo capacity and 2 fuel points, limited by available stock. A complete ammo refill costs 60 SP and a complete fuel refill 20 SP.
Local truck losses can be replaced toward the configured fleet target: the replacement check requires at least 60 depot SP and spends 50 SP. Rail convoy trucks are separate. Reinforcement totals are divided across the selected number of waves; they are not amounts per wave. A destroyed rear depot cancels that side’s pending reinforcements when a wave is attempted. Waves scheduled after the battle ends never arrive.
Time, weather and performance
Daylight smoothly changes scene lighting and detection. Full-night detection is multiplied by Night detection, equally for both sides; there is no thermal or night-vision equipment model. Rain multiplies detection by 0.78 and off-road movement by 0.58. Fog uses 0.5 and 0.9 respectively. Dry weather uses 1. These multiply other applicable modifiers.
Paused scenes redraw on interaction. Mobile rendering targets 30 FPS with simpler terrain and fewer visual effects; simulation steps and collision rules remain the same. Hiding the page suspends updates without catching up later. The standalone file bundles its 3D engine and controls. Actual phone performance depends on the device.
Reading this reference
These are factory defaults, not necessarily your saved settings. General values apply to both sides. Side defaults are listed as Defender / Attacker. A multiplier of 1 means baseline; 0.5 means half and 2 means double the relevant rate before other modifiers. All seconds below are simulation seconds, not accelerated clock minutes. Changing parameters requires Apply & restart.
| Condition | Default / allowed values | Meaning |
|---|
| Weather | Dry / dry, rain, fog | Visibility, off-road speed and ford availability; see Model rules. |
| Seed | 1337 / integer 1–2,147,483,646 | Seeded variation. Keep fixed when comparing changes. |
Every General parameter
| Parameter | Default | Range; increment | Meaning |
|---|
| Temporary bridge build time | 30 s | 10–180; 5 | Uninterrupted work time at a destroyed crossing. Near-bank ground support is required; enemy presence on that bank pauses construction. |
| Starting hour | 7 h | 0–23; 1 | Scenario clock at the start; fictional local time, not geographic solar time. |
| Full day duration | 360 s | 60–1800; 30 | Simulation seconds per 24-hour cycle. Pause and playback speed also control this clock. |
| Night detection | 0.5× | 0.2–1; 0.05 | Detection at full night relative to daylight, applied equally to both sides. |
| Defender bridge policy | 1 | 0–2; 1 | 0: preserve; 1: deny only with another crossing open; 2: permit denying both. Friendly traffic and detected threats still govern the decision. |
| Demolition decision delay | 12 s | 3–60; 1 | Continuous detected threat required; loss of headquarters triples this delay. |
| Crossing restoration time | 45 s | 15–180; 5 | Uncontested work time with a ready friendly ground unit at an approach. Abstract engineering, not real-world timings. |
| Detection distance | 1× | 0.5–1.6; 0.05 | Scales all sensor distances before weather and concealment. |
| Movement speed | 1× | 0.5–1.75; 0.05 | Scales ground and airborne movement; fuel still costs distance. |
| Damage | 1× | 0.3–2; 0.05 | Scales hit and blast damage without changing health. |
| Fire rate | 1× | 0.4–2; 0.05 | Higher values shorten the interval between shots. |
| Suppression effect | 1× | 0.25–2; 0.05 | Scales the pressure caused by nearby impacts. |
| Recovery rate | 1× | 0.3–2; 0.05 | Scales suppression decay and cohesion recovery. |
| Local distribution rate | 1× | 0.25–2; 0.05 | Scales ammunition and fuel transfer rates; stock costs remain conserved. |
| Reserve command delay | 6 s | 0–30; 1 | Delay after a threat is observed; headquarters loss triples it, with a minimum of 12 seconds. |
| Base capture time | 45 s | 20–120; 5 | Base time modified by steady infantry presence in the village. |
| Bridgehead security time | 12 s | 3–45; 1 | Continuous uncontested hold required to establish the supply point. |
| Scenario duration | 420 s | 180–900; 30 | Maximum simulation time before the outcome is assessed. |
Policy precedence: Keep crossings open forces policy 0; Crossing denial forces policy 2 and halves the decision delay. Early reserve halves the configured reserve delay; Village reserve multiplies it by 1.5. Headquarters loss then triples the effective reserve delay with a 12-second minimum. Capture time is a base value, modified by ready units; it is not a guaranteed completion time. Lower Full day duration makes daylight change faster.
