How we build, test and tune instances and bosses

Dungeons, expeditions, raids and world bosses are the part of WildStar the client files tell us least about. The client has the creatures, the spells and the telegraphs; it does not have the spawn lists, the order bosses use their abilities in, how long a fight lasted, or what anything dropped. This page says how we fill that gap, how we test the result without a person at the keyboard, and, boss by boss, how far each fight has got and how sure we are of it. It is written for players who want to know what to expect, for contributors who want to improve a fight, and for anyone curious about how an emulator gets tested.

It builds on the work of the NexusForever project, whose message models and captured world data are the foundation under all of it. For where our data differs from retail and how sure we are, see Bezgelor and WildStar: where we differ.

What we build, and how sure we are

Every instance and world boss is built in three levels (master plan, section 11). They are the instance tests of the one scale on the fidelity map (0 Unbuilt, 1 Working, 2 Complete, 3 Checked; a + is partway to the next level):

1. Level 1, Working. Spawns and trash, bosses using their real abilities, phases, an enrage where there is one, loot, entry and exit, lockouts.
2. Level 2, Complete. Every documented mechanic: telegraphs, interrupt armor, add waves, medals, adventure branches, with timers taken from guides.
3. Level 3, Checked (tuned). Timings checked against video, positions refined, health and damage balanced by simulating a group, and headless-group tests end to end.

Every instance and world boss is aimed at Level 2. Level 3 is only for the flagships (Stormtalon's Lair, Genetic Archives and one world boss), because they build the tooling the rest reuse. The stopping rule keeps this honest: we stop on a detail when the only evidence left is guesswork, or when it needs a check in the real client. That detail stays labelled guess. There is also a time box of about two hours per instance beyond Level 1, so Level 2 means "everything we could document", not "everything".

Where the data comes from

• The client's own tables. Creature, spell, telegraph, public event and objective rows. A boss's ability list is read from the spell rows whose developer text names the encounter (for example e444 - Grond - ...). Creatures link to no spells in the client, so which spells a boss uses is itself something we derive.
• NexusForever captures. Their world database was built from retail captures. It has spawns for some open-world zones and for Evil from the Ether, and nothing for dungeons or raids.
• Guides and wikis. Archived wiki pages, guild guides (Dulfy's veteran guides are the best), patch notes, boss-mod addon source, Jabbithole's archived NPC pages. They give phases, thresholds, interrupt armor and some health values.
• Video. Normal and veteran runs on video show what a fight looks like. See "How we compare with retail" for exactly how little we use them for timing.

The labels

Every generated record carries one of three labels and a note on its basis (provenance convention):

• reference: taken straight from a retail source (a client table row used as it is, the NexusForever world database, a cited guide or patch note).
• derived: computed from client data by a rule we wrote. The inputs are retail; the combination is ours.
• guess: extrapolated or designed, with the reasoning in the basis. A boss cadence nobody published is a guess.

Values we chose in code rather than in data are listed in designed-values.md. The labelled guesses of every instance are collected in wildstar-delta.md.

How an encounter is tested

Four layers, from cheap to slow. None of them needs a person.

1. Zone tests with the client's own parser

Each instance has tests that start a real zone process: the instance's spawns, its map script and its encounter scripts, then drive the boss through its phases and check what happens (a barrier going down, an interrupt breaking a cast, a door opening, a medal being awarded). The packets such a test makes the server send are checked by the client oracle: the WildStar client's own message parser, run on our bytes. A packet the real client would reject fails the test. See the oracle.

2. Headless end-to-end runs

A headless client logs in through the real protocol (the same login, realm and world servers a player uses), forms a group, enters the instance and fights, with the oracle checking what it receives. Four dungeons (Stormtalon's Lair, Ruins of Kel Voreth, Skullcano, Sanctuary of the Swordmaiden), the expeditions, the battlegrounds and Slaughterdome, and Kraggar have one (apps/bezgelor_world/test/e2e/).

