Scope
"Use random currency and hope" is a fair description of crafting without a plan. It is a poor description of crafting with one.
The difference is not access to better currency. It is that a planned craft fixes what is knowable before spending anything, then arranges the random steps so that each one is cheap to retry and none of them destroys the progress of the last.
1 · The determinism inventory
Before planning, know what can actually be guaranteed. There is no operation that adds an arbitrary chosen affix, but there are six distinct kinds of certainty available.
| Kind | Mechanism | What it fixes |
|---|---|---|
| Guaranteed modifier | Essences; abyss bones | One specific modifier, chosen from a fixed list. Abyss bones draw from a pool standard currency cannot reach |
| Off-budget modifier | Runes, soul cores | A chosen modifier that consumes no prefix or suffix |
| Permanent lock | Fracturing Orb | Makes one modifier unremovable and unrerollable — but which one is random. See 1.1 |
| Which half | Sinistral / Dextral omens |
Restricts an operation to prefixes only, or suffixes only |
| Which type | Homogenising omens; catalysts with Catalysing Exaltation |
Adds a modifier of the same type as one already present, or weights the pool toward a type |
| Which one leaves | Whittling omen |
Removes the lowest level modifier rather than a random one |
Sinistral Annulment together with a Greater Annulment removes two prefix modifiers. Combining a positional omen with a quantity omen is how narrow tools become precise ones.
1.1 Fracturing is a late gamble, not a starting anchor
A Fracturing Orb locks one modifier permanently: it cannot be removed, cannot be changed, and its numeric value cannot be rerolled. That is the strongest form of permanence in the game — and it is worth being precise about what it is not.
| Property | Consequence |
|---|---|
| The locked modifier is random | You cannot choose which one, and no omen targets it |
| Requires four or more modifiers | It cannot be applied to a partly-built item |
| Not usable on uniques or already-fractured items | One fracture per item, ever |
| Drop-restricted endgame content | Not a tool available during the campaign |
So fracturing is not a foundation you build on. It is a gamble taken on an item you have already built, at one-in-four odds of locking the modifier you actually wanted.
The eligibility trick
There is one way to improve those odds, and it is a good example of reasoning from a rule rather than from a number.
Desecrated modifiers cannot be fractured — but they still count toward the four-modifier minimum. A desecrated modifier therefore satisfies the requirement while being ineligible for the roll:
four modifiers, one desecrated → your target is 1 in 3
A third more likely to succeed, obtained entirely from understanding which modifiers are eligible. No additional currency, no extra attempts — just the rule, read carefully.
This is the shape of most real crafting improvements. Not a better orb, but a constraint understood well enough to be exploited.
2 · A craft is a flowchart
The single most useful reframing, and the organising idea behind community craft-guide tools: a craft is not a list of steps. It is a decision tree, where every random step has at least two exits.
step: apply currency X
hoping: modifier Y appears
if yes → advance to the next step
if no → repeat this step # only if repeating is cheap
if ruined → abandon, or fall back
Writing the branches down before starting is what converts hoping into planning. A step whose failure branch is "repeat, it costs almost nothing" is a good step. A step whose failure branch is "the item is now worse and I cannot undo it" needs to be either last or avoided.
2.1 Track the affix state at every step
Three prefixes, three suffixes. Every step changes the count, and the count determines what remains possible.
This is where plans most often break: a step adds a modifier to a half that was nearly full, and a later step that needed room in that half becomes impossible. The item is not ruined — it is simply finished earlier and worse than intended.
Before each step, ask "how many prefixes and how many suffixes will this item have afterwards?" If the answer closes off a later step, reorder.
3 · Planning, before anything is spent
- Name the target. Which modifiers, and in which half. Four wanted modifiers split three-and-one across the halves is a different craft from two-and-two.
- Choose the base and check its item level. This sets the ceiling and cannot be changed later. Spending good currency on a low item level base buys a low maximum.
- Identify the anchor. Which of your targets can be guaranteed — by essence, by abyss bone, by starting from a fractured base? Anchor first, then build around it.
- Order by reversibility. Cheap and repeatable first. Irreversible last. Corruption, if at all, only when the item is otherwise finished.
- Set the budget and the bail-out. How many attempts, and what you accept if they run out. Decided now, not mid-craft.
Which step belongs earliest in a craft plan?
