Three fights. Same enemy. One hero swap each time. The 268% buff you never noticed, the burst skill that halved the fight, and eight lies the game pretends are true.
Delta96 — ran the test·Delta96 & CerealKiller — did the math·16 April 2026·Server 497
// Three fights, side by side
Same enemy. One swap per round.
Lv.51 Sombie. Same base, tech, equipment, mod vehicle, Lv.9/10 Fighter troops. One hero slot moved each fight.
F1 — Barnett
Tristan · Francis · Barnett · Guy · Farhad
Ended round 41
2,700
Troops left · 28%
F2 — C&R (winner)
Tristan · Francis · C&R · Guy · Farhad
Ended round 34 — 7 rounds shorter
4,237
Troops left · 43%
F3 — Megan (no Farhad)
Tristan · Francis · C&R · Guy · Megan
Ended round 41
3,061
Troops left · 32%
Stat
F1 Barnett
F2 C&R
F3 Megan
Start troops
9,722
9,752
9,469
HP pool
2,705,851
2,714,201
2,635,436
ATK%
916.4
932.7
921.9
DEF%
934.5
966.7
977.3
Hero NAD
905%
1034%
1028%
Hero power
7.4M
7.9M
7.5M
Faction bonus ATK
68.6%
72.2%
50.3%
Ended at round
41
34
41
Troops left
2,700
4,237
3,061
HP lost
624,780
489,500
569,600
F1 · buff overviewF2 · buff overviewF3 · buff overview
Verdict
F2 won. Clear. Not close.
Swap Barnett → C&R: saved 7 rounds, 1,537 troops. Swap Farhad → Megan: added 7 rounds back, cost 1,176 troops. Same vehicle. Same tech. Same equipment. Every round of gap lives in the hero slot.
C&R hits ~2× Barnett. That is the whole ballgame for F1 → F2. In F3, Megan's Divine Prayer (~20k) is weaker than Farhad's Eagle Strike (~24k) and drops the faction bonus from 72% to 50%. Double loss.
Kenneth's first hypothesis
"Do purple hero troops die faster?"
No evidence. Troops are pooled (proven in kb/damage-system.md). All heroes feed one shared HP bucket. "Your purple hero's troops" is an illusion the roster screen creates.
What a hero actually changes: how many troops you can bring (capacity), how much buff pours into the pool (stack), how big their R11/R21/R31 burst is (spike). That's it.
Data says yes, usually — with a sharper rule: In-faction > Star level > Rarity. Burst damage can override all three, but only when the skill multiplier is ~2× the hero it replaces.
If your swap hero isn't in faction, check one thing: is their burst skill 1.8×+ bigger than the one you're dropping? If not, don't swap. Faction bonus stacks on the whole pool every round. A shiny out-of-faction gold loses ground every single turn.
Passive first. ~1% per book. Multiplies the entire pool. Every turn. Highest leverage.
Active second. ~42pts per book. Fires R11/R21. C&R's Go Rex Go = ~1,800 HP bigger hits per book.
Normal attack last. ~4pts per book. Tiebreaker only. Skip unless maxed elsewhere.
// Section 2 · Vehicle
Vehicle: the 268% nobody talks about.
Mod Vehicle · 5.5M · identical all three fightsMod Vehicle line in the buff stack
268.1% ATK. Same every fight. Formation-agnostic.
268.1% DEF. Same every fight.
15% HP, 15% DMG. Bonus on top.
Biggest single source on the ATK/DEF stack. Bigger than heroes. Bigger than tech. Bigger than equipment.
Always on. Every round. Every skill. Every normal attack.
Actionable
Finish your mod parts before chasing a new hero above 4★.
Whatever mod tier you're at, completing the set scales every skill line and every normal attack. No formation dependency. No roster decisions. Pure multiplier.
// Section 3 · Hero Equipment
Equipment is a bucket. Not a costume.
F1 · Tristan · Francis · Barnett · Guy · FarhadF2 · Tristan · Francis · C&R · Guy · FarhadF3 · Tristan · Francis · C&R · Guy · Megan
ATK sub-buff: 159–174%. Shifts slightly when hero mix changes — pool is re-computed, not re-assigned.
DMG sub-buff: 112.4%. Identical in all three formations.
DEF sub-buff: 161–174%.
Rain Fire ×2 lands identical numbers inside each formation. Tristan's gear did not fire twice. The pool did.
Who wears the gear is irrelevant for skill damage.
Actionable
Upgrade your best D5 set on whoever's easiest. Don't hoard gear on "main hero".
