Corrections
Every factual error this guide has made, with what caused it. Most WoW sites fix mistakes silently. A reference you cannot audit is a reference you should not trust — so this is the audit.
A high correction count on a young site is not a warning sign — it means claims are being checked after publication rather than left to rot. The number to worry about is a site with none. Forever has not launched yet, launch builds can still differ from the beta, and a guide that never changes its mind is not keeping up.
We said: Smelting made Mining “the only gathering profession you can level without leaving a city” (and “indoors”); Cooking’s Artisan quest required character level 40.
Corrected to: Only the lower smelts (copper to gold) give Mining skill; from Steel up the client gives smelts no skill-up stretch, so past roughly 185 Mining levels in the field. The Artisan quest’s level requirement is unconfirmed.
Why it happened: Both came from Classic memory. The client’s recipe tables of build 70532, used for the new profession guides, show the smelting limit, and nothing we hold confirms the level; the independent check of those guides caught both.
We said: Fishing’s Expert rank was “a book from Nat Pagle”, and the Expert books for First Aid, Cooking and Fishing were described as covering 150–225.
Corrected to: The Fishing book, Expert Fishing – The Bass and You, is sold by Old Man Heming in Stranglethorn Vale for 1 gold and needs skill 125 and character level 20; all three Expert books are usable from skill 125.
Why it happened: Our notes carried Classic memory: Nat Pagle gives Classic’s Artisan quest, not the book. The game client’s item table (build 70532) puts each book’s requirement at 125, and Wowhead’s Forever data names the vendor. Found while building the profession guides.
We said: Mage talent tooltips from Wowhead’s datamined dataset: Winter’s Chill applied by any Frost spell for +40% Cone of Cold damage per stack; Missile Barrage also triggered by Arcane Missiles hits; Elemental Precision 5 ranks for 5% hit; Ice Barrier a plain absorb with no pushback.
Corrected to: The client’s own text (build 70532): Winter’s Chill applied by Frostbolt, Frost Blast and Fingers of Frost Ice Lance, +5% Cone of Cold damage and +7% of spell damage per stack; Missile Barrage from Arcane Blast and Arcane Wrath, Fireball, Frostbolt, Frost Blast and Frostfire Bolt, no Arcane Missiles proc; Elemental Precision 3 ranks for 3%; Ice Barrier also +3% Frost damage (+3% more while shielded). The Mage levelling build puts the two freed points into Frost Warding. source
Why it happened: Effect text came from Wowhead’s dataset because the client could not return spell text in game; hotfixes after our 17 October check never reached it.
Lesson: Read the client’s own tables when they are available instead of trusting a dataset’s text.
We said: Champion of the Light 10/20/30% of Intellect; Warlock Amplify Curse +50% on Curse of Exhaustion; Hunter Improved Serpent Sting showed its 5/5 value (10%) as its only rank; a few tooltips showed Wowhead’s raw formulas (“(20 * 60)”).
Corrected to: Champion of the Light 20/40/60% (a client hotfix restored it after the 15 October notes cut it); Amplify Curse +20% on Curse of Exhaustion; Improved Serpent Sting 2/4/6/8/10%; formulas written out in words. source
Why it happened: The same dataset text, unchecked against the client for those talents.
Lesson: A rank list with one entry for a multi-rank talent is a data error, not a design.
We said: Profit and return for a recipe that makes more than one item counted the sale of one: Strong Anti-Venom (makes 3) showed a profit of 1s 75c where selling all three cleared 11s 25c at the same prices. The page also showed 1,334 reagents as bare item numbers (“item:NNNN”).
Corrected to: Sale value and profit count every item a recipe makes. Item names come from the newest client data (the 18 October hotfix data, then the 27 September item sweep); the 79 names no source has are labelled “unknown item #” and explained. source
Why it happened: The generator multiplied the vendor price by the output quantity but not the auction price, and looked names up only in the small 20 September item file.
Lesson: When one column is multiplied by a quantity, check every sibling column is too.
We said: The racials were described as the pre-15 October kits: Human Sword, Orc Axe and Dwarf Mace Specialization needing a weapon, Orc and Troll actives “flagged for change”, Troll Regeneration 10%, Tauren Endurance +1% hit, Skyborne Wind Blessed +1% haste — across Races, all nine class pages and several guides.