What these runs do and do not prove: they prove a group can get in, see the right things, trigger the mechanics and finish the content over the wire. In the Stormtalon's Lair run, for example, the bosses are brought down with a GM damage command and the mechanics that need a strike or an interrupt are triggered on the zone's scripts. The real fights, with every ability on both sides, are the simulator's.

3. The encounter simulator

The simulator (BezgelorWorld.EncounterSim, apps/bezgelor_world/lib/bezgelor_world/encounter_sim*) fights a boss for real, with scripted bots, and measures what happens.

What it does.

• Starts a real zone instance: the instance's own spawns, map script and encounter scripts, the real spell pipeline for every ability on both sides, the real AI, threat, interrupt armor and Moments of Opportunity. Nothing is faked on the zone's side; the bots use the same public calls a player's connection does (cast_spell, player_moved).
• Fields a bot group with roles: a tank, a healer and damage dealers, from a Warrior, Medic, Stalker, Spellslinger and Engineer for a dungeon group, and mixed classes for raids.
• Gives the bots level-appropriate gear from the item tables (green and blue quest gear at level 20; purple dungeon gear at level 50, where a Warrior has about 46,000 health and 10,000 armor).
• Has them dodge telegraphs and ground zones: each bot steps out of a telegraph whose cast gives it half a second or more with a chance of 0.7, and keeps out of a boss's persistent ground zones. It has them interrupt a cast when the boss's interrupt armor is finite, and heal below 95%.
• Handles raid sizes (10, 20, 30 players or any other number; one or two tanks, one healer per eight players, the rest damage) and mid-fight joiners (players who enter the zone partway through, as a world boss's crowd does).
• Runs in real time: a pull lasts as long as it would in the game.

What it measures. Time to kill, deaths, wipes (and the reset), the phases reached, which mechanics hit whom and how often, damage dealt and taken per bot, and the Moments of Opportunity created. Each run is written to a dated JSON file (schema bezgelor.encounter_sim.v1) and a Markdown report with the command that reproduced it, in test-reports.

Known limits (tracked in #171):

• The bots have unlimited resources. Their damage is paced only by global cooldown, ability cooldowns and cast times, so it is higher than a real player's. Fights end faster than they should.
• The healer logic is simple. The Medic bot spends most of its global cooldowns attacking, because its heals are telegraphs around itself and the bot does not yet choose where to stand to cover the group.
• Bots do not use tier-upgraded abilities, AMPs, defensive cooldowns, call-outs or positional play. Players do.
• Damage per mechanic is not measured yet: the report counts hits per spell, not damage per spell.
• Runs are slow (a full Stormtalon's Lair report of five encounters, twice each, takes about 25 minutes of real time).
• The bots are not real characters, so a few database rows (for example public event history) are refused. That is logged and harmless.

Because of those limits, a simulated time is a relative measure that tells us whether a fight is wildly off, not a prediction of what a group of players will do.

4. People in the real client

The last layer is the only one that can say whether a fight feels right: whether a telegraph looks like the footage, whether a mechanic is readable, whether a boss is fun. Each instance has a client-test issue with the exact steps. See "How to help".

How tuning is done

Each encounter in an instance's data file has a tuning block, per difficulty:

"tuning": {"normal": {"health": 117600, "damage_scale": 0.08}, "veteran": {"health": 1000000}}

• Health. Bezgelor targets the final live game. The server's creature formula (BaseHealth' + 14 x Stamina' from the creature's level row, multipliers and Creature2OverrideProperties; semantic registry) reproduces the final-era retail captures and most archived pages, so a boss without a tuned health uses it. Older guides and archived pages that differ (Dulfy 2014, Jabbithole Drop 5) are an earlier era: the final build's rows are higher, and we keep the final build's value. Where we set a number ourselves (a simulator tuning) it is labelled designed. Stormtalon's Lair normal is tuned above the formula, which gives the archived figures exactly (117,590 for Blade-Wind, 137,250 for Aethros, 225,000 for Stormtalon), because the bots have unlimited resources (#171).
• Damage scale. A multiplier on everything the boss deals. Normal-difficulty dungeon bosses use 0.08.