- Selecting the base and checking its item level
- Rerolling the numeric values within their ranges
- Corrupting the item for a chance at more
- Removing an unwanted modifier from the item
4 · A worked craft
A generic example, chosen because the affix bookkeeping is visible. The method is the content here — verify each currency's exact current behaviour before relying on it.
Boots wanting movement speed and life — both prefixes — plus two resistances, both suffixes. Four wanted modifiers, two in each half, leaving one spare slot on each side.
Step 1 — Base
Highest item level boots available of the correct armour type for the character's defence layer. Nothing later can raise this ceiling. Prefixes 0, suffixes 0.
Step 2 — Build through magic, not straight to rare
Upgrade to magic and augment to two modifiers. A magic item holds one prefix and one suffix, so this is a much smaller space than a rare, and rerolling it is cheap.
Fish here for the hardest-to-get target. Movement speed has essentially one home in the game, so it is the constrained modifier and belongs in the cheap phase.
if movement speed present → exit loop
else → repeat # cheap; this is the budgeted loop
This is the step to spend a stated number of attempts on. It is also the step people skip — and skipping it is what makes crafting feel like gambling, because upgrading a blank base straight to rare resolves several modifiers at once with no control over any of them.
Step 3 — Upgrade with the anchor
Use an essence to upgrade the magic item to rare. The essence guarantees its specific modifier, so you arrive at rare with two known-good modifiers rather than a full set of unknowns.
This is the pivotal move. Compare the two routes to the same rarity:
| Route | Known-good on arrival |
|---|---|
| Blank base upgraded straight to rare | 0 |
| Magic fished, then essence-upgraded | 2 |
Step 4 — Count
Stop and take stock. Which half holds what, and how much room is left in each? Every remaining option depends on this number, and this is the step most often skipped.
Step 5 — Fill the empty half deliberately
Now the omens earn their cost. Add to the item, choosing:
- Which half — a positional omen restricts the addition to prefixes or to suffixes, so you cannot accidentally fill the side you still need.
- Which type — a homogenising omen adds a modifier of the same type as one already present, which is how you compound toward a theme rather than drawing blind.
Step 6 — Remove, if something unwanted landed
If a half is full and one of its occupants is filler, removal is a real operation rather than an admission of failure. A positional omen removes from the chosen half; Whittling removes specifically the lowest level modifier.
Note the branch here: removal is random within its constraint. Removing from a half holding one good and one bad modifier is a coin flip, and the failure branch is that you removed the good one. That is a step whose branches must be written down before it is taken.
Step 7 — Socket a rune
A guaranteed modifier that costs no affix. It draws on a different budget entirely, so it is available even when both halves are full.
Step 8 — Tune the numbers, then stop
Rerolling numeric values comes last, once the modifiers are settled — doing it earlier tunes values you are about to replace.
Then stop. Corruption is irreversible and locks the item against all further crafting. It is not a step in a plan so much as a decision to end the plan.
5 · How plans fail
| Failure | Cause |
|---|---|
| The craft felt like pure gambling | Upgraded a blank base straight to rare, resolving everything at once |
| Ran out of room in the half that mattered | Affix state not tracked before an addition |
| Good modifier removed instead of the filler | Removal taken without a positional or targeted omen |
| Currency spent, ceiling too low | Base item level not checked at step 1 |
| Value rerolls wasted | Numbers tuned before the modifiers were settled |
| Item locked before it was finished | Corrupted while steps remained |
| Kept going past the point of value | No budget or bail-out set beforehand |
Every one of these is a planning failure, not a luck failure. The randomness was never the problem.
What to carry out of this
- Six kinds of certainty exist: guaranteed modifier, off-budget modifier, permanent lock, which half, which type, which one leaves. None of them adds an arbitrary chosen affix.
- Fracturing locks a random modifier, needs four modifiers present, and is endgame-gated — a late gamble rather than a foundation. Including a desecrated modifier improves the odds from 1 in 4 to 1 in 3, because desecrated modifiers count toward the minimum but cannot be fractured.
- A craft is a flowchart. Every random step needs its success and failure branches written down before it is taken.
- Build through magic first. Two modifiers is a small space and rerolling it is cheap. Fish there for the constrained modifier.
- Upgrade with an anchor, so you arrive at rare with known-good modifiers rather than a full set of unknowns.
- Count prefixes and suffixes before every step. This is where plans quietly break.
- Irreversible last. Values after modifiers, corruption after everything, or not at all.
- Set the budget and the bail-out before the first attempt.