Start here
How to read this page
A racial is a bonus you get for picking a race. In Forever, Blizzard rebuilt almost every racial. Weapon-skill bonuses became crit bonuses, resistances were removed, and a new race, the High Order Skyborne, was added. Old Classic advice about races mostly no longer applies.
This version also reads the talent trees. A talent such as Ruin or Shadowform makes each crit hit harder, which changes how much a crit racial is worth. The score is therefore calculated per spec, not per class. Terms with a dotted underline have a definition. Hover or tab to them, or open the glossary.
With talents included, the best and worst Alliance race for a spec now differ by about 1 to 2 percentage points of throughput. That is more than the old flat model said, but still less than one good trinket. Use the numbers to choose between races you like, and weigh them against what the score cannot measure.
Want the answer?
Go to The verdict: one card per class with the pick and the reason.
Want the reasoning?
Go to Talents and crit and Class by class.
Disagree with an assumption?
Open the Weight lab. Every ranking on the page updates.
Chapter 1
The verdict
The best Alliance race per class, using the talent-aware model. Where specs disagree, the card says which spec wants what.
Chapter 2
How the score works
The model reads the facts from the game client and applies a short list of weights. This version takes the value of crit from each spec's talents, and only the remaining weights are opinion.
Step by step
- Read what is allowed. Which races may play which class (
CharBaseInfo), and which weapons each class may train (SkillRaceClassInfo). - Read every racial's real numbers from the spell tables, confirmed against the in-game tooltips on 28 September 2026.
- Read the talents that change what a crit is worth for each spec (next chapter).
- Turn each racial into a percentage. A passive counts in full. An active counts at its uptime, its duration divided by its cooldown.
- Multiply by a weight. Crit uses the spec's talent-derived weight. Haste, damage and resource use the fixed weights below.
- Sort into two columns. Output goes in RATE, resource goes in SUST. Everything else is utility and scores 0.
RATE
Percent more damage or healing while you still have resources. Crit, hit, haste and damage bonuses go here. It matters on every fight, and it is the number the class tables rank on.
SUST
Percent more resource, so you can cast for longer. Spirit and maximum mana go here. On a fight that ends before you run dry, SUST is worth nothing. It is shown next to RATE, never added to it.
The fixed weights
| Weight | Default | What it means | How solid |
|---|---|---|---|
| crit | from talents | See Talents and crit. 0.50 to 1.42 depending on spec. The old flat value 0.55 is still available in the lab. | Read from data |
| haste | 0.90 | Nearly linear, held under 1 by the global cooldown and minimum cast times. | Judgement |
| damage % | 1.00 | 1% more damage is 1% more damage. Used for Eureka!'s +10%. | Solid |
| vs creature type | 0.12 | Share of important fights against one creature type. The client cannot tell us this. | Guess |
| Spirit | 0.35 × gate | Spirit only counts in combat through a regen-while-casting talent, so each class gets its own gate. | Derived |
| max mana | 0.60 | A bigger pool only pays on fights long enough to empty it. | Judgement |
| casts per minute | 20 | Used only to price Eureka!, which buffs 3 casts. A placeholder. | Placeholder |
The formula
Why Spirit gets a different value for each class
Mana regen from Spirit normally stops while you cast. A regen-while-casting talent lets part of it continue. Forever raised those talents from 15% to 50% for the classes that had one, and gave one to four classes that never had it. The gate is Forever % ÷ 15.
| Class | Talent | Classic Era | Forever | Gate | Human Spirit |
|---|
More regen while casting makes Spirit stronger, and it is also why mana runs out less often. Gnome's +5% max mana is worth less than it was in vanilla, and the whole SUST column is softer than its numbers look.
Chapter 3
Talents and the value of crit
The old model gave crit one flat weight, 0.55, for every class. The talent trees show that is too low for most damage specs and about right for healers.
Where the number comes from
When you crit, you deal extra damage on top of a normal hit. That extra is the crit damage bonus. In Classic, a spell crit deals 150% (a bonus of +50%) and a melee or ranged crit deals 200% (a bonus of +100%). So 1% more crit chance adds about 0.5% to a caster's damage and about 1% to a melee's.
