A useful score starts
with clear assumptions.

Multiply each reported stat by a weight, then add the results together. The result is a weighted score.

Specialization rankings.

Specialization pages use equipment from the WoW Forever game files and level-aware combat advice. Items the game has not sent yet use Wowhead’s Forever data and are marked provisional. You can adjust the weights and compare two items within a slot. Positive changes mean more weighted points, not a DPS gain.

Calculated: Every build has weights calculated from the game’s files for each level from 10 and each activity: spell ranks, coefficients and what each stat converts to at that level, on a simple rotation and one talent plan. The files do not hold everything, so these parts are estimates, and each build page names the stats they touch at your level and activity: rage per swing (a third-party figure), mana per Spirit and base character stats (server values), attack power ratios Blizzard has not published, and our choices for overheal, how long a pull runs per mana bar, how many mana potions you drink, how hard elites hit and how much threat counts. A Beast Mastery Hunter’s or Demonology Warlock’s pet is scored at an estimated share of your damage, since its own damage is not in the files.

Starter: below level 10, Cat Form before level 20 and the Enhancement tank option use editorial starting weights, which are tuning choices, not measured coefficients. The page says which you are looking at.

Forever class proficiency masks, item class restrictions and level requirements filter the list. Build choices add weapon constraints such as a main-hand dagger for Backstab or a shield for a tank profile. Armor and dual-wield unlock levels are enforced; lower armor tiers remain available. Training is still assumed; race, profession and reputation requirements still need checking on the item.

Only reported passive stats contribute. An item’s procs, set effects and random enchantments are not scored. Calculated weights read hit and expertise below their cap, and each build page names the cap, since a score cannot know how close your gear is. They account for weapon speed and each hand; starter weights do not. Items without usable weighted stats remain unscored. Current database records do not establish that an item is obtainable in the current game phase. Source details and patch information appear with each item.

Feral has separate Cat and Bear priorities and includes equipped weapon DPS. Healer specializations use offensive priorities for solo questing and healing priorities for group play; a calculated healer weighs how much healing your mana pays for. Tank profiles score survival and threat, as a starter mix or, when calculated, as time to live plus a stated share of threat. The level-30 class guides behind the combat advice do not validate optimal advice for levels 31–60. None of these scores reports player DPS, healing per second or survivability percentages.

Weighted score, explained.

With the default example weights, strength is worth 2 points, critical is worth 1 point, and hit is worth 0.5 points. These values were chosen to make the arithmetic easy to check.

The default current item scores 25 points: (10 × 2) + (5 × 1) + (0 × 0.5). Replacing it with this sample item changes the score by +15 weighted points.

A score isn't damage per second.

That +15 is a difference in weighted points. It isn't 15 DPS, a percentage damage increase, or a claim that this item exists in WoW Forever. The preview does not simulate combat.

What this example covers.

A deterministic comparison of artificial items in one independent slot. You can change your current item's stats and the three weights, then compare the results. Scores tied on points are ordered by their artificial item identifier.

The model keeps base stats separate from item stats. The public example uses zero base stats and accepts only the current item's values, so equipment isn't counted twice.

What comes before real recommendations.

Even calculated weights are slopes of a simple priority at one talent plan, not a simulation: no swing timers, proc interaction, movement, group buffs or encounter timing, and raid bosses are not modelled until raids open. Reliable DPS gains would need those, and procs and sets need explicit models. The lists are a starting point.

Each result already identifies its edition, catalog release, preset revision, and engine version. Those boundaries will matter as real data arrives.