How thissite works

Everything on this site is read from Blizzard's official Game Data and Profile APIs. Nothing is scraped from other websites, and no numbers are copied from another meta site. If a figure appears here, it came from an authenticated API response about a real character.

The crawler reads the official leaderboards, then takes up to the top 100 characters per spec in each bracket and fetches each one's talents, gear, enchants, gems and in-game stats. Those characters are the entire basis of every percentage on the site.

That cap is applied to the spec a character was ranked with. We group them by the spec they were actually in when their profile was read, and those two can differ — someone who ranked in one spec and then switched is counted on the other spec's page, with the gear and talents they were carrying at that moment. So a page can show somewhat more or fewer than 100 players. The exact count is always stated on the page itself; never assume it is 100.

Region is EU and the Mythic+ season is Midnight Season 2. Blizzard counts PvP seasons separately, and the rated PvP data here comes from Midnight Season 2. Regions are never merged: every number on this page is measured only from the EU ladder, and a build that is popular in US does not affect it. Use the region switch in the header to see the US measurements instead.

A percentage is almost always how many of the sampled players had that thing — not a rating, score or recommendation. “82%” on a talent means 82% of the measured players had that talent selected.

The denominator changes in five places, and each one is labelled on the page:

• Rings and trinkets — every character contributes two, so the denominator is twice the player count.
• Gems — measured as a share of all gems socketed, not of players, because one character can socket several.
• Hero talents — on an unfiltered page these are divided across two hero trees. Pick a hero tree tab to see that tree's own numbers.
• Embellishments — a share of every embellishment in the sample, not of players, because one character can carry more than one.
• Stat priority — the one place a percentage is not a player count. Each stat's figure is its share of the secondary stat rating sitting on the sampled players' gear, averaged one player at a time so that better-geared characters cannot dominate. The four shares add up to 100%. The dimmer “in game” figure beside it is the real character sheet percentage, which is measured on a different scale and is not comparable between stats.

Sample size is not uniform. A popular Mythic+ spec may rest on a hundred or more players; a quiet PvP bracket can have only a handful. In the current data, 155 of 680 build pages rest on fewer than 20 players.

A spec needs at least 5 measured players before it gets a page at all. At that size a single player is worth 20 percentage points, so those pages carry a warning at the top. The warning is not decoration — read those percentages as a weak signal, not a consensus.

There are no “AoE build” and “single target build” labels here, on purpose. Blizzard's API does not say what a build is for, and we checked whether the split could be inferred: in most specs the second most popular build is a minor variant of the first, and for most specs the top Mythic+ and raid builds are byte-identical. Adding those labels would mean inventing a distinction the data does not contain.

Instead, each spec page shows where players actually disagree — the talents whose adoption sits between 30% and 70%. Those are the real decisions; the rest of the tree is close to unanimous.

The import codes are also not a consensus. A talent loadout is a string of 60-plus choices, and two players almost never match byte for byte — a single filler point changes the code. So each code is one top player's complete loadout, and we show it as that rather than attaching a share percentage that would read as “hardly anyone plays this”.

Also outside what the numbers can support:

• Popularity is not performance. A talent taken by 90% of players is common, which is evidence but not proof that it is optimal.
• Gear reflects what top players have, which is shaped by drops and vault luck, not only by what they would choose.
• Ladder positions are the character's rank in our sample, not an official Blizzard ranking.

Blizzard publishes no tier list, no DPS numbers and no combat logs, so this site cannot rank specs by power. Deriving it from rating does not work either, and that is worth spelling out: the ladder is already sorted by rating, so every player we scan sits in the same narrow band. In 3v3 the scan spans 2440–3140 rating, and the median rating of the top eight DPS specs falls within 72 points of each other. That number measures where we cut the ladder, not how strong a spec is.

What can be measured is concentration: inside one pool, how much a spec's share grows in the top band compared to the pool overall. Numerator and denominator come from the same players, so where we cut no longer skews the answer. A spec at 2.0× is twice as common at the top as its overall share would predict.

It is shown for Mythic+ (measured over every scanned player, not a per-spec sample) and for raid. The top band is not a fixed threshold — it is the highest scores covering about a fifth of the pool, because key levels drift upward as a season goes on and a hardcoded “+20 and above” would cover half the field within weeks.

The tier list is that number and nothing else. A spec is placed against the median spec of its own role: S is at least twice that median, then 1.25×, 0.85× and 0.6×. Those boundaries are fixed and published rather than redrawn each week, so a tier means the same thing on every visit — and a tier is left empty when nothing reaches it.

It covers Mythic+ only. A raid tier list was built, measured and then dropped. Each spec's share carries a 95% confidence interval, and in the raid pool — 31 to 118 players per spec — that interval is wider than the entire spread of the data (±1.10× on average): only 1 placement in 24 fell inside a single tier, and the tank specs do not reach the minimum pool at all. Publishing tiers on that would have claimed a precision the sample does not contain. The same test on Mythic+, where a spec has between 475 and about 14,000 players, is ±0.10× and puts nearly half of the placements inside a single tier. The tier page states how many of its placements are firm rather than leaving that unsaid, and hovering a spec shows its figure together with the margin behind it.

It is not a power rating, and no tier list built on the official API can be. Blizzard exposes no damage, healing or combat log data. A genuine strength ranking needs parses or simulation, both outside this site's only source. What is here answers a narrower question: who actually reaches the top — which spec strength shapes, along with who plays the spec and how demanding it is.

Two limits are worth knowing:

• Compare within a role only. The roles sit at different baselines — in Mythic+ the tank specs sit at a 0.73× median against 0.91× for DPS — so a tank's figure is only meaningful next to another tank's. The tier list already does this for you: every spec is placed against the median of its own role, never against the field.
• The raid pool is not a raid ladder. Blizzard publishes no raid ranking, so it is built from the top Mythic+ and PvP players who have a Heroic or Mythic kill in this season's raid — earlier raids in the tier are not counted. It describes those players, not the raiding population.

Each region is crawled every 8 hours, and the two are staggered, so something on the site is refreshed every four hours. A run can fail or be interrupted, so no page claims to be current just because a schedule exists: each page shows the date its data was actually collected, and pages older than 16 hours (two missed runs) carry a visible stale badge.

The regions are measured separately and never merged, so their dates differ — the badge and the date on every page belong to the region you are looking at.

The data on the site right now was collected Sep 26, 2026, 3:24 AM.

If a number here contradicts what you see in game, that is worth knowing — a wrong figure is worse than a missing one. The quickest way to check us is the leaderboard: it lists the actual characters every figure on this site is computed from, so you can look one up in the armory and compare it against what we report.

If it still looks wrong, tell us at support@boomkin.io. Naming the spec, the mode and the figure you disagree with makes it much faster to check.