Troop compatibility in rallies Совместимость отрядов в штурмах
Русская версияThis guide was written in answer to a personal request. An email arrived at the site address: thank you for the materials, they turned out to be useful, but could you cover the compatibility of different troop types in more detail.
Thank you for the question. It was one of those that make you dig through your own notes and find mistakes in them — and the subject really is described nowhere properly.
If something is missing here for you too, write in: the address is at the foot of the page. Letters like that turn straight into articles.
We do not know the exact maths behind this mechanic. The game never explains it, and it is not in any public source either.
Everything below is guesswork, drawn from analysing a large number of rally experiments: different troop types, different tiers, different buffs. We went through detailed battle reports one by one, did the arithmetic and compared.
I am not prepared to take responsibility for the truth of this information. But it looks true to me, and it works in practice.
If you have data that contradicts it, send it over. We will work it out together.
1. First things first: rallies have different targets
Almost all the confusion here comes from mixing up two completely different tasks.
| Rally target | What decides the outcome |
|---|---|
| Monster Den | Stats and tier. Troop type does not affect your damage — but it does affect the group's losses, see section 7 |
| Another lord's castle | Troop type. Bring what counters his troops |
The rules differ, so the two cases are covered separately below.
2. The troop type triangle
This is the only part of the subject described anywhere at all: the triangle is laid out on the Fandom wiki (Puzzles & Chaos Fandom Wiki — an outside fan wiki about the game, not ours). We have not checked the whole table against the client ourselves.
| Troop type | Beats | Beaten by |
|---|---|---|
| Infantry | archers, vehicles | cavalry |
| Archers | cavalry, vehicles | infantry |
| Cavalry | infantry, vehicles | archers |
| Vehicles | nobody | all three types |
3. Rally damage: tier decides, not troop type
This is the most surprising conclusion from all our measurements, and it reproduces reliably.
The damage you deal in a rally depends only on the tier of your troops and how many of them there are.
Neither your troop type nor your personal buffs enter your own strike. Every participant on the same tier in the same battle deals identical damage per unit — matching to four decimal places.
Each player's contribution is weighted: army size multiplied by a tier coefficient. These are the coefficients we arrived at, normalised to T9:
| Tier | Weight | Share of a T10 |
|---|---|---|
| T6 | 0.44 | one third |
| T8 | 0.73 | 55% |
| T9 | 1.00 | 75% |
| T10 | 1.33 | 100% |
A T6 soldier is worth exactly one third of a T10 soldier. The T10 to T9 ratio came out at 1.3334 time after time, which looks like a constant baked into the game.
What this means in practice: a hundred thousand T6 in a rally is not a hundred thousand, it is thirty-three thousand in T10 terms. The power figure in the interface says one thing, the battle does another.
On how solid these numbers are. The T9 and T10 coefficients are confirmed across six or more battles, and the ratio between them reproduced again and again. T6 and T8, however, rest on a single battle each — trust them less. The order of magnitude is right, the third decimal may move.
4. Losses are shared by everyone
The loss percentage in a rally is the same for every participant. That part is measured: it matches to three or four decimal places across every battle we went through.
It looks as though the game counts losses for the group as a whole and splits them in proportion to each player's army size. But that is our reading of the observation, not the mechanic itself — we cannot see how it works inside.
Everything else follows from this: one person's bad composition is paid for by all, including the strongest in the group.
5. Who actually ruins a rally
Not on the group average, not on the strongest. On the weakest. And if there are several weak players, the effect adds up.
This is the single thing worth taking away from the whole guide.
What is actually measured. Two things. First: the loss percentage is the same for every participant. Second: add certain players to a rally and that percentage rises for everyone at once, including those whose defence is perfectly fine.
What follows is our explanation, not a measurement. We think troops poorly protected against whatever is hitting them die fast, their deaths go into the common pot, and the bill is split between everyone. It hangs together — but there is nothing to check it against: the game never shows whose troops died. It shows one final percentage, the same for all.
What can be said firmly: a weak participant does not degrade your stats — buffs do not spread across the group that way, and that was tested separately. He affects the outcome through what he brought to the battle, and everyone pays for it together.
Weakness is not measured by a player's overall power
Here are three real battles, and they show three different things.
Den 40. A fifth player joined a solid group of four, bringing infantry. His Infantry DEF was 656 — while his own cavalry was levelled to 1311. His active heroes were at levels 9, 56, 63, 10 and 10 — all below 131, which means their buffs barely work.