Every side parameter
| Parameter | Default: Defender / Attacker | Range; increment | Meaning |
|---|
| Strategy approach | Balanced defence / Balanced advance | Five named choices per side | Behaviour preset; keeps your chosen force counts and supply budgets. |
| Temporary bridge kits | 0 / 3 kits | 0–8; 1 | Attackers automatically build temporary decks at destroyed crossings. Each costs one kit and 80 rear-depot SP. Defender kits are reserved and unused. |
| Bridge demolition allocations | 2 / 0 allocations | 0–6; 1 | Finite abstract demolition budget. Only defenders with a denial policy use these allocations. |
| Crossing repair kits | 1 / 3 kits | 0–8; 1 | Each restored crossing consumes one kit and 60 rear-depot SP. Engineering teams are abstracted. |
| Infantry squads | 8 / 8 squads | 0–24; 1 | Initial squads; later arrivals are configured separately below. |
| Tanks | 5 / 8 vehicles | 0–16; 1 | Defenders deploy up to four on the line and hold additional tanks in the local reserve. |
| Infantry fighting vehicles | 2 / 5 vehicles | 0–12; 1 | Defender IFVs begin in the local reserve; attacker IFVs move with the assault. |
| Artillery guns | 3 / 4 guns | 0–8; 1 | Guns start at the existing rear gun lines. |
| Fortified bunkers | 3 / 0 positions | 0–6; 1 | Defender bunkers guard the river; attacker bunkers protect the south-bank rear. |
| Reconnaissance drones | 1 / 1 airframes | 0–3; 1 | Unarmed reconnaissance patrols belonging to this side. |
| Strike aircraft | 1 / 1 airframes | 0–3; 1 | Aircraft use ammunition, fuel and depot supplies from their own side. |
| Guided-missile launchers | 1 / 1 launchers | 0–3; 1 | Launcher count; ammunition load is set below. |
| Anti-air units | 2 / 2 units | 0–3; 1 | Units engage detected aircraft and drones. |
| Ammunition at entry | 100 / 100 % capacity | 0–100; 1 | Starting load for initial units and reinforcements on this side; resupply uses SP. |
| Fuel at entry | 100 / 100 % capacity | 0–100; 1 | Starting load for tanks, IFVs and air units; low-fuel aircraft wait at base to refuel. |
| Rear depot stock | 500 / 800 SP | 0–5000; 25 | Available immediately at the rear depot; also funds aircraft rearming and truck replacements. |
| Forward point stock | 150 / 200 SP | 0–2000; 25 | Available immediately to accessible front-line units. A new north-bank point still starts empty. |
| Rail reserve | 2600 / 3600 SP | 0–10000; 50 | Finite external supply budget delivered by railhead convoys. Zero disables incoming supplies. |
| Local supply-truck fleet | 3 / 4 trucks | 0–8; 1 | Initial fleet and replacement target. Replacements cost 50 SP; railhead convoy vehicles are separate. |
| Rail convoy interval | 34 / 26 s | 10–120; 1 | Spacing between two-truck railhead convoys, while rail reserve remains. |
| Reinforcement tanks | 2 / 9 vehicles total | 0–12; 1 | Total across all waves, in addition to the starting force. |
| Reinforcement IFVs | 0 / 6 vehicles total | 0–12; 1 | Total across all waves, in addition to the starting force. |
| Reinforcement infantry | 2 / 9 squads total | 0–24; 1 | Total across all waves, in addition to the starting force. |
| Reinforcement waves | 1 / 3 waves | 0–4; 1 | The finite budget is split across these waves. Zero disables all reinforcements. |
| First reinforcement arrival | 110 / 70 s | 20–600; 1 | Simulation time when the first wave enters; arrivals after the scenario limit do not deploy. |
| Time between waves | 60 / 60 s | 20–180; 1 | Spacing between subsequent reinforcement waves. |
Resources summary: initial combat/recon units adds configured combat assets, excluding fixed infrastructure and trucks. Initial forces and reinforcement totals are separate. Total SP sums starting depot, forward and rail stocks; engineering kits are separate. Ammunition and fuel entry percentages also apply to later combat reinforcements. Zero waves disables the reinforcement budget. Defender temporary kits and attacker demolition allocations are currently unused; setting them does not unlock an extra behaviour.
