What's changed

Recent changes to how CoffeeAudit audits your character. Only things you would notice are listed here.

People who left the guild stop appearing on the roster

31 August 2026

The character list was keeping everyone it had ever audited under the guild, including people who had since left. Nothing in the tool had any idea that leaving a guild was a thing that happened: a character’s guild was recorded once, when they were audited, and never revisited.
Each sweep now compares the list against the guild’s actual roster. Anyone we still have filed under the guild who is no longer on it gets looked up once more, and whatever guild they are in now — or none — is what gets recorded. They drop off the roster from that point.
Nothing is deleted. Their character page and their whole audit history stay exactly where they were; they simply stop being counted as one of us. If they come back, the next sweep picks them up again like anyone else.
One case this cannot resolve: a character who was deleted, renamed, or moved to another realm can no longer be looked up at all, so there is no new guild to record and they stay on the list. Rare, and better than guessing.

Set piece tooltips were always one piece short

30 August 2026

Hover a tier piece and Wowhead’s tooltip tells you how much of the set you have. It was undercounting every single time, by exactly one — the piece you were hovering. Two Voidheart pieces each read as one; three Corruptor pieces read as two.
The page tells Wowhead which items you are wearing so it can work the set bonus out. It was sending every OTHER piece and leaving out the one under your cursor, on the assumption that Wowhead would count that one itself. It does not — it counts only what it is told.
The practical cost was a set bonus that looked inactive when it was active. If you have four pieces and the tooltip insisted you had three, the four-piece bonus read as unearned. Your gear was always fine; only the count was wrong.

Items with an apostrophe in their name are hoverable again

30 August 2026

A suggestion like "BiS is ‘Vindicator’s Pendant of Dominance’" gave you the name and nothing else — no link, no tooltip, nothing to hover. Anything with an apostrophe in its name was affected, which in this game is most of it: every Vindicator’s and Gladiator’s piece, every tier token, Cataclysm’s Edge. 2895 of the item names we ship carry one.
The cause was mundane. The page finds an item name by looking for the quote marks around it, and it was treating the apostrophe in the MIDDLE of a name as the closing quote. So it went looking for an item called “Vindicator”, found none, and gave up — while the rest of the name printed normally, which is why it looked like an ordinary unlinked sentence instead of a bug.
Names with an apostrophe inside them, like Al’ar, were never the problem and are still fine. Nothing about which item is recommended changes here — only whether you can hover it.

Hunters and feral druids are no longer told to oil a weapon they never swing

30 August 2026

If you are a hunter, your melee weapon is a stat stick you never swing. If you are a feral druid, you fight in a form that replaces your weapon’s damage entirely. An oil or a sharpening stone does nothing for either of you — and the audit was asking for one anyway, on every boss.
That is now dropped, and not just quietly hidden: those weapon slots read "n/a" and count toward neither side of your Weapons score, the same as a shield or an empty hand always has.
This reverses an earlier deliberate decision, so it is worth saying why it changed. The rule was made on the view that ranged dps ought to buff their main hand anyway. Counting it across the whole database told a different story: 13 bare hunter main hands, but 104 of 157 feral ones. About a third of every weapon finding on the site was aimed at somebody who could not have acted on it, which is not a standard — it is noise.
Balance and restoration druids are unaffected: you cast with the staff in your hands, so a wizard oil is expected of you exactly as before. The exclusion is per specialisation, not per class.
Raids already recorded keep the old finding until they age out after thirty days. Nothing rewrites history here — a night that was already audited stays as it was audited.

Every raid preparation claim now links back into the log

29 August 2026

The raid preparation section told you which log a claim came from and left it there: a report code you had to copy out and go hunting with. Now it takes you to the pull. Every boss name in the grid is a link to that exact fight in Warcraft Logs, and the small citation line under each finding opens the report it cites.
That matters most for the findings that only count. "No oil on 2 of 9 bosses" names the two underneath it, and each of those names now goes straight to the pull it is talking about — so a claim about your raid night takes one click to check instead of a search.
This works on every raid already on the site, not only on ones read from tomorrow onward. Both things a link needs — the report and the fight — were already recorded for every pull the site has ever stored; nothing had to be re-read to make the old ones clickable.
One honest limit: a link opens that fight for the WHOLE raid, not filtered to you. Filtering to one player needs an id the site does not keep, and it could only be recovered by re-reading all 424 reports — which would help nothing already here.