Corrected to: The 15 October build: Ruthless (+2% crit), Savage (+1% crit) and Earth Fury (+2% critical damage and healing), none needing a weapon; Blood Fury on a 3-minute cooldown; Shatter Curse a dispel; Regeneration 20%; Endurance +2% critical damage and healing; Wind Blessed +3% haste. Race picks re-ranked. source
Why it happened: Blizzard's 15 October notes were never folded into our race pages before the beta closed; pages kept saying the change was “expected”.
Lesson: A line that says something is expected on a date needs re-checking on that date, not just when the next refresh happens.
We said: Ice Lance deals +300% damage to Frozen targets.
Corrected to: +200% since the 15 October build. source
Why it happened: The class pages had it, but five hand-written pages kept the old figure.
Lesson: When a number changes, search every page for it, not just the class page.
We said: Elemental Mastery was removed; Call of Thunder is Elemental's 16-point talent and Earthbound its 21.
Corrected to: Since 15 October Earthbound is the 16, Lava Burst the 21, and Elemental Mastery is back as the 31-point capstone; Call of Thunder is gone. source
Why it happened: The prose was written from the 7 October deep dive; the talent data had moved on.
Lesson: Class prose that names talents has to be checked against the tree the same page draws.
We said: Moonkin Form grants party crit and is explicitly exclusive with Leader of the Pack.
Corrected to: Since 15 October Moonkin Aura raises spell crit only and Leader of the Pack melee and ranged crit only, and both reach the whole raid. Whether the game still treats them as exclusive is untested. source
Why it happened: Written from the 30 September deep dive; the 15 October build split the two auras.
We said: Levels 10–30 of the Warrior, Mage, Warlock and Priest levelling builds were badged B, “sourced from a published Forever build”.
Corrected to: Badged D: the order for those four classes is ours, built from Blizzard's class notes. The other five follow published builds and stay B.
Why it happened: One rule was applied to all nine classes, but our own class files said the order was ours.
Lesson: A tier badge applied by a rule has to be checked class by class.
We said: Frost row 2 had Permafrost in column 2 and Frostbite in column 4, as Wowhead lays it out.
Corrected to: The game client has them the other way round; the trees now follow the client.
Why it happened: We used Wowhead's layout; the 17 October client walk showed the two swapped.
We said: Improved Serpent Sting was shown as a client-verified Marksmanship talent.
Corrected to: Marked unconfirmed: its node sits off the grid in the client and Wowhead doesn't list it, so it may never show in game.
Why it happened: An off-grid coordinate was read as a real position.
We said: Each race teaches one or two Priest race spells (Undead: Dark Sacrifice).
Corrected to: Every race teaches two (Undead: Touch of Weakness and Dark Sacrifice). source
Why it happened: Our list counted only the new or changed spells from the Priest deep dive and missed the unchanged ones.
We said: Paladin is the only class without a Blizzard deep dive.
Corrected to: Blizzard published it on 15 October. source
Why it happened: Not re-checked after the date it was due.
We said: The talent pages and the calculator said the data was pulled on 9 October.
Corrected to: It was pulled on 18 October; the dates now come from the data file instead of being typed in.
Why it happened: Hard-coded dates were not updated when the data was refreshed.
Lesson: Dates that describe a data file should be read from it.
We said: Zephras Isle has no map entry in the client — a demonstrated absence, probably not enabled in a level-20-capped beta.
Corrected to: Zephras Isle is in the client's map table, under two IDs (2521 and 2665), and Blizzard listed it as week-one beta content, pack-gated.
Why it happened: The claim was never re-checked against our own 20 September map pull, which lists the isle twice — and the zone table on the same page showed it.
Lesson: When a page states an absence, check it against the data printed on that same page.
We said: Every tradeskill pays 3 Legacy Points — 36 of the 65.
Corrected to: Only the six crafting primaries pay, at 150/225/300: 18 Legacy Points. Gathering professions pay none. source
Why it happened: “Each tradeskill” in secondary coverage was read as all twelve professions instead of being counted against Blizzard's own Legacy challenge list.
Lesson: Count against the primary source, not a paraphrase of it.
We said: Spell Power always beats Intellect while levelling.
Corrected to: Below roughly level 20, Intellect wins: spells below level 20 carry a spell-power coefficient penalty (KB 400.50.2).
Why it happened: The rule was derived at mid levels and stated for the whole levelling range without checking the sub-20 coefficient penalty.
Lesson: Test a conclusion at the edges of the range it claims to cover.
