Method

What was simulated

Every spec MythicSim ships a reference build for: 28 DPS builds and 6 tank builds. That includes the weave and turret versions of the three Hunter specs, Survival in melee, both Destruction Warlock builds (Shadow and Fire), and the experimental hybrids MythicSim marks as such (Shockadin, Frostfire, Elehance, Fury-prot, Ret-prot and Holy-prot). Each is run with every Horde race the class can be (Orc, Troll, Undead, Tauren and Skyborne), at every fight length from 20 to 180 seconds in 5 second steps. Healer specs are not included. Each run is 20,000 iterations on one target with no fight-length variance and a fixed seed, so two runs of the same setup give the same numbers.

The engine

MythicSim's own open-source Rust engine, built from its public repository (main as of 2 November 2026) with no local changes: the client 70707 rules, the 26 and 27 October racial changes, the 28 October mage and hunter changes and hotfix pushes 113531 and 113554, exactly as MythicSim implements them. To check the build, every one of MythicSim's 34 published reference sims was rerun here with its own request and seed, and all 34 gave the same DPS as the MythicSim site to the decimal. Earlier versions of this site ran a patched copy of the older Go engine; that is gone, and the numbers changed with it.

Builds and trinkets

Every spec uses MythicSim's reference build as it ships it: its gear, talents, rotation, buffs and consumables. Each build wears its best pair of dungeon and quest trinkets, the pair MythicSim's own sweep measured for that spec, and that pair is the same for every race, so trinkets cannot change which race wins. The race is the only thing changed between runs, along with the fight length. This site only changes the race.

The Hunter builds come in two forms because MythicSim publishes both. The weave builds step into melee between Auto Shots and are about 10% stronger than the turret builds, which stand still and shoot (Beast Mastery 865 against 763 DPS, Marksmanship 945 against 858 on MythicSim's own numbers). MythicSim notes that how well weaving holds up in a real raid is not yet confirmed. Experimental builds are labelled as such.

Fight lengths

The Picker shows one fight length at a time, 60 seconds by default, with the others a click away. There is no combined overall ranking, because any overall depends on how often each kill length comes up, and that is a judgement, not something the sims can tell you. Each card instead lists the winner at 30, 60, 90, 120 and 130 seconds, so you can see whether one race holds across lengths or the pick changes. The Race Crossovers page uses every 5 second step to find where the pick changes.

Ties and close calls

A lead under about 0.3% is a tie, and the pages mark it. Tied races are shown level. Where the order flips between neighbouring fight lengths on a margin that small, treat it as noise, not a crossover; the Race Crossovers page does not count a lead under 0.15% for that reason.

Skyborne

The Horde side plays Windshaper Skyborne, which can be a Druid, Hunter, Rogue, Shaman or Warrior. Its damage is the same as High Order Skyborne, so the two are treated alike. Tauren has no damage racial, so Skyborne is never ranked below it: where the engine puts Tauren ahead by simulation noise or because a preset rotation copes badly with Skyborne's haste, the two are set equal.

Where to be careful

• Builds are MythicSim's defaults, so a different gear set or talent choice could move margins a little. One target only.
• MythicSim marks its tank sims experimental: they measure damage, not survivability, and threat values are engine assumptions. Tanks wear the same kind of trinkets as the other builds, so the race comparison is like for like, not a gear recommendation.
• Weaving Hunters are MythicSim's experimental references and are not confirmed in real raids; the turret builds are the safe reading for a Hunter who stays at range.
• Enhancement Shaman: the engine cannot model Flametongue Totem without Windfury Totem for that build (the preset swaps weapons), so those buff combinations were left out of its fit. Its Comp Sim numbers for that case are extrapolated.
• These are beta numbers. Racial values and talent trees have changed several times in a few weeks and may change again.

Race Crossovers

The Race Crossovers page follows each spec's best Horde race along the fight length. Every spec and race was simmed every 5 seconds from 20 to 180 seconds (20,000 iterations each, one target, the same seed). A crossover is counted when a new race is clearly ahead, by more than 0.15%, and is still ahead at the next length; its second is found by interpolating the gap between the two neighbouring lengths where the order flips, so it is good to a few seconds, not to the second. Races that trade places on a margin smaller than that stay on one chip, and a faded, dashed chip is a near tie with the runner-up. On-use racials (Blood Fury, Berserking) have cooldowns, so a crossover can sit right at a cooldown boundary such as 120 seconds.

Spec Tier List

Every spec ranked by the DPS of its best Horde race at the chosen fight length, from the same simulations as the Race Picker. Tiers are bands of the top spec's DPS: S within 10%, A within 20%, B within 30%, C within 40%, D below. The top spec is the top of whatever is shown, so adding tanks does not change a damage spec's tier unless a tank outdamages the leader. Each spec is shown with the buffs its MythicSim build ships with.

Raid Comp Sim

The Raid Comp Sim is a sandbox. It evaluates whatever raid and buff set you build and does not judge it: no required specs, no required buffs, no suggested comps.