Weapon buffs have names, and a fishing rod is no longer a mistake

28 August 2026

The weapon columns said "buffed" because nothing this tool could reach knew what an oil or a stone is called. They now say "Adamantite Sharpening Stone", "Superior Mana Oil", "Deadly Poison VII", "Windfury Weapon" — twelve of them, read out of your own raid logs and named by hand, because there was no list anywhere to copy.
One of those names fixed a bug you would otherwise have met on your next Serpentshrine night. The Lurker Below has to be fished up, so you swap to a fishing rod for that pull — and the audit was checking your rod for a sharpening stone. Anyone who fished without a lure on was told they had forgotten a weapon buff. A rod is not a weapon and is no longer treated as one; the off-hand check had always known that, and only the main hand was assuming.
It came to light because a fishing lure turned up among the unnamed enchants and somebody recognised it. It had never produced a complaint precisely because the lure was there — it was the raiders who fished without one who were being accused.
An enchant we have not named yet still reads "buffed" rather than a number, and the sweep now says which ones it could not name so they can be added. If your weapon shows "buffed" on a recent raid, that is a gap in our list, not in your gear.

Raid preparation is its own section now, and it names things

28 August 2026

The preparation panel says what you actually used. Each boss gets a row naming the flask or the matched pair of elixirs and the food you had up — "Flask of Relentless Assault", "Skullfish Soup" — instead of a tick you had to take on trust. It is a plain section on your character page, always visible, with nothing to click open.
Preparation problems are no longer mixed into your gear warnings. They were in the same list, so a missing wrist enchant and a night without food read as the same kind of problem. They are not: one is a gap in your gear, the other is something you forgot to bring. Preparation findings now sit in the preparation section under their own heading — and they name the bosses they mean, not just how many. "No food buff on 2 of 9 bosses" now says which two.
The "Prepared on N of M bosses" badge is replaced by two numbers that cannot argue with each other: "Consumables 7 of 9 · Weapons 5 of 9". The old badge counted consumables only, so a green "prepared on 9 of 9" could sit directly above a red weapon column. That was a limit we disclosed when the checks first shipped, and this is the fix for it. What counts as prepared has not changed — only what is said about it.
The weapon columns still read "buffed" rather than naming the oil, stone, poison or imbue. That name exists in no data this tool can reach: a weapon buff never arrives as an ordinary buff, only as an enchant id that neither Wowhead nor the game databases we vendor will name. The ids are being collected from real logs now and named by hand off the reports they came from. Until that table exists, "buffed" is the honest answer; a raw number on the page would be worse than none.
Names arrive one raid at a time. Older raids were recorded before there was anywhere to keep the names, so every report has to be read again — newest first, a batch per sweep. Your most recent raids gain names first and the rest follow over the next few passes. A raid still reading "present" instead of a name has simply not been re-read yet.

What you actually drank, boss by boss

28 August 2026

Your raid appearances now show what you had running at each boss: food, a flask or a matched pair of elixirs, and an oil or stone on your weapons. There is one row per boss, so "not prepared" is a specific claim you can check rather than a verdict you have to take on trust. None of it needs JavaScript.
Four things can now be raised against a raid night: no food, no flask or elixir pair, no oil or stone on a weapon, and no potion used all night. Each names the bosses it means. The first sweep will be louder than you expect — measured against our own logs, a third of eligible weapon slots were bare, and 10 of 25 raiders had at least one boss they went into short. That is the new checks working, not a fault.
Only boss fights count, and only distinct bosses — five Vashj wipes are one boss, not five. "Missing" means you did not have it when the pull started; losing a buff to a death mid-fight is not held against you. Off-hands are only checked when they hold a real weapon, and we do not check whether a stone suits your weapon type, because the game already refuses the ones that do not.
Scrolls and drums are recorded but can never be a finding. They are worth seeing and not worth requiring, which is this tool's usual line: it reports what a raider is expected to bring, never what the most dedicated one could.
One honest limit. Hunters are flagged for a bare main hand, because the rule is that ranged damage needs a main-hand buff; if you never melee, an oil there does nothing for you and the finding is noise.