We said: Auction listings “under vendor” presented as profitable flips.
Corrected to: Auctionator's vendor cache stores the merchant buy price, not the sell price; against the real vendor sell price the genuine margins are 15c or less (KB 400.50.4; 400.80.3 §6).
Why it happened: The cached vendor figure was assumed to be what a merchant pays you, when it is what a merchant charges.
Lesson: Check which side of a price a cache stores before calling a gap a margin.
We said: Caster weapons, including wands, scale with spell damage; Haste sources unknown.
Corrected to: Wands are excluded from spell-damage scaling. Haste comes only from racials, Spellstone and a flask. source
Why it happened: Blizzard's beta development notes settled both points after the pages were written; the pages were not re-read against them.
We said: Reagent-bag margins computed from guessed recipes.
Corrected to: Recipes read from the live beta client (Linen, Woolen and Silk Reagent Bags, with Cerulean Dye) and margins recomputed on them (KB 400.50.5).
Why it happened: The first pass costed recipes from memory of Classic patterns before the client recipe data was pulled.
We said: Item 1482 (listed at 20,000g) described as unnamed.
Corrected to: Item 1482 is Shadowfang.
Why it happened: An item lookup returned nothing and the empty result was read as “no name” rather than as a failed lookup (KB 400.50.5).
Lesson: An empty read is a failed read until proven otherwise.
We said: “It arrives in Winter 2026”, given as the release window on Hardcore, Launch, Legacy, PvP, Systems and Unknowns, with a timeline step placing it after the 9 December raids.
Corrected to: Blizzard's own words since the beta opened: “Hardcore will be available sometime after launch.” No season, no date. Winter 2026 is roadmap and press wording and is now labelled as such. The page also adds the twelve Hardcore Legacy challenges, the planned Hardcore perks, that Hardcore alone does not share battlegrounds, and the self-found, transfer and guild-handoff APIs in the beta client (tier B, signals not announcements). source
Why it happened: A secondary roadmap season was repeated as a date across six pages after Blizzard's ruleset article (21 September) had already used vaguer, different wording. Nobody re-read the primary source once the beta articles started arriving.
Lesson: A date has the tier of whoever said it. When the primary source gets less specific than the coverage, the coverage is not a refinement.
We said: Caps “16 spendable per character, up to 65 earnable”; “Whether Legacy points can be respecced — and at what cost — has not been announced… treat a spend as permanent”; challenge list given as demo examples with PvP “full list not revealed” and Dungeons “requirements not visible”; Reward Track thresholds 15/25/40/55 presented as the track. Elsewhere on the site, professions were counted as 36 Legacy Points.
Corrected to: 65 account-wide, 29 earnable on any single character, 16 spendable per character. A respec exists (beta: 1s, hourly decay). The full challenge list is Blizzard's: Classes 27, Tradeskills 18 (six non-gathering primaries only), PvP 12, Adventure 2, Dungeons 3, Raids 3. Blizzard names the four rewards but no thresholds, so 15/25/40/55 is marked as the demo's. Every beta player was granted 16 points on 8 October. Systems' Legacy section was rebuilt the same way, and its tree values moved from BlizzCon-demo per-rank numbers to the beta client's max-rank tooltips. source
Why it happened: Blizzard's Legacy article (23 September) and the 8 October dev notes answered most of this page's open questions; the page was not re-read against them. The 36 came from reading Wowhead's “each tradeskill” as all twelve professions.
Lesson: An unknowns list is a to-do list for the next primary source. Re-run it every time Blizzard publishes on the subject.
We said: Darkspear Islands “off the coast of Kalimdor”, level-60 content, AB-style vs EotS-style unresolved; returning battlegrounds “could not be tied to any Blizzard statement”; rank names, thresholds, decay and reward tables unpublished; Spearing Strike +80% vs mounted; Improved Counterspell a 2-second silence; Rage of the Farseer +30% melee and cast speed; Feral 16-point Mangle; Gnome Eureka +10% damage or healing; ruleset transfers “not announced”; faction balance “no specifics”.