The whole group's losses went up 8.6 times. The kill percentage did not move at all: he affected nothing about the damage, only the losses.
Same den. A cavalry player joined an archer group — and he is not weak. He is personally stronger than the player who ran these measurements: he takes a den one level higher solo. His cavalry is levelled and its defence is 1401.
Losses still tripled. The obvious explanation is that his defence counts on the cavalry type while it is the enemy archers — the counter type — that hit his cavalry. But it is a guess, and we have nothing to check it against.
So an objectively strong player can sink the group too, if he brings the troop type that is being punished in that particular battle.
A player behind the strongest member on overall stats joined with a full march of his own levelled type, matching the group's composition, and with developed heroes. On attack for his own type he ranked second of four in the group: no star, but no hole either.
The group's losses fell by half again. More troops, shorter battle.
A caveat, so as not to misrepresent it: he cannot be called a weak player. This is not an example of weak players being harmless. It is an example of trailing on overall power doing no harm by itself — provided the man turns up with his own type and developed heroes.
What hurts is not a weak player and not a low tier. What hurts is a poorly protected troop type in that particular battle — whether because the player never levelled it, or because he did but it is out of place in this group.
The scale of the harm depends on the number of bodies
That same cavalry player, joining with a single unit, produced exactly zero effect — the battle result was indistinguishable from one without him. With a full march, he tripled the losses.
This is why "join with one unit if you are unsure" is mechanics, not politeness.
With one unit you physically cannot sink the group. And you still hand the group your best buff: it looks like the group takes attack buffs as the maximum across participants, so if you are the best in the group on one of them, a single unit of yours is enough.
Two special cases you deserve to be burned for
Both sound like common sense, and both wreck the rally for the whole group.
Here we step outside the measurements. Neither of these two cases was measured separately — this is the logic of the battles above carried over. The logic is direct, but there is no measurement under it.
The reasoning sounds sensible: "I level archers, they are strong, but they have all gone out with other rallies. I will send cavalry here so the slot is not wasted."
Never do this.
Your cavalry is not levelled, which means its defence is low, which means it dies instantly — and the bill goes to everyone in the rally. You are a strong player, but you turned up to this rally as a weak one, and the group pays for it.
Remember case two above: there a levelled cavalry force from a strong player tripled the group's losses. An unlevelled one will do worse.
"I am short on archers, they went out with other rallies — I will scrape up the odds and ends of cavalry and infantry, troops are troops."
You deserve to be burned for that. It is the worst thing you can bring to a rally.
None of those scraps is protected: you never levelled them. They all die cheaply, and every one of them adds to the common pot of losses that is split between everyone. One player like this and the group takes enormous losses.
Better to stay out entirely or join with a single unit of anything than to send a ragbag.
With one unit you harm nobody, and if you are the best in the group on some buff, the group still gets it. Staying out is fine too, nobody will hold it against you. Saying in chat "no main troops available, sitting this one out" is honest and correct.
6. Legends and myths
Plenty of confident claims circulate in the chats about this subject. Here are the ones our measurements do not support.
- "My troop type affects the damage I deal." It does not. Only tier and headcount do. Players on the same tier in the same battle deal identical damage per unit.
- "A mixed composition is bad in a den." There is nothing to counter in a den: all four troop types are there in equal numbers. A mixed bag is a normal composition.
- "A weak player ruins the group with his buffs." Buffs do not spread that way. He ruins it by bringing troops that die cheaply, while the bill is split between everyone.
- "Low tier is what kills you." It is not the tier. A full T9 march from a player with weak stats reduced the group's losses. What kills you is leaky defence on the type the player came with.
- "A strong player cannot ruin a rally." He can, and it has been measured: a player stronger than the author tripled the group's losses by joining with the wrong troop type.
- "The power figure shows my contribution." It does not. A T6 soldier weighs one third of a T10, and the power figure says nothing about it.
You often hear that "troop type incompatibility is made up, stats decide everything, troop-wide stats above all". Our data says the opposite.
The cleanest test we have: the same group took on an extra player with an army of about 227,000. A cavalry player — losses tripled. An archer of the same size — losses went down. Same headcount, opposite sign.
Alongside it: between two battles the group's defence on its troop types barely moved, while losses tripled. Defensive stats did not explain that. Troop type composition did.
Why exactly — we do not know, and we say so honestly.