Five attacker approaches
| Approach | What changes |
|---|
| Balanced advance | Split forces across the existing routes; initial columns depart at 14 seconds. |
| West crossing focus | Initial forces and reinforcements prefer Mill Bridge; unavailable crossings use normal alternatives. |
| East crossing focus | Initial forces and reinforcements prefer Station Bridge; unavailable crossings use normal alternatives. |
| End-ford approach | Use the nearest end ford from the outset; rain switches units to bridge routes. |
| Staged advance | Infantry depart at 14 seconds, IFVs at 24 and tanks at 34; later waves use the same 0/10/20-second spacing. |
Five defender approaches
| Approach | What changes |
|---|
| Balanced defence | Use the configured reserve delay and bridge policy; reserves react to detected north-bank threats near the village. |
| Early reserve | Commit available reserves when a detected attacker approaches within 100 units of the village; use half the configured delay. |
| Village reserve | Keep reserves back until a detected north-bank attacker is within 30 units of the village; use 1.5 times the configured delay. |
| Keep crossings open | Disable deliberate bridge demolition; retain normal reserve response. Combat can still damage bridges. |
| Crossing denial | Allow deliberate denial of both bridges and use half the configured decision time. Friendly traffic checks and finite allocations still apply. |
Selections are independent and saved with parameters. They change behaviour, not force counts, ammunition or budgets. West/east focus is a preference, not a guarantee that a route is open. End-ford approach uses bridges in rain. Staged advance delays departure of existing units; it does not create extra reinforcement waves. Reserve approaches only affect units already designated as reserves, not every defender.
What this model does not claim
This is a fictional sandbox with abstract map units, compressed time and generic units. It does not predict real combat outcomes or casualties. History and doctrine motivated earlier design ideas, but the numerical settings and strategies are authored game rules, not validated doctrine or measured equipment performance.
There is no real-map importer, individual engineer simulation, civilian model, medical evacuation, comprehensive communications network, adaptive commander AI, remembered enemy-track model, or general building-destruction system. Sensors and air systems are simplified. A real map would require new terrain, routes, passability and scale handling; changing a label would not provide that.
Exporting and reproducing results
Batch JSON and optimisation JSON include the engine version, parameters, exact seeds, settings and raw results. Partial exports identify unfinished work. CSV contains per-run batch metrics; retain JSON alongside it for the complete configuration. Results remain in memory until refresh, so export before closing the file. Engine versions change when model code changes; comparisons across versions may differ.
Interpreting a stalled run
Pause and select the leading unit. Check its status, fuel, steadiness, cohesion and nearby wrecks. Then inspect bridge health/work progress, available engineering kits, depot stock, ford weather and the chosen strategy. Some congestion and repair conditions can cause stalemates. Defender victory at the time limit does not establish that the chosen defence would work in the real world.
Comparing settings fairly
Use the same seed, weather, time limit and resource budgets, changing one factor at a time. Replay resets the battle. Try multiple seeds before drawing a conclusion. Compare capture, north-bank supply deliveries, remaining strength and consumed supplies together; no single percentage describes success. The guide documents implemented rules, including their limitations, rather than proposed future features.