Corrected to: Darkspear Islands is off Dustwallow Marsh, levels 30–60, four capture points plus a central flag, first to 2,000. Alterac Valley is confirmed and four battleground reputations exist. Blizzard published the 14-rank reward table, a weekly Rank Point cap, blue and epic Honor sets with Emboldened Seals, and the 12-point PvP Legacy list; the Honor cap is 25,000. Spearing Strike doubles damage; one point of Improved Counterspell always gives a shorter silence (no duration); Rage of the Farseer is attack speed only; Mangle is now Primal Bite; Eureka is a 10% resource discount. The PvP ruleset is one faction per account, rulesets cannot be switched, and battlegrounds may mix rulesets except Hardcore. World's PvP section was corrected the same way. source
Why it happened: Blizzard's 7 October PvP and Darkspear Islands articles, the class deep dives and three weeks of dev notes superseded most of a page written from BlizzCon coverage and demo data.
Lesson: The least-announced area of the game changes fastest once it is announced. A page that is mostly unknowns needs the shortest re-check interval, not the longest.
We said: Phase 2 “no date announced”, Excavation Site 24–29; Deadmines and Wailing Caverns “still unconfirmed”; the surname-uniqueness question unresolved; 16-point talents “waiting until October”. A first pass on 9 October then wrote that level 30 gives 16 talent points.
Corrected to: The cap rose to 30 on 1 October, with Excavation Site 26–31, Razorfen Downs and Uldaman; City of Dalaran opened 8 October; nine classic dungeons were already open. Surnames are shareable. Talent points are level − 9, so the 16-point talents arrive at level 25 and level 30 gives 21. Also added: the CE restock, Blizzard's GPU floor, layering, the CurseForge “Forever” flavour and the Cooldown Manager. source
Why it happened: The refresh followed a KB beta guide line that said 16 points at 30, while the KB's class overview had already corrected it to level − 9. Two KB files disagreed and the page took the wrong one.
Lesson: When a refresh touches arithmetic, redo the arithmetic. A number copied from a source is only as good as that source's own last correction.
We said: Raids “Barrow Deeps, Hyjal Summit and Onyxia's Lair” all on 9 December; Hardcore Winter 2026; which half of a name is unique unresolved; ruleset transfers unannounced; “no sharding or layering described”; Onyxia's attunement “not stated”; dungeon quest XP “raised enormously”.
Corrected to: Blizzard: “New raids unlock on December 9”; Onyxia's timing is disputed. Surnames are shareable; rulesets cannot be switched; the PvP ruleset is one faction per account; layering exists; Onyxia is the only raid with a traditional attunement at launch (tier B); the dungeon-quest bonus was halved on 1 October and kill XP raised on 8 October. GDKP is banned. source
Why it happened: Blizzard's September ruleset and naming articles and the October dev notes answered questions the page still listed as open.
Lesson: “Unresolved” is a dated statement. It needs the date it was last checked, or it quietly becomes false.
We said: Sharding/layering, Haste, attunements, returning battlegrounds, PvP reward tables, Legacy respec, ruleset transfers, the camp-object list and the beta NDA listed as unknown; Phase 2 “no date”.
Corrected to: Each is marked answered with its source; the list now leads with what is still open — population caps, lockouts, Hardcore's date and rules, rank thresholds, AH scope, Server Slam, the 15 October racial rework and the missing Paladin deep dive. Three new “said, but wrong” rows: Phase 2 on 8 October, Hardcore in Winter 2026, and 36 Legacy Points from professions.
Why it happened: Routine refresh after three weeks of Blizzard articles and dev notes.
Lesson: An unknowns page that never shrinks is not being maintained.
We said: Names “each field 2–12 characters”; reserving “is free”; Shen'dralas as a 20–30 levelling zone; Excavation Site 24–29; dungeon quest XP “front-loaded and one-time… where the experience now is”.
Corrected to: Blizzard publishes no name-length limit; surnames are shareable; reservation is included with any upgrade purchase. Shen'dralas is an endgame zone with no published range. Excavation Site is 26–31. Dungeon quests still pay a bonus, but it was halved on 1 October and dungeon XP per hour is meant to match questing. The party-level XP rule changed on 8 October. source
Why it happened: The 2–12 figure was never Blizzard's; Shen'dralas was slotted by geography, not by any level statement; the XP advice predated the October rebalance.
Lesson: A planning page inherits every error of the pages it summarises, so it needs re-checking whenever they change.
We said: Mount Hyjal and Shen'dralas “60, unconfirmed”; Excavation Site 24–29 “Phase 2”; City of Dalaran “borderline”; “all three raids unlock 9 December”; Onyxia's attunement not stated; “dungeon mob XP is cut way down while dungeon quest XP is raised enormously”; Darkspear Islands off the coast of Kalimdor, returning battlegrounds unconfirmed.
