Exact engine (port of wos-sim.ratme.org, matches ~540 real fights). Enemy: 50/20/30, this gen's three heroes, balanced build at my average (inf def/hp and every attack/lethality 1.3× the rest); Gen 1–5 only, later gens still vs a flat-stat enemy; their garrison has 4× Patrick joiners, their rally 4× Jessie. T10 troops both sides. Each focus plays its standard ratio: balanced 50/20/30, marksman 50/5/45, lancer 60/40/0. 1 pt = ×1.25 attack.
Points the hero adds, skills only (no base stats). Joiner: skill 1, in my rally. Bear: damage over a rally with no heroes. Lead is always in the fight the hero's widget (4th skill) works in: rally, garrison, or field if it has none (shown under the name), so its 50/20/30 column equals the skill breakdown's total. Click a column to sort, a hero for its skills: the in-game text, how ratme models it, and what each is worth alone.
Top formations head to head
No heroes, same average stat, each on its natural build (50/5/45 and 50/10/40 marksman, 60/40/0 lancer, the rest balanced). Points the row is ahead of the column.
Best answer to each
Of these five, which beats it by the most (no heroes, natural builds).
Marksman build vs lancer build, by gen
Each with its best heroes and joiners for that gen and fight, vs the same average enemy: marksman = 50/5/45 with marksman-heavy stats, lancer = 60/40/0 with lancer-heavy stats. Field / rally / garrison: points over the enemy. Bear: damage, shown as points of attack (ln ratio / ln 1.25). Δ > 0 = marksman ahead. Rows where FC changes are marked.
Which stat to raise
+10% on one stat, scaled so +10% attack on every type = 1.00. Attack and lethality are worth exactly the same, and so are defense and health.
FC levels
Points vs FC0 (stats and skills). Rows for one type hold the others at FC5.
Troop tiers in a rally
Points over T10 FC10, rally vs garrison. T12 includes its native skills.
How a fight works
- Rounds. Both sides fire each round from the same snapshot (attacker first, then defender). Each troop type hits the first living enemy row: infantry, then lancers, then marksmen. Losses land at the end of the round. Overkill is wasted: it doesn't spill into the next row. Fights here last rounds.
- Damage per hit ∝ √(your troops of that type) × √(the smaller army) × attack × lethality ÷ (their defense × health) × skills. So attack = lethality and defense = health exactly, and doubling troops is worth ×1.41 per round.
- Skills: same category adds, different categories multiply. Four Jessies are 4 × +25% lethality = +100% = ×2. Patrick (+25% health), Ling (−20% enemy attack), Lumak (−20% enemy lethality) and Norah sit in four categories: ×1.25 × 1.2 × 1.2 × … That's why a mixed joiner set beats four copies of the "best" one.
- "−X%" is secretly "+X%". "Damage taken −20%", "enemy attack −20%": damage is divided by 1.2 (×0.83), not multiplied by 0.8. Small reductions are about what they say; −50% is ×0.67, not ×0.5.
- Chance skills count at their average (40% × +50% = +20%).
- "The target" means one troop type: the row being hit, usually the enemy's front infantry, which deals little damage. Debuffs on the target are worth far less than the same debuff on all enemies.
- Hero base stats are big. Leading a hero adds that generation's stats to its own troop type (Gen 1 +200% attack/defense, +50% lethality/health; Gen 5 +444% / +111%). Stats are reported with your best heroes, so leading an older hero costs the difference, on that hero's troop type only.
- A slot whose troop type is missing is a buff slot. At 60/40/0 the marksman hero's own-troop skills do nothing; only all-troop buffs count, and its stat cost hits troops you don't have. That's why odd heroes (Zinman, Patrick) show up there, tied with many others.
- Joiners bring skill 1 only; the top 4 by skill level count.
Every hero: what the game says vs what the engine does
In-game text, then ratme's model of it, then what that skill alone is worth (50/20/30, field, 1000% stats). Notes flag where the two disagree or the value surprises.
Intuitive = bucket math, no fight sim: same bucket adds, different buckets multiply, 1 pt = ×1.25. Engine-calibrated (default): each skill scaled so that alone it's worth what the engine says (Heroes tab, 50/20/30), shown as face % → counted %; skills that aren't a bucket (dodge, stacking, hits on every row) get their own row. Rule of thumb: face × a rough uptime (chance, every-N, duration). Counted: effects on all troops; our damage on "the target" (the row we hit); effects on some of my troop types, at that type's value share (Stats tab: e.g. infantry deals ~0–16% of the damage). Not counted: defensive debuffs on "the target" (only the row we hit), shields, troop skills, overkill, which row dies first, and the enemy's own skills. Use it to see why; trust the sim for which.
Intuitive vs sim
Best joiner for a slot
What's in each bucket
Which skills share a bucket
Inside a bucket the percentages add; different buckets multiply. So a skill in an empty bucket is worth its face value, and one next to a big number (Hector's +200%) is worth little: +25% on top of +200% is ×3.25/3 = ×1.08, not ×1.25. Label = hero, skill number (1 = joiner skill, W = widget), max-level value. Gen 1–5 and epic heroes; troop skills up to T11 / FC5.
Before the buckets: your fight stat = permanent stat (gear, charms, research, War Academy, hero stats) × each fight buff (city buff, chief boost, pet, widget; same widgets merge) ÷ each debuff.