Many Forever talents raise that bonus, or add an effect each time you crit. Those are read from the talent data below. The last factor is the share of your output that can crit at all. A Beast Mastery hunter's pet damage does not use the hunter's crit chance, for example.
Two talents list damage-over-time spells among the spells whose crit damage they raise: Shadowform lists Shadow Word: Pain and Devouring Plague, and Pandemic lists Corruption and Bane of Agony. A crit-damage bonus on a spell that cannot crit would do nothing. The model therefore assumes DoTs can crit. You can switch this off in the lab, and one Corruption on a target dummy in the beta settles it.
Crit weight per spec
Open a spec to see each talent's exact text and the arithmetic. Numbers marked assumption are not read from data.
Chapter 4
Weapons: the real cost of a weapon racial
Yes. In Forever the racial is not a weapon-skill bonus any more. The client's own text for Mace Specialization (spell 1259719) reads:
Human Sword Specialization (20597) has the same wording with swords and 2%. The weapon only has to be equipped. Every spell you cast gets the crit.
The real cost is the weapon slot. Priests cannot train two-handed maces, so a Dwarf priest has to use a one-handed mace plus an off-hand instead of a staff. If the staff is the better item, the racial costs you stats. That part of the objection is right, and the model now shows it.
Who can hold what
| Weapon | Classes that can train it | Consequence |
|---|---|---|
| One-handed swords | Warrior, Paladin, Hunter, Rogue, Mage, Warlock | Human Sword Spec works for all of them. Priests cannot use swords, so it is dead on a priest. |
| Two-handed swords | Warrior, Paladin, Hunter | Mages and warlocks must use a sword plus an off-hand instead of a staff. |
| One-handed maces | Warrior, Paladin, Rogue, Priest, Shaman, Druid | Dwarf Mace Spec works for these. Hunters cannot use maces, so it is dead on a hunter. |
| Two-handed maces | Warrior, Paladin, Shaman, Druid | Priests use a mace plus an off-hand instead of a staff. |
Break-even: when to keep the staff
A weapon racial is worth exactly its RATE score, in percent of output. If your best staff gives more than that over the best sword or mace plus off-hand, keep the staff and the racial is worth 0. This is something you can check with real items at level 60.
| Spec | Racial | Worth | Keep the staff if it is better by more than |
|---|
Weapon talents that compete with the racials
Two talents give a bonus that depends on the weapon type. They can pull a warrior or rogue away from the racial weapon.
| Talent | Axe / Polearm | Dagger / Fist | Mace / Staff | Sword |
|---|---|---|---|---|
| Weaponmaster Warrior · Arms row 5 | +5% crit | – | ignore 15% armor | 5% chance for an extra attack |
| Hack and Slash Rogue · Combat row 5 | 5% chance for an extra attack | +5% crit | ignore 15% armor | 5% chance for an extra attack |
Arms warrior. An axe gives +5% crit from Weaponmaster. A Human with a sword gets +2% crit from the racial plus the sword's extra-attack chance instead. So the sword only wins if 5% extra attacks are worth more than 3% crit (about 4.3% damage at Arms crit value). That is close, and the model does not score extra attacks.
Dagger rogues. Assassination and Subtlety use daggers for Backstab, Mutilate and Hemorrhage, and Hack and Slash gives daggers +5% crit. Neither weapon racial works with a dagger, so the model scores Human and Dwarf at 0 from their weapon racial for those specs.
Mace and armor. A Dwarf with a mace gets 1% crit plus 15% armor ignored, against the axe's or dagger's 5% crit. Armor reduction is not in the model, so a Dwarf mace build may be stronger than its score.
Chapter 5
Priest: can talents change the race?
Yes, through the healing build you pick. This chapter explains why Human is missing from the priest ranking, which kinds of talent can move a race pick, and runs a full 51-point build optimiser for the three priest specs.
Why Human is not in the priest RATE ranking
Spirit is mana, so it lands in SUST
The Human Spirit (+5% Spirit) gives mana regen. With Meditation at 50% it is worth 5.83 SUST, the best sustain on Alliance. It does not make a single heal bigger, so its RATE is 0. Human's other output racial, Sword Specialization, is dead on a priest because priests cannot train swords.