Why 0.08. Found by the simulator (#169): five level 20 bots in level 18 to 20 green gear (about 11,500 health each) wiped against every Stormtalon's Lair boss within 20 to 90 seconds at the damage the client's creature formula gives, whatever they did. A scale of 0.08 is the value at which the simulated group clears the bosses with few deaths. It is a designed value, labelled as such, not a retail one.

Was the formula wrong? We checked, and found no formula bug. With the damage calculator on the real spell rows, the inputs match the client's shape and NexusForever's; one of Aethros's veteran Torrent rows comes out at 20,500 to 28,200 after a level 50 player's armor, against the guide's "about 20k"; and the ratio of a hit to a player's health is the same at level 20 and level 50. What the bots lack that players have is what probably caused the wipes: shields, upgraded abilities and AMPs, defensive cooldowns, call-outs for interrupts and positions, and a healer that heals. So the damage scale is a tuning knob for what the bots cannot do, not a fix for a formula, and the per-encounter values stay in tuning. We expect them to move toward 1.0 as the simulator learns to play better; #169 stays open for that. One real bug turned up on the way and is fixed: scripted spawns without a captured level took a random level from 1 to 50, so Kraggar spawned at level 11.

Target kill times. The Stormtalon's Lair test asserts a kill within 5 minutes for Blade-Wind, 6 for Aethros and Stormtalon, and 4 for the minibosses; these are designed upper bounds, because retail times are only roughly known (next section). For Kraggar the target is a kill by 20 bots within 10 minutes with most of the group alive, inside a 5 to 15 minute range that is a guess. Within those bounds, health stays at the reference value until the simulator models resources; then we retune from a new run.

How we compare with retail

This is the part to read carefully, because it is easy to overstate.

Nobody on the project has timed footage. The "retail" kill times below are estimates made from video lengths, descriptions and guide text, not measured from footage. So:

• The "retail" kill times in our reports are estimates, made from the lengths and descriptions of normal-mode videos and from guide text. They are not measured from footage and not timed against any particular run.
• Every one is a guess in our labelling. The only exception would be a source that states a kill time outright, and for these encounters we have found none.
• The comparison is a coarse sanity check: is the simulated kill within a factor of two or three of "a few minutes", or is it ten seconds or an hour? It is not a timing match.

Exactly what was used:

• Stormtalon's Lair. The dungeon doc says normal-mode videos show kills of about 1.5 to 4 minutes per boss and a 25 to 35 minute run, and that no guide gives ability cooldowns. The report's per-boss figures (Blade-Wind 2 to 3 minutes, Aethros 2.5 to 3.5, Stormtalon 3 to 4, minibosses about 1.5) sit inside that span; they are not individually sourced. Only Blade-Wind's carries the words "guess from their length" in the report.
• Kraggar. No retail kill time is published at all. The only retail fact is an official 2015 post saying he was balanced for 20 players. The 5 to 15 minute range is a guess for a boss of that size, and the simulated 6:32 is, in the research doc's words, "a target, not a measurement".
• Level 2 dungeons and expeditions. No retail time is recorded, and none is simulated.

The Stormtalon's Lair report also says that Blade-Wind and Aethros "fall in or near the retail range" and that Stormtalon and the minibosses die "2 to 4 times faster than in the normal-mode videos". Read those as "against our estimate": Aethros at about 2:00 is below his 2.5 to 3.5 minute estimate, not in it, and "the videos" were not timed.

A timed comparison needs a person (or footage with a timestamp). If you have one, see the next section.

Numbers are different from timings. Boss health, level, interrupt armor and phase thresholds do not need footage: archived Jabbithole pages and Dulfy's guides state them, and the ground-truth reconciliation checks our data against them boss by boss. Those comparisons are labelled reference, unlike the kill times above. They find, for example, that our interrupt armor agrees with the veteran guides for 7 of 8 bosses and mechanics checked, and that the formula gives the six veteran bosses of Stormtalon's Lair and the Ruins of Kel Voreth 1,396,641 health each where the guides say 1,000,000 to 1,200,000.