• What was run. Each DPS and tank spec was simulated at 30, 60 and 120 seconds with each of 29 buffs and debuffs switched on or off, about 19,000 runs at 12,000 iterations plus about 25,000 more at 6,000, on the best Horde race for that spec and length.
• 90 seconds. The Comp Sim has fitted models at 30, 60 and 120 seconds. For 90 it blends the 60 and 120 second models half and half, then anchors the level to a real 90 second sim of each spec with every buff on, so the total is exact in a full raid and the buff effects are interpolated between the neighbouring models. Checked against 2,027 fresh 90 second sims of all 34 specs (random buff sets, every buff on or off alone, and none), it misses by 0.9% on average. On random half-buffed raids alone (1,012 sims) it misses by 1.5% on average, 1.3% at the median and 3.2% at the 90th percentile, and the worst single miss was 7.9%. No spec averages above 2%, which is a little looser than the fitted lengths.
• How the page answers instantly. A model was fitted to those runs for every spec and length: a base DPS, a weight for each buff, and weights for buffs that interact (armor debuffs with each other, mana buffs together). Each model was fitted on about 410 runs and checked on 20 or more random buff sets it had not seen. It missed by 0.9% on average; 63 of the 102 spec-lengths averaged under 1%, 38 between 1 and 2%, and the loosest (Feral Druid at 120 seconds) 2.0%. The single worst miss was 8.7%, on one random buff set for Frostfire Mage at 120 seconds.
• Mana. At 120 seconds the eight mana buffs get a full lookup of all 256 combinations, because running out of mana is a cliff for Hunters. At 30 and 60 seconds mana barely matters. The 130 second view uses the 120 second buff model, with race numbers from the sims at 130 seconds.
• Rules built in. Windfury Totem and Flametongue Totem do not stack (Blizzard's Shaman deep dive), so ticking one clears the other, and Windfury is kept when both would be added by default. Totems reach the whole raid within 30 yards since the 15 October build, and Grace of Air no longer blocks Windfury, but one Shaman can hold only one Air totem, so running both takes two Shamans. Mana Spring Totem and Blood Pact reach only the caster's own group. Blood Pact needs an Imp and every build runs a Succubus, so nothing ticks it by default. One Blessing per Paladin and one Curse per Warlock are flagged when the comp cannot field them.
• Campfire buffs. Each of nine camp objects (Incense Candle, Enchanted Lute, First Aid Kit, Faction Banner, Camp Chair, Sharpening Wheel, Fish Bowl, Lodestone, Mana Well) copies a class buff and is exclusive with it, so a buff is off, the class version, or the campfire version. The values are the level 60 numbers from the Endgame Tools and Conquest Capped camping guides (18 to 22 October 2026). Each object was run in the sim as bonus stats on top of the full buff set, and the page scales the class buff's effect by what the sim measured. Sharpening Wheel, for example, delivers about 64% of Strength of Earth's effect and Lodestone about 68% of Might's. Fish Bowl's +8% stats cannot be added as flat stats, so it is estimated as 80% of Blessing of Kings. The Camp Chair adds 2% crit to spells and attacks and is measured as its own bonus. The guides list it as exclusive with Moonkin Aura only, but that would let it stack with Feral's Leader of the Pack, so the sim treats it as exclusive with both auras. A Balance Druid already carries his own aura, so the Chair adds nothing to him. A Basic Campfire holds 3 objects, a Journeyman 5 and an Expert 10, each placed by a different player with the matching profession.
• Sense checks. A buff that gives a spec nothing it can use is worth exactly zero to it, whatever the fit says: mana buffs for Warriors and Rogues, melee totems and auras for casters, Hunter's Mark for anyone but Hunters, armor debuffs for non-Warlock casters, and so on. Stamina buffs are deliberately not on that list: max health feeds Undead's Touch of the Grave (5% of max health as damage), Warlock Life Tap, Bloodrage and tank rage, so Fortitude and Blood Pact are worth about +0.18% and +0.14% to an Undead player (the engine drains 5% of the player's current max health, so it does count health buffs), and the Race dropdown adds that when you pick Undead for a spec whose best race is not Undead. Each rule was checked against the held-out sim runs, and the ones the sims contradicted were dropped: Flametongue Totem and Curse of Elements both really help melee specs in this engine, so they stay. Beyond that, adding a buff never lowers a player's DPS and removing one never raises it, and Power Infusion or Innervate on a player never takes DPS away.
• Race. Each slot starts on the best Horde race for that spec and fight length. Picking another race scales the spec's DPS by that race's ratio from the Race Picker data.
• Power Infusion and Innervate are limited to one per Priest Healer and one per Druid. Only a Discipline Priest has Power Infusion (the talent sits about 30 points deep in Discipline, which neither the Smite nor the Shadow build reaches), so the Priest Healer card is the source and the damage priests are not. They are placed on the players who gain most unless you move them by hand.
• More than one on a player. Click a player's PI or Inn button again for a second copy (it turns green and reads 2x), and again for a third (gold, 3x), as long as the raid has a Priest Healer or Druid to spare. One more click turns it off. The copies run back to back, not stacked: a second Power Infusion starts when the first ends, and a second Innervate lands about 20 seconds after the first. Each extra copy is worth less than the one before. In the sims a second Power Infusion adds roughly half of what the first does at 30 to 60 seconds and less on longer fights, and a second Innervate adds nothing on a short fight and a few percent at 120 seconds, more with fewer mana buffs already up. These extra-copy effects were measured separately for the caster specs (and the hunters that use Power Infusion) from about 30 random buff sets per spec and length. For every other spec an extra copy adds nothing.
• Who casts the debuffs. A debuff only has to be cast once, so the sim does not make every Warlock cast Curse of Elements or every Warrior cast five Sunders. The buff models were fitted with each spec's own debuff casts taken out of the rotation, so a debuff is an ordinary buff that costs nobody anything until somebody is assigned to cast it. Each Warlock, Druid, Warrior and Rogue card has a button for the debuff that class can bring (Curse of Elements, Faerie Fire, Sunder Armor, Improved Expose Armor). By default the cheapest caster takes it: one Warlock for the curse, one Druid for Faerie Fire, the Rogue for Expose Armor, and every Warrior shares Sunder. Pressing a button hands the work to that player. The fight needs 1 curse cast, one Faerie Fire every 40 seconds, one Expose Armor every 30 seconds, and 5 Sunders to stack plus one refresh every 30 seconds after the first, split evenly among the Warriors assigned. The caster loses the DPS of those casts and everybody else keeps theirs, so a second Warlock gains the global he no longer spends on the curse. The cost of a cast was measured in the sim by adding one cast to a spec's rotation while the debuff was already up from the toggle, averaged over six seeds, and smoothed as a fixed number of seconds per cast (about 1.7% for an Affliction Warlock's curse at 60 seconds, 1.3% for a Feral Druid's Faerie Fire, about 3% for a Rogue's Expose Armor). Sunder costs Warriors very little because they are limited by rage, not by the global; that figure comes from the sim's own rotation (removing Arms' Sunder changes it by about 1%), not from an inserted cast, and Rogue and Faerie Fire costs are rounder than the rest. Battle Shout is still cast by every Warrior build. Who casts what is saved in the link.
• Tanks. The six specs on MythicSim's tank list are simulated like any other spec, with an attacking boss, and their damage counts toward the raid total. The Fury-prot, Ret-prot and Holy-prot hybrids are Discord-community builds that MythicSim runs through its tank rotations. Protection and Fury-prot Warriors can be assigned Sunder Armor like any Warrior, and Bears Faerie Fire.
• Not modelled. Healer slots bring buffs but have no DPS. Buffs are raid-wide rather than per party. One target, no movement, no deaths.
• When class changes land, the simulations are re-run and one data file is swapped.

