Threat, Loot Rolls & Item Generation
What determines threat, how item power is rolled, how source rarity weights work and what loot bonuses actually change.
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.
Threat follows level, area and an offset
Recommended threat = ceil(hero level / 5) in the shipped baseline. Levels 1–5 map to threat 1 and levels 96–100 to threat 20. The selected routine does not read gear power.
Current threat = max(area threat, recommended threat) + threat offset. This calculation has no clamp of its own. The universal mapping from every game-facing difficulty label or encounter rule to the offset remains unresolved. Party world threat uses the ceiling of the mean current threat of the player actors counted by that helper.
Evidence for this section
- Recommended threat follows character level
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/0.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Current threat takes the higher of area and recommendation, then the offset
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/1.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Weapon multipliers also include global and selected difficulty curves
native call chain plus authored curves. Record:
reference/local-weapon-power-evidence-1.1.1.0.json/findings/2.Reviewed record SHA-256:
e6dc83dce5131896015667bd9bfdc557abf5323849f6f22af54f570a6caa56b8 - Shipped Active balance inherits SecondPass by copying the parent struct
Native registration/import/merge/copy call chain plus shipped table; 32 controlled CPU cases. Record:
reference/local-balance-resolution-evidence-1.1.1.0.json/findings/0.Reviewed record SHA-256:
ca95aa704324ad911f7ee380e0aa3663773fc003f3b32fde4e9e0a81e7086cae
Dropped Item Power uses a centered random roll
For the selected generator, C = trunc(float32(hero level × 1.5)). low = max(1, C + rarityOffset − 4); high = max(1, C + rarityOffset + 4). The generator sums two half-range random draws, creating a center-weighted distribution rather than a uniform one. It rounds the float32 result to nearest even and caps it at the threat benchmark plus 15.
That benchmark uses max(recommended threat, area threat). The item's rarity offset and resolved source still have to be supplied. This establishes one item-generation path; it is not a universal +9 rule, a Blacksmith upgrade formula, or a promise about all special rewards.
Evidence for this section
- Item generation is centered on character level, with a threat cap
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/2.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Shipped Active balance inherits SecondPass by copying the parent struct
Native registration/import/merge/copy call chain plus shipped table; 32 controlled CPU cases. Record:
reference/local-balance-resolution-evidence-1.1.1.0.json/findings/0.Reviewed record SHA-256:
ca95aa704324ad911f7ee380e0aa3663773fc003f3b32fde4e9e0a81e7086cae
Rarity curves are multipliers on source weights
The generator merges source weight operations, restricts the selected item to its possible rarities, multiplies retained rarity weights by their threat curves, then performs weighted selection. A single positive candidate or a source that skips threat skew bypasses that adjustment.
The generic rarity table alone cannot give the chance of finding a specific Unique. A zero ordinary threat multiplier does not erase a guaranteed reward or a source that bypasses that branch. We have preserved 163 authored loot-source layers and 191 slot plans; unresolved parent inheritance and selection rules prevent treating them as final item odds.
The rarity-bonus branch can promote Common to Rare or Rare to Special when the requested next rarity is retained in the source list. It does not promote Special to Unique. A displayed bonus percentage is therefore not an unconditional Unique-drop probability.
Evidence for this section
- Threat values multiply source rarity weights
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/4.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Opulence has a one-step Common/Rare upgrade branch
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/5.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Exact named-item odds depend on history and eligible source pools
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/8.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7
Extra drops, duplication and multiplayer
A resolved slot has guaranteed drops and separate random attempts. When its modifiers are allowed, a random attempt compares its draw with baseChance × multiplayerScale + DropChanceIncrease. Successful creation and eligible pools are additional conditions.
LootingMultiplier supplies a chance for each configured extra payout attempt per successful random base roll. Guaranteed drops do not increment that counter. DropDuplicationChance is a separate event roll: it copies one highest-rarity generation plan, choosing randomly between ties. It is not the same as rerolling every item independently.
For configured multiplayer-scaled reward slots, the selected helper uses spawnMultiplier[1] / spawnMultiplier[playerCount], followed by an optional slot override. The shipped spawn multipliers for one to four players are 0.8, 1.2, 1.6 and 1.9. Missing context/configuration returns 1. These factors apply to the selected random-drop chance, not every guaranteed reward.
Gear slots track three recent drops for repeat prevention; book slots track four and also request one-of-a-kind prevention. Talismans request one-of-a-kind prevention. Ownership, history, pool modifications and encounter inheritance must be included before calculating named-item farming odds.
Evidence for this section
- Guaranteed drops and random rolls are separate
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/3.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Looter rewards successful random rolls; Lucky selects one best-rarity plan
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/6.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Co-op scales random gear chance through spawn multipliers
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/7.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Exact named-item odds depend on history and eligible source pools
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/8.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7 - Shipped Active balance inherits SecondPass by copying the parent struct
Native registration/import/merge/copy call chain plus shipped table; 32 controlled CPU cases. Record:
reference/local-balance-resolution-evidence-1.1.1.0.json/findings/0.Reviewed record SHA-256:
ca95aa704324ad911f7ee380e0aa3663773fc003f3b32fde4e9e0a81e7086cae
Rarity and Storm Called effect capacity
The shipped ordinary rerollable counts are Common 0, Rare 1, Special 2 and Unique 1. A Unique's fixed trait is separate. Rare, Special and Unique carry a guaranteed extra-effect entry for the StorminatorReward tag; Common does not.
Fixed effects are stored before random effects. The selected generator filters candidates by base effect family, so a fixed Unique template can exclude ordinary tiers of that same family on the item. Complete roll pools and reroll-position legality remain under review.
Build Lab now uses these capacities. Older builds with excess rolls keep every selection and show a review warning; they are not silently converted to legal builds or trimmed.
| Rarity | Ordinary random slots | Storm Called random slots | Fixed traits |
|---|---|---|---|
| Common | 0 | 0 | If authored separately |
| Rare | 1 | 2 | If authored separately |
| Special | 2 | 3 | If authored separately |
| Unique | 1 | 2 | Separate from random slots |
Evidence for this section
- Unique weapon fixed traits are separate from one baseline rerollable effect
Shipped rarity row, native inheritance and effect-count routine. Record:
reference/local-item-effect-generation-evidence-1.1.1.0.json/findings/1.Reviewed record SHA-256:
ca4bb9a8be2c3f7b513351cb542f67256dcf568ebd8e37f4ef0b953c0634f406 - Fixed and rolled copies of the same effect family are filtered on one item
Native generation call order and 12 family-filter cases. Record:
reference/local-item-effect-generation-evidence-1.1.1.0.json/findings/2.Reviewed record SHA-256:
ca4bb9a8be2c3f7b513351cb542f67256dcf568ebd8e37f4ef0b953c0634f406 - Storm reward tags add a rerollable effect for supported rarities
recovered native path with stated inputs. Record:
reference/local-threat-loot-evidence-1.1.1.0.json/findings/9.Reviewed record SHA-256:
49fb28c82e4ea0003497324dab152a5ee44ba743923199fa8bd0d150f966f5b7
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.