Results

Updated automatically from the instance data files and the newest simulator report of each instance in test-reports; see how.

How to read the columns:

• Level: the fidelity level reached (1 Working, 2 Complete, 3 Checked, which for an instance means tuned; see above).
• Health (data) and Damage scale: the tuning of the encounter in the data file, normal difficulty. "(formula)" means no override: the server's creature formula (the final build's tables, BaseHealth' + 14 x Stamina'; semantic registry) at the instance's normal level. A damage scale of 0.08 means a boss deals 8% of the damage the client formula gives (see how tuning is done).
• Simulated kill: median and range over the simulator runs of the newest report, and how many of the runs killed the boss. The bots are not players: read it as a relative measure, not a prediction.
• Wipes / deaths: totals over all runs.
• At Level 2 (not simulated), health is the published figure where a guide gives one and an estimate otherwise (Sanctuary of the Swordmaiden: only the Ravager's veteran health is published), and the damage scale is the 0.08 of Stormtalon's Lair; no simulation has tuned either for these dungeons.
• Retail reference (label): what the report compares against. Always guess: see "How we compare with retail".

Dungeons

Stormtalon's Lair (dungeon, level 20)

Fidelity level 3. Simulated with report stormtalons-lair-sim-2026-10-13.md (commit 64505404), 2026-10-13.

Issues: #179 (client test); #169 damage scale, #170 telegraph sizes, #171 simulator limits, #168 leftovers.

What the retail reference rests on: The dungeon doc (docs/dungeons/stormtalons-lair.md, "Timings") says: no guide gives ability cooldowns, and normal-mode videos show kills of about 1.5 to 4 minutes per boss and a 25 to 35 minute run. Those are read from the videos' lengths and descriptions, not from timed footage. The per-boss figures in the table are the simulator test's own estimates inside that span; none is cited individually.

Ruins of Kel Voreth (dungeon, level 25)

Fidelity level 2. Not simulated yet.

Issues: #177 (client test); #176 Level 3, #172 veteran, #173 veteran, #174 Gurka, #175 challenge 846.

Skullcano (dungeon, level 35)

Fidelity level 2. Not simulated yet.

Issues: #202 (client test); #207 Level 3, #203 jump-aware tremor, #204 Mordechai, #205 veteran, #206 environment.

Sanctuary of the Swordmaiden (dungeon, level 40)

Fidelity level 2. Not simulated yet.

Issues: #201 (client test); #199 Level 3, #195 Shallaos, #198 unmodelled spells, #194 Rayna challenge.

Protogames Academy (dungeon, level 10)

Fidelity level 2. Not simulated yet.

Issues: #223 (client test); #227 Level 3, #224 veteran, #225 interactions, #226 unguided fights.

Ultimate Protogames (dungeon, level 50)

Fidelity level 2. Not simulated yet.

Issues: #246 (client test); #250 Level 3, #247 player-action rooms, #248 stealth, #249 unbuilt challenges.

Adventures

Crimelords of Whitevale (adventure, level 40)

Fidelity level 2. Not simulated yet.

Issues: #262 (client test); #267 Level 3, #263 bikes and carrying, #264 CSI and escorts, #265 boss mechanics, #266 notoriety.

The Hycrest Insurrection (adventure, level 15, Exile)

Fidelity level 2. Not simulated yet.

Issues: #268 (client test); #269 Level 3, #270 achievements, #271 stand-in mechanics, #272 open questions.

The Malgrave Trail (adventure, level 45)

Fidelity level 2. Not simulated yet.

Issues: #281 (client test); #285 Level 3, #282 Path actions and carrying, #283 boss mechanics, #284 mine, resource bar, achievements.

The Siege of Tempest Refuge (adventure, level 30)

Fidelity level 2. Not simulated yet.