Your raids are audited now, not just your logout gear

28 August 2026

The officers put it plainly: "If I log off Tirii in feral gear it gets proper confused. If Sly logs off in PVP gear likewise." They were right, and it was never fixable from the armoury alone — Blizzard only refreshes your profile when you log out, so the tool could only ever describe whatever you happened to be wearing at the time.
So it now reads your raids from Warcraft Logs and audits the gear you actually raided in. Your character page keeps its Armoury tab, as of your last logout, and gains one per role you have raided. A druid who tanks and cats gets both, each judged against the right caps, instead of one confused audit against the wrong ones.
Every appearance is audited, not just the most recent or the best one, and it does not matter who uploaded the log. Raid appearances are kept for 30 days; your armoury history keeps its own 90.
This is not attendance tracking and will not become it. It records what you wore and what you had running, not who turned up. The bot also stays quiet about raid runs — a finding about a raid three weeks ago is not news worth pinging anyone over.
Raids appear as the logs are read, a batch at a time rather than all at once, so a backlog takes a few passes to work through. A character with no raid tabs yet simply has not been reached.

Enchant advice says what it actually means

27 August 2026

"Head could use a better enchant" — followed by "EP gain: -12.76". Reported, and fair enough: the sentence and the number were arguing with each other. The advice was right. When your enchant differs from the best-in-slot list, we tell you, and the list's choice can genuinely score lower on paper — Glyph of the Defender loses to Glyph of Ferocity on attack power while being the thing that keeps a Phase 3 bear from being critically hit. That is a real trade the list has already weighed.
So the wording now matches the claim. A differing enchant reads "your enchant differs from best-in-slot", and only a suggestion that genuinely outscores what you wear is called better. The number is labelled "EP change" rather than "EP gain", because it is signed. Nearly four in ten enchant suggestions were affected by this wording.
Enchant suggestions are also scored against your caps now, the way gear already was. Before this, a hit-capped rogue could be told to put more hit on their boots and a hit-capped warlock more hit on their gloves — stat that does nothing once you are at the cap.
And the bigger one, from the same player: they were ALREADY immune to critical strikes, and were still being told to make that swap. We now only tell you your enchant differs from the list when the list's choice is either better for you, or is closing a cap you are actually short of. If it is worse for you and you do not need it, we keep quiet — a Combat rogue was being warned to give up 12 agility for boot run speed, and a mage to lose 2 EP on a shoulder inscription. Both of those are real guild characters, and across the tool's own lists roughly half of all enchant suggestions were of that kind.
This is judged on cap headroom, never on the EP number, because the number cannot tell the two cases apart. Glyph of the Defender scores negative for a bear either way — it trades attack power for defence. For a bear under the line that trade is the whole point, and you still get told. For one already over it, it is just lost threat.

