The genre is unfalsifiable. This version does not have to be.
Every “mistakes to avoid” page and video for this game is somebody’s recollection of their own run, and there is no way to check any of it. That is not a complaint about the format — a creator forty hours in knows things nobody else does — but it does mean the advice arrives with no way to tell a measured claim from a remembered one. And in this game the difference is load-bearing: the single most-recommended early skill is the subject of an unresolved four-way disagreement about whether it is safe against bosses, and a list that simply says “take Helm Splitter” is passing on one side of an argument as if it were settled. So every row below names who established it and links to the page on this site that holds the evidence. A row that cannot do both is not on the list — the data file will not accept it.
The ledger, filtered by when it bites
Sorted the way a run runs rather than by severity, because a reader is somewhere specific and wants what applies now. The cost column separates the three that are expensive from the seven that are merely annoying, and the standing badge separates what was read off the game’s own text from what sources agree on from what they are still arguing about. Two of the rows here contradict advice that appears on most other sites for this game.
10 errors · 2 of them contested by the sources themselves
Filtered by when it bites rather than ranked by severity — you are somewhere in a run, not choosing which error to care about most. Every row names who established it and links to the page holding the evidence, because a mistake nobody can check is the format this page exists as an alternative to.
Read off the clientFirst hourExpensive
Holding block when the entanglement gauge is empty
The gauge is also the block reserve. Guard an attack without enough of it and you are staggered and completely defenceless for a few seconds — which is worse than taking the hit you were trying to avoid.
Watch the bar rather than the health. If it is low, take the hit or move; a parry refunds gauge, a failed block spends what is left of it.
IGN, from play, and the tooltip figures behind the gauge were read off the client.
The gauge, what drains it, and the eight stones that put it back→
Read off the clientFirst hourExpensive
Taking the advice to “prioritise Entanglement Overdrive” as a single pick
It is one of Emma’s three branch roots, not a node. Committing to it commits roughly a third of her tree, and it gates on Strength 9 rather than on the Florescence the colour scheme suggests. Iaido is inside the branch, not beside it.
Decide the branch, not the node. The two published routes disagree about whether to go up the tree toward the gauge or down it toward a different branch, and neither has tested the other’s order.
Read off the node’s own tooltip and Emma’s attribute panel in a capture we hold.
EstablishedFirst hourRecoverable
Copying a creator’s attribute spread without copying their difficulty
The one specific published recommendation — points into Vitality — is scoped by its own author to hard difficulty. On normal or easy he says to balance attributes or spend them elsewhere, and hands the decision back.
Take the difficulty with the spread or take neither. Nothing published establishes what any attribute does per point, or whether thresholds gate equipment.
RPG Division, who states the scope on camera.
Attributes: the one explicit recommendation, and its condition→
Published advice conflictsMid-runRecoverable
Building around Helm Splitter and staying airborne against bosses
Three guides recommend it. The one source who had finished the game says enemies track you in the air effectively, their overheads reach you up there, and Emma can neither block nor parry while airborne — so it is “much safer” to stay grounded against bosses.
Read the scoping before copying the build. The safety claim is made about normal enemies and groups; the warning is made about bosses. Both can be true, and nobody has filmed the ten-second test that would settle it.
RPG Division, Arekkz and IGN for; Skill Up against, reviewing after completing the game.
Published advice conflictsMid-runRecoverable
Choosing a sword by its attack figure
The figure separates nothing. Every sword Fextralife lists reads 51 or 45; every sword Game8 lists reads 27 or 21. The column a tier list would sort on holds two values on either site, and neither site sorts by it.
Choose on the damage type and the parry effect, which are the fields that actually vary — and keep a second sword upgraded, because swords share their stats and combo string and differ mainly in what they are good against.
Fextralife against Game8, read against one client frame showing a third value between them.
Why the two wikis disagree by a factor of two, and what to pick on instead→
EstablishedMid-runRecoverable
Letting the hunger meter empty on a long exploration run
An empty meter costs healing, not health — it stops the passive recovery that runs between fights. Chip damage from trash stops healing back and you arrive at the next fight already down.
Feed both of them; Emma and Koo carry separate meters and his treats are a separate item. If you are dying to a boss, hunger is not what is wrong — if you are arriving at bosses at 60%, it probably is.
NerdsChalk is the only source that pins the out-of-combat scoping down.
What an empty meter actually costs, and the claim only one guide makes→
EstablishedMid-runRecoverable
Wearing the heaviest armour you have found
Defence climbs monotonically across the set and the other two published axes do not. Heavier armour raises how easily enemies see you, and armour also decides how many Spirit Stones you can equip — so on two of three axes the best piece is not the last one you found.
Pick against the axis you are short of. There is a set with non-existent enemy awareness for the stretches where being seen is the problem.
IGN for the awareness and slot mechanics; the defence ladder is Fextralife’s table.
Where the best piece is not the last one, on two of three axes→
EstablishedMid-runRecoverable
Running a “levelling” farm to raise your character level
The loops everyone posts drop skill points. The campfire menu keeps “Level Up Attributes” and “Acquire Skills” as separate entries and the HUD counts them separately; no source we hold measures what those kills do to the experience bar.
Run them when you want skill nodes. If you are grinding toward an attribute threshold, nothing published says these are the fastest route.
Two creators independently describe one skill point per kill; neither mentions the experience bar.
EstablishedMid-runExpensive
Not going back to the Cleanse Walker between zones
New conversations open there constantly, and some of them pay — three attribute points in one case, a large amber payout in another. Enhancement is also gated there: past a point you cannot upgrade further until you return generator devices.
Treat a return trip as part of the loop rather than as a detour, and talk to everyone each time.
IGN, describing the walker’s functions before playing.
The enhancement tracks, and the walker power that gates them→
EstablishedEndgamePermanent
Activating Overdrive inside the arena encounter that is known to break
Doing it there has been reported to soft-lock the fight. This is a bug rather than a mechanic, and it is the one documented way the state fails outright.
Check the tracker before assuming a stuck encounter is your build.
Player reports, tracked on this site rather than mentioned in passing.
The issue tracker, with what is official and what is community-reported→
The two the sources contradict each other on
Worth pulling out, because they are the rows a confident list would have got wrong in one direction or the other. The airborne question is the big one: three guides build around getting Emma into the air, and the only source who had finished the game when he spoke says enemies track you up there, their overheads reach you, and she can neither block nor parry while airborne. Read carefully those positions may not even conflict — the safety claim is scoped to normal enemies and groups, the warning to bosses — which is a distinction no list carries. Underneath it sits one mechanical question that would settle everything, and it is a ten-second clip nobody has published. The second is quieter and easier to check: the two wikis publish attack figures for the same swords that differ by roughly a factor of two, so whichever one you read, the other contradicts it. A third row is not marked contested but has the same shape in miniature — one creator makes one specific attribute recommendation and scopes it to hard difficulty in the next sentence, and every page that reprints the spread drops the scope. That is not sources disagreeing; it is one source being quoted badly, which the grade deliberately does not conflate with the other two.
What is not here, and the page it belongs to
Nothing on this page is about content you can permanently lose. That is a different question with a different answer, and it has its own page — the checklist of things that become unavailable, including the one blocker with evidence behind it that the three big portals all omit. The single row here marked permanent is a bug rather than a choice, and it points at the issue tracker instead. Keeping the two apart is deliberate: two of our own pages bidding for one query is a defect this site has now paid for five times, and it is the reason a check exists that reads Search Console for exactly that shape.