The game doesn't read "this is Tristan's helmet." It reads "+X ATK in the bucket."
// Section 4 · Special Weapons (Exclusive Equipment)
Tristan's cannon fires Rain Fire twice. Free.
F2 · R11 · Rain Fire fires, then fires againFrancis's Requiem · Shield every round post-R11
That's a Guy Blade Storm's worth of damage. Every skill round. Free.
Francis's Requiem shields every round post-R11. You'll see "Requiem! gained Shield effects" in R12–R15, R22–R25, R32–R35. That shield absorbs hits that would otherwise kill troops.
Actionable
Star up Tristan's Twin Ballistics if the path exists. It's doing real tonnage.
// Section 5 · Troops
We had half their strength. We won three times.
Us: Lv.9/10 · ~9,500 troops · 5.1–5.2M Unit Strength.
Enemy Sombie: Lv.10 · 13,600 troops · 10.9M Unit Strength — 2× our strength per unit.
Enemy Chief Zombie: 1581.9% ATK and DEF from heroes. Nothing else. No tech. No vehicle. No equipment. No season. No faction.
F1 · Victory · 2,700 leftF2 · Victory · 4,237 leftF3 · Victory · 3,061 left
Correction — the big one
More troops = wins fights. TRUE.
But only if your stack is in the fight. Sombie had 2× our troops and still lost three times because their buff stack was a single number (1581.9% heroes) against our multi-layer stack (~1,800% effective).
Read it this way: stack × troops = damage. Kill either side of the equation and you lose. We won because our 1,800% on 9,500 troops outpunched their 1,581% on 13,600. Feed both sides of the equation.
// Section 6 · Research
Research: 180% ATK you can't see.
Tech ATK: 180.4%. Identical all 3 fights (account-wide).
Tech DEF: 182.9%.
Tech HP: 7.2%.
Tech DMG: 7.8%.
3rd-biggest ATK/DEF source. After heroes and vehicle.
Faction-specific tech adds ~11% more for concentrated troops (per kb).
Actionable
Tech grows silently. Keep it fed.
It can't be tested marginally from one dataset, but its size tells you it deserves continuous investment — especially faction tech if you've already picked a lane.
// Section 7 · Buffs overall
Where damage is actually made.
Ranked by contribution to Troop Total ATK (F2 numbers):
F2 reached round 31 earlier. Finished round 34. F1 and F3 ground to 41.
Every skill round you reach earlier = fight ends earlier = fewer troops lost. That's it. That's the whole story.
Degradation: all damage drops ~3.4% per 10 rounds. Eagle Strike drops ~16% per 10 — the Farhad anomaly. Still unsolved. See kb/unsolved/.
// Lies the game quietly tells you
Eight illusions. Now exposed.
"Each hero commands their own unit."False. One pool. One blob. Equipment distribution irrelevant for skills.
"Gold hero better than purple hero."Conditional. In-faction fit + passive books + burst size matter more than rarity color.
"Star level is the power driver."Partial. Passive skill books produce more pool-multiplier than the star itself.
"Heroes are the biggest buff source."Tie, not win. Mod Vehicle ≈ Heroes on ATK/DEF.
"More troops = wins fights." · (we previously called this false)True. But only if your stack is in the fight. Sombie's 2× troops lost because their stack was 0%. Feed both sides of stack × troops.
"Vehicle is just cosmetic."False. 268% ATK, 268% DEF, always on. Biggest single source.
"Tristan's second Rain Fire is a visual effect."False. Twin Ballistics fires a real second shot. ~32k HP free, every skill round.
"Skin buffs matter in combat."Barely. 5%. Bottom of the stack. Cosmetic with a sticker.
// Do this tomorrow
Eight moves. In order.
Concentrate one faction. Whatever you play, don't mix. Highest-ROI change you'll ever make.
Fix your Faction Bonus number. F3 dropped that from 72% to 50% and paid in blood. Any out-of-faction hero better be worth the trade.
Upgrade passives before stars. ~1% per book, compounds through the whole pool.
Test designed and executed by [-qp-]Delta96 on Server 497, 16 April 2026. Three sequential battles against the same Lv.51 Sombie, with full pre-fight buff stacks, round-by-round combat logs, and battle reports captured across 130+ screenshots. Analysis, burst-damage reading, and mental model by Delta96 & CerealKiller. Raw source data in /kb/reference/test-reports/.
Known unresolved: Eagle Strike degrades at ~16% per 10 rounds while every other skill sits at ~3.4%. The Farhad anomaly. No confirmed cause. Flagged in kb/unsolved/.