Gear advice will not un-cap you any more

27 August 2026

Reported by a bear tank: the audit was suggesting best-in-slot pieces that would have left them able to be critically hit. It was right about the pieces and wrong about the timing. A best-in-slot list reaches crit immunity as a COMPLETE SET, and often by very little — the Phase 3 bear list clears the line by 0.12 defence rating. Take one piece out of it, keep the rest of your own gear, and that property does not come with it.
Nothing checked this, because nothing could: the tool ranks gear with stat weights, and a weight is a slope while a cap is a cliff. Losing the last point of defence before crit immunity scored the same as losing the first — about a third of a point — when in the game it costs you the whole thing. Measured across the tool's own tank lists, 249 suggestions did this, every one of them shown as an upgrade with no warning attached.
A suggestion that would cost you a cap your gear already reaches is now withheld, and the audit says which slots went quiet and why. This covers hit and expertise too, not just crit immunity. Run against the tool's own lists it holds back about one suggestion in 25, and closer to one in five for bears and protection warriors — a real character wears a messier mix than a list does, so treat those as the shape of it rather than the rate you will see.
The pieces themselves are still the right target — they are only unsafe on their own. Once the defence, hit or expertise is made up in another slot they come back.
Enchants get the same guard. A head glyph is 16 defence rating on a bear — nearly 7 points of defence skill — so an enchant is every bit as able to cost you a cap as a piece of gear is.

Bears get the tanking set, not the threat set

15 August 2026

The best-in-slot lists for feral tanks at Phase 3 and later have changed to sets that actually keep you immune to critical strikes. The old ones did not — at Sunwell the list was about 48 defence rating short, which is a boss critting you.
This turned out not to be a mistake in the lists. Burning Crusade feral tanks had two recognised gear sets per tier: a mitigation one and a threat one, and the guides of the day said outright that a threat set is not supposed to be crit immune, because you only wear it on content that cannot hurt you. Our source published just one bear list per phase and it was the threat one. A tool that hands a tank that list and then warns them they can be crit is arguing with itself, so the tanking set is now the default. If you want the threat set, that is a real choice — but it should be one you make knowingly.
Tanks were also being judged very slightly too kindly. Defence skill only counts in whole points in this game — 10 defence rating and 11 both round down to the same 4 skill — and the tool was counting the fractions. That let a character who could still be crit be told they were safe, which is the wrong way round for a tank to be wrong.

Healers are finally judged on their healing

15 August 2026

Restoration druids, holy paladins and restoration shamans were being scored on about a third of their gear. TBC healing pieces carry two separate lines — a big healing number and a smaller spell damage one — and the tool was reading only the second. The clearest case: a restoration shaman was told the ELEMENTAL Tier 6 chest was a 48% upgrade over their own restoration one, off the same token, from the same boss. Their own tier piece now leads by 43%, which is the right way round.
The old numbers were never TBC numbers. They came from the simulator we take our data from, and they turned out to be Cataclysm weights — two expansions later — carried into a Burning Crusade project by a repository rename and never replaced. The three healer simulators behind them register no spells at all and are switched off on the simulator's own site, so nothing was ever measured for them.
The replacements come from published Burning Crusade sources: Icy Veins for shamans, the seventyupgrades community default for paladins, the sixtyupgrades one for druids. The druid set was the hard one — nothing turned up in any guide, and it was briefly worked out by hand before the published version was found hiding inside a gear export. The reasoning behind every number, and what each was before, is written into the data file next to it.
Mana is now valued consistently across all four healers, priests included. The priest weighting undervalued mp5 by roughly thirty times against every other published healer source, which pushed priests away from gear that keeps them casting.
One honest limit: haste for a restoration druid is a threshold, not a slope. Enough of it fits a fifth spell between Lifebloom refreshes, and below that it does very little — your heals over time do not tick faster for it in this expansion, and they cannot crit at all. No single number can express a threshold, so treat the tool's view of druid haste as a rough guide rather than an answer. The source we took these numbers from publishes a second version with haste removed entirely, for exactly this reason.

Holy and Discipline priests get advice again

14 August 2026

Gem, upgrade and best-in-slot suggestions now run for healer priests. They were silent for every one of them, and not because there was nothing to say: the tool had no way to weigh healing gear at all, so it could not compare two items and said nothing rather than guess.
The weighting comes from a Holy priest stat priority the guild supplied and covers Discipline too. Healing power is the yardstick; a point of haste is worth about two and a half times a point of it. Spell damage counts for nothing on purpose — a healing item's tooltip figure already includes the damage it gives, so paying for both would price the same line twice.
This does not fix restoration druids, holy paladins or restoration shamans. Their weighting still values spell damage and ignores healing, so their suggestions are being measured against the wrong third of their gear. Their own numbers are the next thing to put right.