Changes

• 1 Nov 2026 Rebuilt everything on MythicSim's current Rust engine and reference builds, replacing the patched Go engine, the two setups (MythicSim and TECHNIQUE Sim) and the provisional Arcane Mage. All 34 MythicSim builds are in, including the turret Hunters, Survival in melee, both Destruction Warlocks and the experimental hybrids, each with its best trinket pair, and the Race Picker, Race Crossovers, Spec Tier List and Raid Comp Sim all run on the new numbers. Spec names changed to MythicSim's. Expect numbers to differ from earlier versions, especially for mages (the 28 October talent changes are now MythicSim's own) and for Balance and Feral Druid.
• 30 Oct 2026 Method page only: extended the Provisional Arcane Mage section with the three-rotation test (MythicSim's rotation, WoWSims' live rotation, Pyroblast precast). No simulation numbers changed.
• 29 Oct 2026 Applied the 28 October hunter hotfix and re-simmed all six hunter builds: the Race Picker, Race Crossovers, Spec Tier List and the Raid Comp Sim's hunter models. The ranged builds lose about 0.5 to 0.9% (the Trueshot Aura change), the Survival weave build about 0.4% and the Survival melee build nothing. Beast Mastery and Marksmanship were briefly published about 4 to 13% lower than this; that came from my wrongly treating Aspect of the Hawk's rank 5 as its top rank, and it is corrected. Arcane Mage is simmed on a provisional build for the unconfirmed 28 October mage changes (see "Provisional Arcane Mage" below); the other mage specs are unchanged. The Raid Comp Sim now lets one player carry a second or third Power Infusion or Innervate (click the button again), with the extra copies measured in their own sims. The Race Picker got class colours, and the "One race the whole way" section on Race Crossovers was removed.
• 28 Oct 2026 Added the six tank specs to the Race Picker, the Race Crossovers page (formerly the Race Atlas) and the Raid Comp Sim, removed the raid-spec view from the Picker, launched the Raid Comp Sim and added the Spec Tier List. Added campfire buffs and the Windfury and Flametongue stacking rule to the Raid Comp Sim.
• 28 Oct 2026 Added the TECHNIQUE Sim setup (shared trinkets, cooldowns at pull and Recklessness), the Horde Race Picker and this site. The Picker defaults to a 60 second kill with no combined overall. Fixed a pre-pull timing error that understated Skyborne on Balance Druid and Elehance Shaman. Confirmed Undead's 15 second Touch of the Grave against the MythicSim site.