Absy Labs ABSY LABSKingshot Theorycraft Launch the simulator

Mechanics · what the engine reads

What actually counts: rally, garrison and troops

Two things decide a fight before a single number is rolled: who in your rally actually contributes, and the real base stats your troops carry. The game UI shows you neither cleanly. Here is what counts, with each claim graded for how sure we are.

Confirmed Probable Speculative every claim below carries one.
If you only remember a few things
  1. Your 3 leaders do nearly all the work. Joiners add at most four first-skills on top.
  2. A joiner's gear and account stats are ignored: only its first hero's first skill and its troops count.
  3. The four contributing joiners are the highest-skill-level ones, ties broken by join order.
  4. Chance-based joiner skills do not stack: four Jabels fire as one.
  5. Troops fight with hidden base stats: base Lethality and Defense are both 10, and only Attack and Health change by tier.
  6. Cavalry bypass hits archers about 20% of rounds; archers volley for a second hit about 10%.

The reasoning for each is below, for whoever wants it.

Who counts in a rally?

The Rally Captain picks three leader heroes: one Infantry, one Cavalry, one Archer, no duplicates. Any slot can be an epic or a mythic (epics can lead). The captain stays the leader for the whole battle, even if a stronger member joins later. Confirmed.

Alliance members then join and bring troops. Only four of them also contribute a skill: the four with the highest first-skill level, with join order as the tie-break. The rest add troops only.

Confirmed Contributing joiners are picked by skill level, not by who joined first. A level-2 first skill beats an unfilled one; join order only breaks ties. The simulator deliberately models the realistic case where your top joiners are maxed.

Add it up and a fully-stacked rally is running a lot at once:

Leaders: 3 heroes × (3 expedition skills + 1 widget skill) = 12
Joiners: up to 4 × 1 first skill                          = 4
                                                    Total = 16 active skills

Plus the passive stat bonuses from the three leaders' base stats, gear and widgets. The leader trio does the overwhelming majority of the work, which is why Best Counter spends most of its search budget on the trio and treats joiners as a smaller, de-duplicated set.

What does a joiner actually bring?

  • Only the first skill of its first hero, plus its troops. No gear, no other stats, no widget, no second or third skill.
  • A joiner's gear and account stats are completely ignored. A barehanded joiner and a fully-geared one produce the identical result. A stat-granting joiner still matters, just through the hidden channel, not the battle report.
  • Diana is not a valid combat joiner: her first skill is a non-combat stamina reduction and does nothing in a fight.
Probable Chance-based joiner skills do not stack across copies: four Jabels produce one roll, not four. Non-RNG skills stack additively in their op family. The search de-duplicates identical first-skill joiners before it ever simulates. How the stacking works.

Leader vs joiner, line by line

The split between what a leader contributes and what a joiner contributes is half of building a good rally:

SourceLeaderJoiner
Tech, governor gear, charms, pets, VIP, town skinAppliedIgnored
Alliance tech, city buffsApplied (leader's)Ignored
Hero base stats (leader Atk/Def %)Applied, per trio leader, class-specificIgnored
Hero gear statsApplied (3 leaders)Ignored
Widget stat bonusesApplied (leaders' widgets)Ignored
Helga / Amadeus unique passivesApplied if in the trioIgnored even if joined
Hero expedition skills (sk1 + sk2 + sk3)All 3 per hero (9 total)First hero's first skill
Widget expedition skillApplied (leaders' widgets)Ignored
Troop countCountedCounted

Probable. Per community testing (Daryl): account-level stats, gear, charms, pets and buffs all come from the rally leader; joiners contribute only troops and up to four first skills.

Epic leaders

An epic in a leader slot contributes its leader bonus and its first skill only: no widget, and no second or third skill. The "all 3 skills, 9 total" row above is the mythic-trio case.

Garrison: same rules, one twist

A garrison works the same way as a rally: a leader plus up to four counted joiner skills. The one difference is that if a higher-power player joins the garrison, they replace the leader, who is demoted to joiner. Probable.

Confirmed The simulator does not model automatic leader replacement. You enter the active garrison leader and joiners directly, so what you simulate is exactly the formation you specify. If your real garrison swaps leaders, simulate the one that ends up leading.

The real troop base stats

The numbers in the in-game training panel are display placeholders. These are the actual base values the engine uses. Every troop, every tier, also has base Lethality = 10 and base Defense = 10: only Attack and Health vary by tier. Confirmed.

TierInf AtkInf HPCav AtkCav HPArc AtkArc HP
T6243730730243974183
T9400120012004001600300
T10472141614164721888354
T10.1491147314734911964368
T10.2515154615465152062387
T10.3541162416245412165406
T10.4568170517055682273426
T10.5597179017905972387448
T11566169916995662266390

Probable. T11 (TrueGold) base stats are in the simulator, and troops fight with their TrueGold procs across the full T1–T11 × TG0–TG8 grid. T12 (Tempered TrueGold) has its slot reserved with no values yet: it will be filled from data, not guessed.

Speculative T11 archers look like they take a survivability nerf, with arc HP / inf Atk dropping to about 0.689 versus the steady 0.75 that holds T1 to T10. This is not yet confirmed: it needs more data.

The pattern behind the tiers

The tiers are not arbitrary. One set of ratios holds across all of them, which is a useful sanity check whenever a new tier's values surface:

attack_inf  = health_cav
attack_cav  = 3 × attack_inf
attack_arc  = 4 × attack_inf
health_arc  ≈ 0.75 × attack_inf   (T1–T10; ≈0.689 at T11, unconfirmed)

Archers hit hardest per troop and die easiest; infantry are the wall. The percentages you stack on top do not change this shape, they scale it.

Cavalry bypass and archer volley

Two troop traits bend targeting and damage, and either can swing a fight:

  • Cavalry bypass (Ambusher): about 20% of the time, a cavalry squad skips the frontline and strikes archers directly. One roll per round per cavalry-attacking fighter. Confirmed ability, community-quoted rate. (In old State of Survival this was deterministic every 20 rounds; KingShot made it a per-round roll.)
  • Archer volley: about a 10% chance to fire twice in a round. Confirmed ability, community-quoted rate.

Expected mode bakes both in as fractions; Monte Carlo rolls them per round. The bypass especially can decide a fight, because archers are the squad least able to survive being hit early.

One more confirmed number while you are here: when your attacking squad counters the type it is hitting (Infantry over Cavalry over Archer over Infantry), it deals a flat +10% damage. There is no defensive type bonus, so you only ever deal more, never take less. Confirmed. Where it sits in the formula.


Related: the damage formula · hero skill stacking · every hero

Build the rally that actually wins

Best Counter searches your owned heroes for the leader trio and troop ratio that trades best against a garrison.