Gem advice follows your best-in-slot list

14 August 2026

Where the list has an opinion about the gems in a piece, that is now what you are told to fit. The tool used to ship the list's gear and then score its gems itself, so it recommended against the very list it had just handed you — 358 warnings across the guild, all of them the tool arguing with itself. Slots the list does not pick still get a scored suggestion.
Eleven of the lists carried a meta gem their own other gems could never switch on, so the meta sat inert and its stats did nothing. Every list now lights its own meta. Four hunter and six warlock lists take their gems from the simulator's version of the same build while keeping every item; the healer priest's Phase 4 list had no such version to borrow from and one gem was chosen by hand.
Gem suggestions were being compared against the wrong sockets whenever a helm's meta socket was not listed first, which shifted every coloured gem one socket along and quietly gave up the socket bonus. On a Phase 1 bear helm that made the list's own gems look 3 points worse than what was worn when they are in fact better — so the answer came out backwards, not merely imprecise.

Findings are ranked by what they actually mean

14 August 2026

Critical was firing on 116 of 120 characters wearing their list to the letter. A mark that lands on nearly everyone who did everything right is a banner, not a warning. It now lands on 52, and nothing has been hidden — every finding still appears, they are just ordered honestly.
Expertise is a Warning rather than Critical. No physical damage list reaches that cap and none tries: a Fury warrior in the full Phase 2 list is 78 rating short, a Combat rogue 24, and unlike hit no raid buff supplies any of it. It is a stat to gear toward, not a defect. Hit and defence are still Critical.
Hitting a cap exactly now counts as hitting it. The requirement is worked out from conversion rates carried to six decimal places, so a mage needing "164 spell hit" was really being asked for 164.000005 — and both Phase 4 mage lists, which put exactly 164 on you, were told they had failed. Five millionths of a rating point is not a gear problem.
A slot your list deliberately leaves bare is no longer a Critical. The ranged slot is enchanted in all 18 hunter lists and none of the other 108, because a scope is worth having on a hunter's weapon and not on the stat stick every other melee parks there. A hunter with no scope is still Critical, because their list does name one.

Hit rating keeps its worth past the cap when you dual-wield

14 August 2026

The 9% hit cap is the cap for special attacks. Past it your Sinister Strikes stop missing — but your ordinary swings do not, because dual-wielding carries an extra 19% miss chance that cap says nothing about, and white damage is most of a Combat rogue's output. Hit past the cap was being priced at almost nothing, which is why the tool disagreed with lists that stack it deliberately.
It now keeps about 60% of its worth past the cap, for dual-wielders only — read off the weapon actually in your off hand, so an Arms warrior and a shield are unaffected. That share is an estimate and is meant as one, but it is far closer than treating the stat as dead.
This is what was wrong underneath the ranged-slot half of the route problem below. For a Phase 2 rogue the Arcanite Steam-Pistol now scores above the Phase 1 Sunfury Bow, where it used to score below it.

Tanks: crit immunity counts resilience and talents

14 August 2026

All 18 tank lists were being reported short of the crit-immunity cap — the tool's own reference gear failing its own check on every single tank. It was reading defence rating and nothing else.
Resilience now counts, at 1.5 defence rating per point, and so do the talents that buy defence skill: Anticipation for warriors and paladins, and Survival of the Fittest for bears. Eleven of the 18 now come out immune, with four more within 4 rating of the line. Bear Phase 3 through 5 are genuinely short, which is a question about those lists rather than about the check.