Corrected to: No level range is stated for either endgame zone. Excavation Site 26–31 opened 1 October, Dalaran 8 October, with the Horde key chain. Blizzard says “new raids” unlock on 9 December; Onyxia's timing is disputed and it is the only raid with a traditional attunement (tier B). Boss counts added: Hyjal Summit 13, Barrow Deeps 8. The dungeon-XP row now carries the 1 and 8 October rebalance and GDKP ban. PvP section corrected as on the PvP page. source
Why it happened: The page predated the October dev notes and Blizzard's 7 October PvP articles; the “60” for the new zones was secondary reporting presented in the level column.
Lesson: A level column is a claim of the same tier as its source. If the source has no number, the cell should not either.
We said: “Camps last roughly 10 minutes”; the tiers listed were “the trainer line, not the Blueprint line”.
Corrected to: Lifetime is disputed — 10 minutes (developer interview, tier C) against 15 (Icy Veins hands-on, tier B). Tier 2 is the Blueprint line; the Greenhouse Blueprint drops from Shadetooth in Excavation Site. Added the 24 September resting postures, the 1 October campfire-damage change and Boosted Rest. Systems' camping section was corrected the same way. source
Why it happened: The KB had corrected the Blueprint point on 20 September from the beta Auction House; this page was not updated with it.
Lesson: When the KB corrects itself, every page citing the old line is a follow-up, not an optional sync.
We said: Haste “unannounced… neither confirmed present nor excluded”; language preference with no mechanics; Hardcore “from Winter 2026”; two-part names with no word on surnames.
Corrected to: Haste exists (Wind Blessed, Berserking, Spellstone) but is not yet on gear; wands no longer scale with spell damage; Wizard Oils are back to Classic values; dungeon quest rewards were lowered to Uncommon on 8 October; selling an uncollected BoE now soulbinds it. Rulesets: no switching, PvP is one faction per account, battlegrounds may mix rulesets, layering exists, language is account-wide with a 24-hour cooldown, surnames are shareable. Legacy and Camping sections rebuilt as noted on those pages.
Why it happened: Routine refresh against the 2026-10-09 KB.
Lesson: Systems pages summarise; they go stale at the speed of the fastest-moving page they summarise.
We said: “Wowhead's Forever quest XP is understated by a constant factor — 3.40× in one dungeon, 3.75× in another… which fits a multiplier that rises with level.”
Corrected to: The client's quest cache stores a per-quest RewardXPMultiplier. Dungeon quests carry 2.90× (Wailing Caverns) to 4.35× (Shadowfang Keep), set per dungeon and, in Ruins of Lordaeron, per faction (3.75× Alliance, 3.20× Horde on the same level-22 quest). Wowhead shows base XP. The multiplier is not level-driven.
Why it happened: The first finding came from two compiled sources, seven quests and two dungeons. Two dungeons can't separate a per-dungeon constant from a level trend, so the level theory was a guess. Reading the client settled it, and the client data was read the same day the page went up. The page kept the old framing for eleven days.
Lesson: A pattern from two data points is a hypothesis, and the page should say so. When a better source turns up, the page is due an update that day, not whenever someone next looks.
We said: No talent trees on the site at all — the class pages carried row-by-row tables hand-transcribed off the calculator, and nothing showed the grid, the point gates or the per-rank text.
Corrected to: The full Mage trees, drawn: Arcane, Fire and Frost, 7 rows × 4 columns, every rank's effect text, the prerequisite links and the row gates. Built from the JSON payload behind Wowhead's Forever talent calculator — the same data the calculator itself draws, so the page cannot silently drift from it. The other eight classes are scaffolded and will fill in the same way. source
Why it happened: The data is Tier C and the page says so. Wowhead's own footer: “These talent trees use information gathered from testing at Blizzcon, as well as streams from the event. The data and icons will be refreshed and fully accurate once we have the beta client available to datamine.” Structure — rows, columns, rank counts, prerequisites, gates — is reliable. Every number inside the effect text is a BlizzCon demo value. Six Mage talents where this dataset and Icy Veins' beta guide disagree are flagged rather than quietly reconciled; Winter's Chill remains unresolved.