One talent turns Spirit into healing
With 400 Spirit, Human's +5% is +20 Spirit, which becomes +5 healing. Against a Greater Heal (about 2,287 effective base plus 800 healing power) that is +0.16% at 5/5. Real, but too small to move the ranking. assumption: 400 Spirit, 800 healing
Which talents can move a race pick
Talents that multiply all your damage or healing equally, such as Darkness (+10% Shadow damage) or Spiritual Healing (+10% healing), make every race stronger by the same factor. They cannot change the order. Only talents that interact with a specific racial can.
| Kind of talent | Examples from the Forever trees | Effect on the race choice |
|---|---|---|
| Crit damage bonus | Ruin, Pandemic, Shadowform, Ignite, Ice Shards, Elemental Fury, Vengeance, Impale, Mortal Shots, Divine Aegis | Raise every crit racial (Human, Dwarf, Night Elf). The biggest effect in the game. See Talents and crit. |
| Weapon-type talents | Weaponmaster, Hack and Slash; Mutilate, Backstab and Hemorrhage reward daggers | Can void Sword or Mace Specialization, or compete with it. See Weapons. |
| Spirit into power | Spiritual Guidance (Priest only) | Gives Human's Spirit a small RATE value. No other class has one. |
| Regen while casting | Meditation, Arcane Meditation, Reflection, Reverence, Mindfulness, Bestial Discipline | Multiply Human Spirit's SUST by up to 3.33. |
| Mana savings | Improved Healing, Mental Agility, Litany of Light, Soul Warding, Shadowform's −50% cost | Make mana run out less, so all SUST (Human, Gnome) matters less. |
| Intellect into power | Champion of the Light (Paladin), Mental Quickness (Shaman), Careful Aim (Hunter) | None any more. No Alliance racial gives Intellect in Forever (Gnome's became max mana). |
| Creature-type talents | Improved Tracking (Hunter, +5% vs the tracked type), Murder (Rogue, Humanoids) | Add on top of Beast/Elemental racials but do not change their value. |
| Flat % damage or healing | Darkness, Shadow Weaving, Spiritual Healing, Twin Disciplines | No effect on the order. |
The meta calculation
The optimiser reads the real Priest tree from the client: every talent, its ranks, its row and its prerequisites. It then finds the best 51-point build under these rules:
- Point rules from the data. 51 points. A talent in row n needs 5 × (n − 1) points spent in that tab first. A prerequisite must be at full rank.
- A value for every talent, in % RATE and % SUST, for each spec profile. These values are judgement and are listed below.
- Score = RATE + λ × SUST. λ says how much a point of mana is worth next to a point of output: 0.3 for healers, 0.2 for Shadow.
- Exact search. Every legal way to spend points in each tab, row by row, then every split of 51 points across the three tabs. No guessing which build is best.
- Race check. Each racial is re-priced on the winning build, because the build decides what a crit is worth.
Show the value table and assumptions
Results
What the meta calculation shows
It takes Penance and the cheap Discipline talents, then goes deep into Holy for Divine Fury, Holy Specialization and Spiritual Healing (+10% healing). It skips Divine Aegis and Power Infusion. The top three splits (23/28, 22/29, 21/30) are within 0.1 of each other, so the exact split hardly matters.
Forcing Divine Aegis 3/3 and Power Infusion gives 32/19/0 at 48.5, against 53.8 for the hybrid. For deep Discipline to be the meta, Divine Aegis and Power Infusion together would have to be worth about 5% more output than this table gives them. That is the number to watch once real logs exist.
Power Infusion needs 30 Discipline points before it, so it is your 31st Discipline point. Spiritual Guidance needs 20 Holy points before it, so it is your 21st. 31 + 21 = 52, one more than you have. This comes straight from the tree structure, not from the value table.
In the hybrid healer build, crit is only worth the base 0.50 because Divine Aegis is not taken. Dwarf is then 0.50 plus the Beast bonus, and without the Beast bonus Gnome's Eureka! (0.75) wins. In deep Discipline, Dwarf's 0.725 and Gnome's 0.75 are a tie. In Shadow, Shadowform makes Dwarf clearly best (1.00 against Night Elf's 0.83).