7. Monster dens: there is nothing to counter
A den holds all four troop types in equal numbers. Checked against the den composition screen and consistent with the Fandom wiki (Puzzles & Chaos Fandom Wiki — an outside fan wiki about the game, not ours).
| Den | Army | Breakdown |
|---|---|---|
| 38 | 1,080,000 | 4 types of 270,000 |
| 39 | 1,200,000 | 4 types of 300,000 |
| 40 | 1,320,000 | 4 types of 330,000 |
Whatever you bring, the matchup is symmetrical: you counter a quarter of the den, a quarter counters you, the rest is neutral. Counter-picking a den does not work at all.
So bring a mixed bag. What matters is not who came with which type, but that tiers are comparable and everyone's stats are decent.
But "mixed bag" here is about the group, not about an individual player. Different people bringing different troop types to one rally is fine — each brings what he levelled. One person dragging several of his unlevelled types into the battle wrecks the rally for everyone — see section 5.
The den's own buffs are troop-wide only, incidentally: at den 40 that is 1300% to Troop ATK, DEF and HP, with zeroes on the individual types. Against multipliers like that, only your own stats carry you.
In a series of measurements at den 40, group losses rose monotonically with the share of cavalry in the composition: 26.5% cavalry gave 0.05% losses, 33.5% gave 0.06%, 34.7% gave 0.09%, 42% gave 0.16%.
A clean test: adding 227,000 cavalry tripled the losses. Adding the same 227,000 archers brought them down. It was the share that mattered, not the number of riders: two riders at a low share gave the same result as one.
An important caveat: the group was archer-based. It may be the combination rather than cavalry as such. We do not know the mechanism. The triangle suggests itself as the explanation, but why it would produce a threefold difference in losses here is not clear.
8. Another lord's castle: here troop type does decide
This is about events that allow attacking castles, and about castles outside the NAP. In normal times we leave alliance and pact castles alone — there is a separate guide about that.
A castle defends with whatever its owner has been levelling, and people normally level one troop type. So unlike a den, the target is uniform — and a uniform target can be countered.
Archers inside — bring infantry. Infantry inside — bring cavalry. Cavalry inside — bring archers.
If a rally is built with the right troop type and people join with the wrong one, everyone suffers. Losses are shared equally, after all.
The rally leader picked his type to counter the defender. Join with something else and you have brought troops that stand badly against that castle, while the whole group pays for it.
So: look at what type the rally is built with and join with the same. No troops of that type? Join with one unit, or do not join at all.
To be straight with you: this section rests on practice, not on measurements. Our measurements are from dens; we never ran a series like that against castles. But the loss mechanics are shared, and the logic follows directly.
9. Vehicles are a separate matter
Vehicles counter nobody and lose to all three types. We never take them into rallies, neither T1 nor T10. The reasoning is in a separate guide.
10. Formations: who stands in front
The Institute has a Formations branch, unlocked at Institute level 28. Infantry stands in front by default. Research moves cavalry or archers to the front instead.
As long as you march with one troop type this is irrelevant. The moment different types go into battle together, it matters who takes the first hit. This is the one part of the mixed-troops subject that the game acknowledges openly.
11. The short list of rules
- Den — bring any mix. Watch tiers and stats, not troop types.
- Castle — bring the type that counters the defender. Join someone else's rally with the type it was built with.
- Unsure about your stats — join with one unit. That is honest, safe and lets nobody down.
- Level the defence of your own troop type and keep your active heroes above level 131. Below 131 the buffs work at half strength and you become the weak link.
- A low tier is not a death sentence. A full T9 march beats an empty slot. Just do the arithmetic: T6 weighs one third of T10.
- Never bring vehicles to a rally.
12. What we do not know
So that nobody takes this text as gospel, here is an honest list of the holes:
- There is no exact loss formula. It is clear that the weak link drags the group down, but not how — a multiplier, a threshold, steps? Unknown.
- How the enemy distributes its damage between participants — unknown.
- Why the cavalry share moved losses so much — unknown.
- Whether it is one model across all dens or different regimes at different levels — untested.
- There are no measurements against castles at all — only practice.
All the numbers above come from a single account in spring and summer 2026 — and it is not the account this guide is signed with. None of the personal stats carry over; only the formulas and the ratios do. The game changes on top of that. If your results differ, you are more likely right than this text is.
Get in touch
Found a mistake or an inaccuracy? Got a question about a mechanic? Need a guide on a topic that is not here yet?
Write to pnc.wiki.101@gmail.com — I will fix it, answer it, or write it.
One special request: if a mechanic in the game does not look the way it is described here, send a screenshot from the client. A screenshot settles an argument faster than any amount of reasoning, and in these guides it outranks any outside source.