Lesson: A page is not allowed to be ahead of the KB. This one was, for a few hours, and the drift checker caught it — the trees now have a source of record behind them in 400.30.6, covering where the data came from, what tier it is and which numbers are still contested.
We said: “Mage and Warlock are the consensus first and second best solo levellers in any vanilla-lineage game.”
Corrected to: They are the two fastest casters. Hunter is at least as fast as either and is more commonly ranked first — a pet that tanks, damage at 40 yards, and the lowest death rate in the game. The page is a caster guide and its lede had quietly generalised to all nine classes.
Why it happened: A claim about the whole class roster was made on a page scoped to two casters, and dressed as “consensus” with no source behind it. It surfaced only when a second page ranked all nine classes and disagreed with this one.
Lesson: A lede that widens a page's scope is making a factual claim, and inherits the burden of proof that goes with it. “Consensus” is a citation, not an adjective.
We said: “Restless Blades does not exist anywhere in the Forever Rogue tree — a Wowhead search returns zero hits.” Published on 19 September as a correction of record.
Corrected to: Disputed, and we are flagging it rather than quietly keeping it. Icy Veins' Forever class guide, written by a hands-on beta tester, describes Restless Blades in the Combat tree with a specific effect — finishing moves reducing the cooldowns of Adrenaline Rush, Blade Flurry, Sprint, Evasion and Vanish. That is too specific to be a copy of the aggregator we originally faulted. An attempt to re-read the Forever talent calculator returned an empty page, which settles nothing either way. source
Why it happened: Our correction rested on a Wowhead search returning no results — and this site's own stated rule is that an empty read is a failed read until proven otherwise. We applied that rule to other people's data and then broke it with our own. Flawless Execution at the 16-point slot is separately evidenced and is not in question; whether Restless Blades also exists elsewhere in the tree is.
Lesson: A correction is a claim and gets checked like one. The rule you wrote down applies to you.
We said: Battle.net location: “Sources disagree — PC Gamer places it under the World of Warcraft tile, Icy Veins under WoW Classic. Check both.” Separately, Darkspear Islands listed as not testable.
Corrected to: Blizzard's beta-live article resolves both: choose World of Warcraft, then the Game Version drop-down under “In Development”. And “additional zones, dungeons, and the new battleground will open for testing as the beta continues.” source
Why it happened: We recorded a disagreement between two secondary sources without checking whether the primary source had already answered it.
Lesson: When two outlets disagree, go to the source they are both paraphrasing.
We said: Name reservation: 'how many characters — not announced. The up-to-3 figure on aggregator sites traces to no Blizzard source.'
Corrected to: Blizzard states it outright: 'You can create and customize up to three characters during this period.' source
Why it happened: We dismissed a correct figure as unsourced aggregator noise without checking Blizzard's own pre-purchase article — which this site already cited for other facts. Worse, the site's own index page said 'up to 3' while the beta page denied it.
Lesson: Check the primary source before calling a widely-repeated number unsourced. And run an internal-consistency check: two of our own pages disagreed.
We said: Dungeon level ranges attributed to Blizzard Watch and Wowhead — “not from Blizzard's own recap”; Krol'dok Stronghold marked as a disputed range; City of Dalaran carrying a competing “panel said 20–35”; the dungeon named “Alcaz Island Prison”; Riverglades given an invented “~30–45”.
Corrected to: Blizzard published its own nine-dungeon table by blue post: Krol'dok is 40–45, Dalaran 28–33, and the name is Alcaz Prison. Riverglades now reads “mid-30s to mid-40s”, which is Blizzard's own wording — they give no numeric range. source
Why it happened: We had better sourcing than we claimed and hedged correct figures with a caveat that was itself wrong. A fact-check pass also reported that this page listed Mount Hyjal and Shen'dralas as dungeons — it did not, they were already correctly listed as zones, and that finding was discarded after checking the page directly.
Lesson: A caveat that understates your sourcing is as much an error as one that overstates it. And a fact-check is a claim like any other — verify its findings before acting on them.
We said: 'No racial grants weapon skill any more — only Orc Axe Specialization and Human Sword Specialization survive, as crit bonuses.'
Corrected to: Dwarf Mace Specialization also grants crit (+1%) — so three weapon racials give crit, not two. Note it is a new Dwarf racial rather than a survivor: Classic's Mace Specialization belonged to Humans and was removed. source
Why it happened: Blizzard's Deep Dive recap names Mace Specialization in the Dwarf paragraph; we read the Orc and Human paragraphs and stopped.
