← Back to How it works

Comparing Weapons by Power Level

"Are these weapons about the same power level?" is a question the calculator can answer properly, because both engines already do the hard part: resolve a real profile against a real target and report what happens. This page shows the method, the measured numbers for the KT24 heavy weapon lists, and the sets that came out balanced — including what the model does not count, which matters as much as what it does.

The numbers here are produced by a harness in the test suite, not by hand. Anyone can re-run it and get the same values — see Reproducing this at the bottom.


The method

One weapon, one slate of targets. Comparing two weapons against a single target says very little: the answer flips depending on the target's save and wound count. Every weapon here is scored against the same five defenders, chosen to span the profiles a kill team actually meets:

TargetMelee profileShooting profile
Horde3 dice, 5+, 2/3 damage, 7 woundsSave 5+, 7 wounds, no cover
Chaff3 dice, 4+, 3/4 damage, 8 woundsSave 4+, 8 wounds, no cover
Trooper3 dice, 4+, 4/5 damage, 10 woundsSave 4+, 10 wounds, in cover
Marine4 dice, 3+, 4/5 damage, 13 woundsSave 3+, 13 wounds, no cover
Elite5 dice, 3+, 5/6 damage, 15 woundsSave 3+, 15 wounds, in cover

Melee defenders need a full operative profile because they fight back; shooting defenders need a save, a wound count, and whether they hold cover.

Melee is scored as a fight, not as output. A melee weapon that deals big damage while eating a parry-heavy counterattack is not the same as one that wins clean, so melee runs through the Fight engine as a real one-round combat: both fighters set to Max Damage, alternating strike and parry (see Fight rules). The carrier is fixed at 13 wounds for every weapon, so the comparison is between weapons rather than between operatives. Three things get recorded: expected damage dealt, expected damage taken, and the chance the target is reduced to 0 wounds. The Fight engine is Monte Carlo, so the harness pins the simulation count (40,000) and the seed (24601) — re-running reproduces the published numbers exactly instead of landing near them.

Shooting is exact. The Shoot engine enumerates every attack-and-save outcome, so no sampling is involved: expected damage and kill chance are computed, not estimated. Cover, Piercing, Devastating, and Saturate are modeled as the engine handles them (Combat rules).

Kill chance is the headline number for melee, damage for shooting. In a fight, killing the enemy is what stops the counterattack, so kill chance is the outcome that decides games. A shooting attack is one-directional and usually part of a turn's worth of fire, so expected damage is the fairer summary; kill chance is reported alongside it.

The buff is scored too. A set that is balanced plain can come apart the moment a team upgrade lands on it, so every weapon is scored twice: as printed, and with the buff that upgrades it. For melee that is Paired Weapon (+1 Atk); for shooting, Twinned (Ceaseless, or +1 Atk on a Burst Cannon). Three spreads then matter, not one:

  • plain spread — how even the weapons are as printed;
  • buffed spread — how even they are once everyone is upgraded;
  • gain spread — how differently the buff treats them.

A set can look even and still be broken, if the buff is worth twice as much on one member as on another. That is the failure the third number catches.


Melee results

Ten profiles, one round of combat, averaged over the five defenders. Kill chance is the chance the target is reduced to 0 wounds.

WeaponKill %With +1 AtkGainDamage dealtDamage taken
Scything Talons85%92%+6.810.25.0
Flesh Mower81%89%+7.710.15.4
Power Fist77%88%+10.79.74.5
Power Scourge77%85%+7.79.75.2
Thunder Hammer76%87%+10.99.64.6
Power Weapon72%84%+11.89.45.1
Chain weapon72%85%+12.89.35.6
Power Talon71%84%+12.99.35.5
Dread Saw71%80%+9.49.55.6
Close Combat Weapon24%38%+14.35.77.0

Three things fall out of this table:

  • Scything Talons and Flesh Mower have no bad matchup. Ceaseless on 5 dice, and hitting on 2+, put them above everything else against hordes and against Marines. Nothing in the list can be balanced against them.
  • Close Combat Weapon is the baseline, not a peer. At 24% it is less than half of the next weapon up. It is what a weapon looks like with no rules attached.
  • +1 Atk is a flat buff. Every real weapon gains 7 to 13 points. That is what makes a buff-resistant set possible here; a percentage-style buff would not behave so evenly.