Upgrade routes no longer send you backwards

14 August 2026

The route to best-in-slot has stopped listing gear from phases you already left behind. A rogue wearing a Phase 1 ring was being sent to earn a pre-raid reputation ring first, on the way to the Phase 2 one they actually wanted. The note on 4 August said a step is only shown when it beats what you already wear -- true, but not enough, because a stat you have already capped counts for almost nothing and that let older pieces score higher than the gear you had on.
A slot where you already hold this phase's best-in-slot now says nothing at all. It used to keep talking and would name an earlier phase's item as your goal, so a fully geared rogue was told a Phase 1 trinket was their Phase 2 target. Rings and trinkets still count either way round -- wearing the right one in what the game calls the other slot has never counted against you.
Sidegrades worth a fraction of a point are gone from the route. They were being listed as though they were a plan.
Trinkets and relics whose worth is a proc the tool cannot measure -- Warp-Spring Coil, Shard of Contempt, Blackened Naaru Sliver and 38 others -- are no longer treated as replaceable. Their procs score as nothing here, which made pre-raid trinkets look like upgrades over a current best-in-slot one. Those slots are now left alone rather than guessed at, and still tell you what the list picks.

Discord commands work wherever you allow them

6 August 2026

The bot no longer has its own channel allowlist. Which channels and roles can use each command is now set in Server Settings → Integrations → CoffeeAudit, so officers can change it without a redeploy.
Discord allows a newly registered command everywhere by default, so set those overrides if some channels should stay off-limits.

Discord: professions

6 August 2026

/profession <character> <profession> <skill> records who can craft what, and /professions shows everyone grouped by profession — so "who can make me these boots" has an answer.
This is self-declared and nothing checks it against the game: Blizzard's Classic API exposes no professions at all. Every reply says so, and shows how old each claim is.

Discord: /digest — how the guild moved this week

6 August 2026

/digest shows how many characters were audited, whether Critical findings went up or down, who improved most, who reached zero, who went backwards, and who still has the most to do. Defaults to the last 7 days; pass days to widen it.

Discord: /compare two characters

6 August 2026

/compare <first> <second> puts two guild members side by side: their Critical, Warning and Info counts, then what each has wrong that the other does not.
Problems they both share are left out on purpose -- those are the ones a comparison has nothing to say about.

Discord: /bis shows what you still need

6 August 2026

/bis <character> reports every slot that is not yet best-in-slot, with the upgrade path where one exists -- the same route the character page shows. Add a slot to see just that one; the slot is picked from a list, so there is nothing to spell.
Like /audit it takes no region or realm.

Hover any suggested item, gem or enchant

6 August 2026

Suggested gems, enchants, best-in-slot targets and every step of an upgrade path are now Wowhead links, with the same hover tooltip the equipped items on your character sheet already had. They open in a new tab.
One name is deliberately left unlinked: "Band of Eternity" is a different item for hunters than for casters, and both go in the same finger slot -- so rather than guess, it stays plain text.

Discord: /audit no longer asks for region and realm

5 August 2026

The /audit command now takes just a character name. It always looks up EU-Thunderstrike, which is what everyone was typing anyway -- and a typo in either field used to come back as an unhelpful "no stored audit found".
/guild still takes a region and realm, since looking up a guild somewhere else is the point of that command.

Upgrade advice now shows a route, not a shopping list

4 August 2026

Instead of naming whatever item scores highest, each slot now shows the path to best-in-slot: the earlier-phase pieces worth picking up on the way, ending at the current phase's BiS as the goal.
A step is only shown when it actually beats what you already wear, so you will not be sent to fetch something worse than your current gear.
Every best-in-slot list, enchant and gem recommendation now comes from wowtbc.gg, for all 18 specs and every phase. Previously these came from a raid simulator's own presets, which disagreed with the community lists -- most visibly for protection paladins, who were being told to gem and enchant for survival rather than threat.
Rogues, warriors, hunters, mages and warlocks are now matched against their own build's list rather than one shared list per class.

Healers, feral druids and a false alarm

3 August 2026

Healing specs now have best-in-slot data. Restoration druid, holy paladin, restoration shaman and both healing priest builds were previously compared against nothing, because the underlying simulator does not model healers.
Healers are no longer told they are short of the spell hit cap. Healing spells cannot miss in TBC, so that warning was never real.
Feral druids are now audited as cat or bear based on the gear they wear, rather than both being judged against one list.