Lesson: Counting survivors and counting new additions are different questions. We answered the second while asking the first.
We said: Darkspear Islands listed under 'not testable — on neither phase list'.
Corrected to: Blizzard: 'Additional zones, dungeons, and the new battleground will open for testing as the beta continues.' No date given. source
Why it happened: We inferred absence from a phase list that was never exhaustive.
Lesson: Absence from a list is not a statement of exclusion.
We said: Rogue's 16-point Combat talent listed as 'Restless Blades'.
Corrected to: Flawless Execution — reduces the Energy cost of Eviscerate by 10. Restless Blades returns zero hits across the entire Forever dataset.
Why it happened: Taken from a BlizzCon-footage transcription that was simply wrong. Checking all 54 milestone entries against real tree data found 53 right and this one wrong.
Lesson: 'Mostly right' still contains a wrong row. A 98% hit rate is not a reason to skip the check.
We said: Winter's Chill 'reworked into a self-facing proc rather than the 5-stack raid-wide Frost crit debuff — a real cut to Frost's raid-support value'.
Corrected to: Narrowed, not restructured: 'now only benefits Frostbolt and Ice Lance'. Whether it still stacks to 5 and applies raid-wide is unresolved.
Why it happened: We read a placeholder tooltip value ('stacks to 1') as a design change.
Lesson: Wowhead flags Forever tooltip VALUES as placeholders. Structure is reliable; numbers are not.
We said: 'The +5% experience bonus on crafted food is not in the item data' — flagged unverified.
Corrected to: Confirmed real. Icy Veins' beta guides state food buffs give +5% XP. The datamined tooltips simply omit the line.
Why it happened: We applied the empty-read rule to a missing tooltip FIELD and produced a false negative.
Lesson: Data not saying something is not data denying it. A missing field is not an empty index.
We said: 'Blacksmithing's Sharpening Wheel grants Attack Power' and 'Alchemy's skill-20 object is the Alchemy Lab'.
Corrected to: Client data reads Strength for the Sharpening Wheel, and the Alchemy skill-20 object as the Mana Well. ⚠️ Blizzard's own Deep Dive recap says Attack Power and Alchemy Lab — so this is a genuine source conflict, now shown as such rather than as a flat correction. source
Why it happened: We corrected Blizzard using client data without flagging that Blizzard had said otherwise in print.
Lesson: When you contradict a primary source, say so out loud rather than quietly overwriting it.
We said: Affliction's 31-point capstone listed as 'Drain Hope'.
Corrected to: Wrack — a 6 sec Shadow DoT that also raises the target's damage taken from your other Shadow DoTs by 10%.
Why it happened: Early list predated the live calculator.
We said: Repeated a claimed hidden beta achievement, 'Patience, Young Grasshopper', granting a Leeroy Jenkins pet.
Corrected to: Recorded as probably fabricated. It appears only on SEO aggregators and boosting sites; nothing in Blizzard, Wowhead, Icy Veins, Blizzard Watch or Warcraft Tavern.
Why it happened: It surfaced in a search summary and read as plausible.
Lesson: Name suspected fabrications explicitly so a later pass cannot launder them back in as fact.
We said: 'Campfire' recorded as a separate unannounced social system alongside Camping.
Corrected to: Withdrawn. 'Basic Campfire' is the object inside Camping; outlets use 'Campfire system' as a synonym.
Why it happened: Two names for one thing read as two things.
We said: Beta level cap 30; beta ends 22 October.
Corrected to: Cap starts at 20 and rises to 30 later. Beta ends 21 October.
Why it happened: Both were wrong in the first research pass and were caught by a dedicated verification pass before publication.
The patterns worth knowing
Read together, these corrections fall into four repeating shapes. They are worth stating because they will happen again:
- Placeholder values read as design changes. Wowhead stamps Forever tooltips as carrying placeholder numbers. Tree structure and tier positions are reliable; the values behind them frequently are not.
- Absence read as denial. An empty search index is a failed read — but so is a missing tooltip field. Neither is evidence that a thing does not exist.
- Secondary sources trusted over primary ones. Twice a figure was taken from a transcription or aggregator when Blizzard had published the answer directly.
- Internal inconsistency. On one occasion two pages of this site contradicted each other for two days. Cross-page checks now run on every build.
Generated from data/changelog.json · Last updated 2026-10-24 (the newest correction)