Issues: #295 (client test); #298 Level 3, #296 boss mechanics, #297 lanes, troop numbers, vignette, achievements.

War of the Wilds (adventure, level 25)

Fidelity level 2. Not simulated yet.

Issues: #319 (client test); #318 Level 3, #313 map: control points, lanes, #314 armies, promotions, level, #315 champion kits, #316 boss mechanics, #317 rookie challenge, achievements, rewards.

The Bay of Betrayal (adventure, level 50)

Fidelity level 2. Not simulated yet.

Issues: #339 (client test); #334 Level 3, #335 boss mechanics, #336 Technopathy, #337 places, hazards, flavour, #338 rookie challenge, achievements, rewards.

Expeditions

Evil from the Ether (expedition, level 20 to 26)

Fidelity level 1-2. Not simulated yet.

Issues: #167 (client test); #146 kits are guesses.

Outpost M-13 (expedition, level 12 and up)

Fidelity level 1-2. Not simulated yet.

Issues: #167 (client test); #146 kits are guesses.

They Came from Fragment Zero! (expedition, level 6)

Fidelity level 2. Not simulated yet.

Issues: #362 (client test); #355 Level 3, #357 beams, Boon, veteran objectives, #361 veteran, doors.

Infestation (expedition, level 19)

Fidelity level 2. Not simulated yet.

Issues: #362 (client test); #355 Level 3, #358 vignette, decompression, the fiend, #361 veteran, doors.

Rage Logic (expedition, level 27)

Fidelity level 1. Not simulated yet.

Issues: #362 (client test); #356 Level 2, #359 speeders, hoverbikes, #355 Level 3, #361 veteran, doors.

Space Madness (expedition, level 31)

Fidelity level 1+. Not simulated yet.

Issues: #362 (client test); #356 Level 2, #360 stages, sanity, #355 Level 3, #361 veteran, doors.

Deep Space Exploration (expedition, level 33)

Fidelity level 1. Not simulated yet.

Issues: #362 (client test); #356 Level 2, #359 ship transitions, #355 Level 3, #361 veteran, doors.

The Gauntlet (expedition, level 41)

Fidelity level 1. Not simulated yet.

Issues: #362 (client test); #356 Level 2, #360 arenas, dance-off, #355 Level 3, #361 veteran, doors.

Raids

Datascape (raid, level 50, 20 players)

Fidelity level 2. Simulated with report datascape-sim-2026-10-14.md (commit 7d770218), 2026-10-15.

Issues: #261 (client test); #256 pairs, #257 minibosses and trash, #258 geometry and menu, #259 unbuilt mechanics, #260 Level 3.

Genetic Archives (raid, level 50, 20 players)

Fidelity level 2. Simulated with report genetic-archives-sim-2026-10-14.md (commit 7d770218), 2026-10-15.

Issues: #241 (client test); #208 platforms and tiles, #209 phase realm, throw, #210 minibosses and trash, #211 Kuralak, Convergence, timings, #212 simulator.

Redmoon Terror (raid, level 50, 20 players)

Fidelity level 2. Simulated with report redmoon-terror-sim-2026-10-14.md (commit 7d770218), 2026-10-15.

Issues: #333 (client test); #328 minibosses and trash, #329 unbuilt mechanics, #330 places and geometry, #331 Level 3, #332 loot.

Initialization Core Y-83 (raid, level 50, 20 players)

Fidelity level 3. Simulated with report initialization-core-y83-sim-2026-10-14.md (commit 2272db22), 2026-10-14.

Issues: #352 (client test); #347 unbuilt mechanics, #348 places and geometry, #349 loot and quests, #350 Level 3 follow-ups.

World bosses

Kraggar the Earth-Render (world boss, level 50)

Fidelity level 3. Simulated with report kraggar-sim-2026-10-15.md (commit 8d2df792), 2026-10-15. Health is the client formula for a level 50 raid boss; the damage scale is 0.35.