Recommended melee set

Power Scourge, Thunder Hammer, Chain weapon, Power Weapon, Power Talon. Kill chance plain → with +1 Atk:

WeaponHordeChaffTrooperMarineEliteAverage
Power Scourge96% → 98%96% → 98%88% → 94%69% → 81%37% → 53%77% → 85%
Thunder Hammer91% → 96%89% → 95%87% → 94%67% → 82%47% → 68%76% → 87%
Chain weapon88% → 95%86% → 94%86% → 94%54% → 74%45% → 66%72% → 85%
Power Weapon94% → 97%86% → 94%86% → 94%54% → 72%42% → 63%72% → 84%
Power Talon88% → 95%86% → 94%86% → 94%53% → 74%44% → 65%71% → 84%

Spreads: 5.8 points plain, 3.1 buffed, 5.1 on the gain. The set gets tighter when Paired Weapon is in play, because the weapon slightly ahead of the pack (Power Scourge, already at 6 dice) is the one with the least room to gain.

The set also keeps distinct roles, which is the point of a balanced set rather than five copies of one weapon: Power Scourge is the horde eater (96% into 7-wound models, 37% into Elites), Thunder Hammer is the elite killer (47%, the best in the set), and the other three are generalists within a point of each other.


Shooting results

Twenty-five profiles, one shooting attack, averaged over the five defenders. Supercharge profiles are used where a weapon has one.

WeaponDamageWith buffGainKill %
Thermal Spear14.520.2+40%65%
Heavy Rail Rifle14.420.2+40%64%
Macro Plasma Incinerator (supercharge)12.517.1+37%60%
Splinter Cannon10.514.8+40%48%
Meltagun10.314.8+43%48%
Multi-melta10.314.8+43%48%
Lascannon10.314.7+43%47%
Heavy Onslaught Gatling Cannon10.014.1+41%49%
Flamestorm Cannon9.812.2+25%44%
Heavy Bolter9.312.9+38%45%
Autocannon9.213.3+45%44%
Plasma Cannon (supercharge)9.112.7+39%41%
Plasma Gun (supercharge)9.112.7+39%41%
Starcannon9.112.7+39%41%
Cyclic Ion Raker8.612.2+42%42%
Shuriken Cannon8.511.8+39%42%
Heavy Phosphor Blaster8.311.5+40%39%
Burst Cannon8.210.8+31%37%
Missile Launcher (krak)7.911.4+46%36%
Heavy Flamer7.89.8+26%32%
Heavy Stubber7.410.7+45%35%
Rocket Launcher6.710.1+51%31%
Stranglethorn Cannon5.68.0+44%24%
Havoc Launcher5.48.0+49%25%
Missile Launcher (frag)4.26.3+49%15%

Worth noticing:

  • Plasma Cannon, Plasma Gun and Starcannon are the same weapon on paper — 4 dice, 3+, 5/6, Lethal 5+, Piercing 1 — and produce identical single-target numbers. Their differences are Blast, Heavy, and Hot, none of which show up in a single-target comparison.
  • Thermal Spear and Heavy Rail Rifle sit 40% above the pack. Like Scything Talons in melee, they cannot be balanced against the rest of the list.
  • Ceaseless is worth roughly +40% damage across the board. The exceptions are the weapons that already hit on 2+ or already reroll: Flamestorm Cannon (+25%), Heavy Flamer (+26%), and Burst Cannon (+31%, since Twinned gives it +1 Atk instead).

Recommended shooting set

Heavy Bolter, Autocannon, Starcannon, Shuriken Cannon, Heavy Phosphor Blaster, Missile Launcher (krak). Expected damage, plain → with Ceaseless:

WeaponHordeChaffTrooperMarineEliteAverage
Heavy Bolter11.4 → 15.010.1 → 13.78.4 → 11.98.9 → 12.47.8 → 11.29.3 → 12.9
Autocannon12.4 → 16.910.2 → 14.48.2 → 12.28.1 → 12.16.9 → 10.79.2 → 13.3
Starcannon11.2 → 15.19.9 → 13.68.3 → 11.88.6 → 12.17.7 → 11.19.1 → 12.7
Shuriken Cannon10.8 → 14.49.2 → 12.67.8 → 11.07.7 → 10.96.9 → 10.08.5 → 11.8
Heavy Phosphor Blaster10.3 → 13.88.6 → 11.98.6 → 11.96.9 → 10.06.9 → 10.08.3 → 11.5
Missile Launcher (krak)10.5 → 14.48.8 → 12.56.9 → 10.37.1 → 10.66.0 → 9.37.9 → 11.4