The default prices Eureka! at the average cast. A Gnome healer can save it for Greater Heals. If those are 1.5 times the average cast, Eureka! is worth about 1.13 instead of 0.75. Gnome then beats Dwarf in every healing build unless you count the Beast bonus, and even with the Beast bonus it edges ahead in the hybrid build (1.13 against 1.10).
Shadowform, Darkness, Shadow Weaving and Mind Flay in the Shadow tab. Twin Disciplines, Meditation and Mental Strength from Discipline. Nothing in Holy pays for Shadow.
The point rules, talent texts and prerequisites are read from the client. The value of each talent is judgement, and nobody can test a 51-point build while the beta stops at level 20. Treat the results as "what the tree structure implies if these values are roughly right". The numbers will be updated with better values after launch.
Chapter 6
The five Alliance races
The complete kit for each race, including the parts that score 0. NEW marks something Forever added or reworked compared with Classic Era.
Every resistance racial is gone, so you no longer pick a race for a resistance fight.
Every weapon-skill racial became a crit racial that works on spells too. That is why Human and Dwarf now matter to casters.
Chapter 7
Class by class
Each class has a race × spec table of RATE scores, the SUST per race, a written case and the full breakdown. The star marks the top RATE race for each spec.
Chapter 8
Weight lab
Every assumption in one place. Change one and every score, table and pick on this page updates. Red rows are specs whose top race differs from the default.
| Class | Spec | Top RATE | RATE | Next |
|---|
Chapter 9
What flips
The creature-type weight (0.12) is the least defensible number in the model. This table removes those bonuses and shows which spec's top race changes, using the talent-aware crit values.
| Class | Spec | Top RATE with the bonus | Top RATE without it | Result |
|---|
In the flat model, Skyborne's Elemental bonus decided Warrior and Rogue. With talent-aware crit, Human's sword racial wins those without any creature bonus. The guess still decides in the specs listed here, where the top race's lead is mostly a Beast or Elemental bonus: .
Chapter 10
What the model cannot see
A 0 means "this racial does not add throughput". Several racials that score 0 can still be the right reason to choose a race.
Item stats
The model cannot price the stats you give up for a sword or mace. Use the break-even table once real level-60 items are known.
Crit-triggered talents
Flurry, Hot Streak, Nature's Grace, Seal Fate, Illumination, Master of Elements and Improved Shadow Bolt all add value on a crit. They are listed per spec but not scored, so crit is if anything undervalued.
Tanking and survival
Stoneform (Dwarf) gives −10% physical damage and removes bleeds, poisons and diseases. For a tank that can be the whole argument. Threat is not scored.
PvP and levelling
Will to Survive, Escape Artist, Shadowmeld, Read Ley Line and Walk on Air score 0 but matter a lot outside raids.
Professions
Gnome Engineering Specialization belongs in the profession plan, and it is still a reason the Warlock could be Gnome.
The level-60 game
The beta stops at level 20. The talent values here are read correctly, but nobody has played a 51-point build yet.
Chapter 11
Beta checklist
Three checks change the picks above. All can be done in one evening before the beta closes on 21 October.
| Test | How | Affects | If yes | If no |
|---|---|---|---|---|
| Can DoTs crit? | Cast Corruption or Shadow Word: Pain on a target dummy and watch the ticks. | Warlock, Shadow Priest | The default model stands. Human leads for Affliction with a sword. | Affliction drops to near the flat model, and Gnome comes back. |
| Does the crit racial show on the character sheet? | Equip a sword (Human) or mace (Dwarf) and compare spell crit in the character panel with and without it. | Casters, priests | The racial works on spells as the tooltip says. | The tooltip is wrong or hotfixed. The tooltip-based scores fall. |
| Does mana run out? | On a long pull, count how often you drink or Life Tap. | Priest, Mage, Warlock | SUST matters, so Human and Gnome gain. | SUST is soft, so the RATE picks stand. |
Reference
Glossary
Every key term on this page, in plain words.
No terms match that search.