Deaths are totals over the runs; nobody is resurrected during a pull. Issues: #184 (client test); #180 Seismic Wave, Landslides, #181 Data Cache, respawn, #182 ground telegraphs, #171 simulator limits.

What the retail reference rests on: No retail kill time is published. The 5 to 15 minute range is a guess for a world boss that an official 2015 post says was balanced for 20 players (1 to 2 tanks, 2 to 3 healers, the rest damage).

Metal Maw (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #279 (client test); #273 instant casts, #274 jump-aware Shockwave, #278 simulator, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published for him and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20.

Metal Maw Prime (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #279 (client test); #273 instant casts, #277 aura, dome, drones, #278 simulator, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published for him and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20.

Scorchwing (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #279 (client test); #273 instant casts, #274 Wind Buffet push, #275 egg, Scorchlings, yetis, #278 simulator, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published for her and she has not been simulated: the damage scale (0.4) is Kraggar's (tier 5 instead of 6), tuned for a group of about 20.

Dreamspore (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #279 (client test); #273 instant casts, #276 spore heal, hallucination, #278 simulator, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published for her and she has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20.

Gargantua (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #288 (client test); #273 instant casts, #274 jump, pulls, #278 simulator, #289 bubble, Lethargic, Euthanize, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

Grendelus the Guardian (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #288 (client test); #273 instant casts, #278 simulator, #290 bomb, tower, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

King Honeygrave (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #288 (client test); #273 instant casts, #278 simulator, #291 heal phase numbers, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

Mechathorn (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #288 (client test); #273 instant casts, #274 pull, float, #278 simulator, #292 plates, Accelerite Dust, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

King Plush (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #304 (client test); #273 instant casts, #274 jump, #278 simulator, #299 jump, Dust Devil, #303 delayed Proxy effects, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

Zoetic (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #304 (client test); #273 instant casts, #278 simulator, #300 Storm Shield, tornado, armor, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

Tower-Engineer Renhakul (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #304 (client test); #273 instant casts, #278 simulator, #301 machines, armor, the cycle, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

Sulak the Xenophage (world boss, level 50)

Fidelity level 2. Not simulated yet.

Issues: #304 (client test); #273 instant casts, #278 simulator, #302 Binding, Pheromones, the cave, #181 Data Cache, respawn.

What the retail reference rests on: No retail kill time is published and he has not been simulated: the damage scale (0.4) is Kraggar's, the same creature class tuned for a group of about 20 (#278).

Doomthorn the Ancient is retired in the final game: the client keeps his corpse and no fight, so there is nothing to build and he has no row (Level 1 is Level 2 for him; #288, docs/world-bosses/doomthorn-the-ancient.md).

Everything else

Researched, not built to Level 2

These have client data and a research doc (labelled reference, derived or guess) and no encounter script yet, so they have no tuning and no simulation. They have not been built beyond that.

Every world boss with an encounter in the data has a script (Level 2 or 3).

How to help

• Run a client checklist. Every instance and world boss with a client test has an issue with exact steps (linked under each results table). You need the WildStar client and a test account; the checklists say which GM commands to use to get there. Report what you saw: did a telegraph match the footage you remember, was a mechanic readable, did a fight feel too short or too deadly, did anything look wrong?
• Report feel and timing. For a fight you have done, tell us how long it took, how many deaths, what the group looked like, and what killed people. Say which version or era (launch, Reloaded, veteran, normal).
• Contribute footage-based timings. A video link with timestamps for the pull and the kill of a boss, or a cooldown you timed yourself, is the best way to turn a guess into a reference. It beats a guess in a report; please say if you can be cited. We label it reference with your citation in the basis.
• Fix a fight. Everything under "Issues" in the results is open. Many are small. Each fight's design doc (docs/dungeons/, docs/world-bosses/, docs/raids/) lists what is built and what is not.
• Improve the simulator. The healer logic and the resource model (#171) are the two changes that would make the simulated numbers most trustworthy.