Kill chance for the same six:

WeaponHordeChaffTrooperMarineEliteAverage
Heavy Bolter83% → 96%76% → 94%36% → 67%26% → 51%6% → 18%45% → 65%
Autocannon75% → 92%62% → 86%49% → 78%20% → 42%12% → 30%44% → 66%
Starcannon69% → 89%60% → 83%49% → 76%18% → 38%11% → 27%41% → 63%
Shuriken Cannon78% → 94%68% → 89%40% → 63%19% → 39%6% → 15%42% → 60%
Heavy Phosphor Blaster75% → 92%62% → 86%36% → 63%17% → 35%6% → 15%39% → 58%
Missile Launcher (krak)69% → 87%51% → 76%37% → 65%17% → 34%6% → 16%36% → 56%

Spreads: 1.5 damage plain, 1.8 buffed, 0.8 on the gain — the best gain spread available from the pool once weapons carrying Selection are excluded, since taking an extra weapon slot is a real cost the dice never show.

Roles inside the set: Heavy Bolter and Autocannon are the horde clearers, Starcannon and Missile Launcher (krak) are the anti-armor picks, and Heavy Phosphor Blaster is the cover-breaker — Saturate means its Trooper number is the only one in the set that does not drop when the target takes cover.

Adding a seventh weapon widens at least one spread no matter which is chosen, so six is the natural stopping point. If a seventh is needed anyway, Heavy Stubber keeps the gain spread at 0.8 while costing plain spread (1.5 → 1.9); Burst Cannon does the reverse, holding plain spread at 1.5 but falling behind once buffed, because Twinned treats it worse than everything else in the list.


What this does not measure

The numbers above are a single attack against a single target. Several real sources of value sit outside that frame, and every one of them favors particular weapons:

  • Blast and Torrent hit extra operatives. Every number here is single-target, so a Torrent 2" weapon is worth more at the table than its row suggests. This is the largest omission on the shooting side, and the reason the recommended set was checked for how many of its members carry one.
  • Hot can damage the shooter. Supercharge profiles are scored for what they do to the target only.
  • Heavy, Selection, and Range are restrictions with no dice effect. Heavy limits movement, Selection costs an extra weapon slot, Range caps the engagement. A weapon that looks even on damage can still be worse to take.
  • Stun, and other rules that hit the next activation. Thunder Hammer's Stun does nothing to the dice in the fight it happens in, so its melee row understates it.
  • One round, one attack, no ploys. No APL costs, equipment, ploys, or multi-round attrition. A weapon that wins the second round is scored the same as one that wins the first.

None of these break the comparison — they are the reason the recommendation names them explicitly rather than pretending the ranking is the whole story.


Reproducing this

The harness lives in the test suite and is skipped unless BALANCE_OUT is set, so it costs nothing in a normal run:

CI=true BALANCE_OUT=/tmp/balance.json npx react-scripts test --watchAll=false \
  --testPathPattern WeaponBalance

It writes one JSON file holding every weapon, plain and buffed, with per-target damage, damage taken (melee), and kill chance, plus the defender slate and the Fight engine's seed and simulation count. Editing the profile lists at the top of src/WeaponBalance.test.ts re-scores any other set of weapons the same way.

The same comparison can be done by hand in the calculator for a single matchup: set the attacker, note the expected damage, change the one input, and compare. That is the approach behind When not to take cover saves, and the same discipline applies here — an unexplained difference between two profiles is a question to answer, not a result to publish.

← Back to How it works

Open source on GitHub — Pull requests welcome!
Forked from https://jmegner.github.io/KT21Calculator/
⚔️ Alternative Tools
Ballistica Imperialis →
Alternative KT math tool with preloaded weapon & defence profiles and a cool design.
NemesisForge →
Build your own nemesis operatives.
How KT Calc works → (opens in a new tab)
Deep-dive notes and the Kill Team rules the calculator is built on.