What this site answers

Four questions

What this site refuses to do

• No single-card composite score. A single ranking systematically pushes "scales over time" and "scales with the enemy" cards to the bottom — the method below explains why.
• No invented numbers. Where the game still stores a variable, the site writes "value not measured yet" instead of a plausible guess.
• No single authoritative tier list. When community rankings conflict, they are shown side by side with dates and sources.

Method: what any single-card ranking leaves out

The card-output reference computes √(damage × HP), divided by cost or not. It only answers "paper output per gold or per plot". The nine items below are what it structurally cannot measure — which is why every ranking page states its premise up front.

1. ① Crit multiplier is not uniformCriticalMultiplier sets the ceiling for crit builds: cards at 1 can reach ×8, the rest ×16 — a one-step difference in ceiling.
2. ② Crit growth is rareOnly a handful of cards have MCritChanceBoost > 1; every other card gains no crit chance from levels, so crit builds have far fewer carriers than they look.
3. ③ Range decides how much paper DPS landsThe formula assumes 100% uptime. A unit with range 1.1 has to walk to the front before firing; paper HP is not effective HP.
4. ④ AOE and pierce multiply by targetsHit N targets and effective DPS is roughly paper × N. Single-target math systematically undervalues both.
5. ⑤ Adjacent support buildings multiplyBeacon, blacksmith, orchard and camp boost neighbours by a percentage every year. The same unit between four boosters and in a corner diverges by several times over a long run — that is placement, not card choice.
6. ⑥ Summons, splits and revives raise effective HPDuplicators refill units, slimes split, pagodas summon, Embalming revives: the health bar is a pipeline and MaxHP is one section of it.
7. ⑦ Thorns and damage reduction never enter a DPS formulaThe defender attacks for almost nothing and ranks last on both tables, yet its job is soaking and reflecting. Steel coat stacks are mitigation.
8. ⑧ Not every card caps at level 3Cards with LimitlessLevels do their real work well past level 3; any table that stops at MaxLevel writes them off.
9. ⑨ Equal DPS/HP weighting is a choiceThe geometric mean √(DPS×HP) assumes both matter equally. Losing HP ends a run and dead units deal no damage, so pure damage units are worth less than the table says.

The per-item data lives on thecard output reference; use it as "compute first, adjust second", never the other way round.

Evidence levels

Coverage

680 visible entries · baseline 2026-11-02 · Unity 6000.3.8f1 (IL2CPP) · 9 Kings (AppID 2784470)

Data baseline

Every number on the site comes from the same extraction. If a value disagrees with the game, check the date here first — the data layer has to be re-extracted after v1.0 ships.

Version baseline

Data was extracted from the local client on 2026-11-02 (late Early Access). v1.0 is planned for 2026-11-12; card, policy and perk sets will change and the data layer will be re-extracted. Start, systems and mechanics content survives the patch, but its numbers are version-tagged.

This is a personal, unofficial reference. Data comes from local game-file extraction; some descriptions reference 9 Kings Wiki (wiki.gg, CC BY-SA 4.0). Editorial notes are original.