Weapon Power, Gear Score & Base Damage

Which balance settings apply?

The shipped Active ruleset inherits SecondPass through a native parent-copy operation. The values below are therefore the resolved local baseline for build 1.1.1.0. They are stronger evidence than the executable's startup defaults or an isolated table row. They are not a capture of an online server's current settings.

A weapon uses its own Item Power

Melee scaling reads the equipped melee weapon's power; ranged scaling reads the ranged weapon's power. The three artifact slots have separate power attributes. Replacing an unrelated armor piece does not change the direct weapon-slot power input in these paths.

Let P be the weapon's Item Power and T the captured current threat. trunc means discard the fractional part. First compute Q = trunc(1 + max(trunc(T) − 1, 0) × 5 × 1.5). Then F = max(0.1, 1.5 / (1 + 0.5 × exp(0.15 × (Q − P)))). At P = Q, F is 1. The lower floor is 0.1; 1.5 is an asymptote approached from below, not a reachable hard cutoff.

The weapon multiplier is F × globalOutgoing(worldThreat) × selectedDifficultyOutgoing(worldThreat). The world-threat helper takes the ceiling of the mean current threat of the player actors it counts. The shipped Low outgoing curve is 1.5 and High is approximately 0.6; the neutral factor is 1. This describes the selected outgoing-damage stage, not all difficulty effects.

Armor power and overall equipment average are different

EffectiveArmorPower = (helmet + chest + leggings + boots + mean(artifact 1, artifact 2, artifact 3)) / 5. Each armor piece contributes one fifth; all three artifacts together contribute the remaining fifth. RawArmorPower is the mean of the four armor pieces.

The separate equipment-average routine excludes talismans, cosmetics and private or unequipped entries. It rounds each eligible power, skips values below 1, truncates the resulting mean and returns 1 for an empty set. This is not a reason to average all twelve Build Lab slots. The arithmetic is recovered; every UI display binding and the complete armor-mitigation path are still being traced.

How to read the weapon records

All 58 melee and 22 ranged weapons now have local base-data supplements on their individual pages. Normal melee primary damage is Damage × DamageRatio. Each additional damage type receives Damage × (1 − DamageRatio). Splash begins at Damage × clamp(splashAttribute × splashMultiplier, 0, 1), then follows the same type split.

The selected ranged launch path starts with projectile base damage, applies the charge multiplier when charged, and applies a qualifying jump-shot multiplier. Impact prefers a positive supplied damage override over the projectile's original base. The separately stored ProjectileDamageFactor remains unresolved across other firing paths and is displayed as such.

Attack variants, stored loop counts, cooldown fields and attack-rate fields are authored data. They do not establish an animation's hit count, total firing cycle or DPS. Jump attacks, procs, splash and fixed traits must be accounted for separately. Fixed traits have been resolved across all 80 weapon records; a restricted direct-hit prototype supports 46 of the 58 melee weapons and rejects the remaining 12 rather than inventing their contributions.

Try the conditional melee calculator

Open Research Lab → Melee damage to compare one normal attack variant. The worked Axe example uses power 143, captured/world threat 20, Normal difficulty, Sharpness III and an already-critical hit against a hypothetical neutral target. Its internal packet total is 9.828 and its display-scale sum is 221.13 before popup rounding. Changing the defenses or qualifying conditions changes the result.

The display stage divides out the combined weapon multiplier and applies the Item Power display factor in this path. Therefore threat or outgoing difficulty can change the internal result while leaving the display-scale sum unchanged for otherwise identical inputs. This does not mean difficulty has no effect on combat.

Exporting keeps the scenario inputs, result and exclusions in a JSON file. The calculator does not read a hero save or connect to the game.

Sources & data coverage

Reviewed local game tables and selected native instructions from Windows build 1.1.1.0. Original-instruction tests use synthetic actors and engine services. They are not gameplay measurements; online overrides have not been observed.

Reviewed evidence manifest and scalar data ↗