[Bug]: Expected damage overestimates actual; add diagnostic for missing-overlay spells

[Xerrion]

Problem

Two related accuracy / coverage issues observed in real combat on TBC Anniversary as a Warlock:

1. Expected damage is consistently 5-10% above actual sustained damage. Magnitude points at modeling bugs rather than a missing major mechanic.
2. Hellfire (spell ID 1949) does not show an action-bar overlay even though all four ranks are present in Data/SpellData_Warlock.lua and the engine has no skip path for channels, AoE, or self-damage spells. Spell is placed on the bar via direct spellbook drag (no macro).

Root cause analysis

A static audit identified two engine-level bugs that together explain the 5-10% overestimate:

Bug A: spell crit chance double-counted

StateCollector reads GetSpellCritChance(school) (StateCollector.lua:202-207). The WoW API value already includes class talent crit bonuses such as Devastation, Demonic Tactics, etc.

Several Data/TalentMap_*.lua entries declare effect = MOD.CRIT_BONUS for the same talents (e.g. TalentMap_Warlock["3:11"] Devastation, line 271). ModifierCalc then sums those into mods.critBonus, which CritCalc adds on top of the already-talent-adjusted base crit. The bonus is counted twice.

Effect: with 5/5 Devastation and Ruin (crit multiplier 2.0x), spells crit ~5% more often in the model than in reality, inflating expected damage by ~5%.

Bug B: stacking debuffs assumed full-stack

Stacking debuffs declared in AuraMap_*.lua (Shadow Vulnerability / Improved Shadow Bolt 17800/17803, Shadow Weaving, etc.) apply their full multiplier the moment a single stack is detected. The current code path does not read auraData.applications. The Sunder Armor handler in StateCollector.lua:87-92 is the only place stack count is honored today.

Effect: a fresh Shadow Bolt landing into a 1-stack ISB shows *1.20 in the overlay when actual damage is only *1.04. Adds another ~3-15% overestimate during ramp-up.

Hellfire missing overlay

Static analysis cannot identify the runtime cause - data is present, no skip filter exists, BuildSpellIDMap correctly maps all four ranks back to base 1949. Speculative fixes risk shipping a non-fix. A diagnostic command is needed to capture GetActionInfo, ResolveSpellID, the data lookup, and Pipeline.Calculate output for a user-specified action slot.

Proposed fix

Single PR addressing all three:

1. Crit double-count audit and fix. Walk every Data/TalentMap_*.lua and remove CRIT_BONUS effect descriptors that target talents already reflected in GetSpellCritChance(school). Keep entries that grant spell-specific crit bonuses (e.g. Improved Searing Pain on a single spell). Add busted test that asserts a Destro warlock with 0% base API crit and a separate API-already-included +5% crit produces the same expected damage as the version where talent map adds the +5% (regression guard).

2. Stack-aware aura modifiers. When applying an aura entry whose flagged effect scales linearly with stacks, multiply by (auraData.applications or 1) / aura.maxStacks of the declared bonus. Update Shadow Vulnerability (assumed-5-stack hard-coded behavior) and Shadow Weaving as the seed cases. Add busted tests at 1, 3, 5 stacks.

3. /phd debug diagnostic command. New slash subcommand that takes an optional action slot (default: currently hovered or action slot 1) and prints:

   • GetActionInfo(slot) raw return values
   • ResolveSpellID outcome
   • Spell data table lookup
   • Pipeline.Calculate result or skip reason
     Output is plain print() to chat - no UI. Used to debug Hellfire and any future missing-overlay reports.

Out of scope (deliberately deferred)

• Average partial-resist modeling (would risk swinging into underestimate once Bugs A+B are fixed)
• Channel coefficient sourcing for Hellfire / Rain of Fire (open question whether 0.4286 is correct)
• Sustained-DPS execution-efficiency multiplier (movement, clipping, latency)

These will be filed as follow-up issues after the diagnostic command surfaces real data.

Verification

• luacheck . passes 0/0
• busted --verbose all tests green including new regression tests
• In-game smoke test on Warlock: /phd debug works on a Hellfire button; expected damage on Shadow Bolt with full Devastation/Ruin no longer inflates by ~5% relative to actual; Shadow Bolt expected scales correctly with ISB stack count