Scope
There are a great many currency items and they change between patches. Memorising them is not the skill.
The skill is recognising that they fall into a small number of operations, knowing which operation a given situation calls for, and being able to decide in advance how many attempts a plan is worth. That structure survives patches; the item names do not.
1 · The operations
Nearly every crafting currency performs one of these.
| Operation | Effect | Example |
|---|---|---|
| Upgrade rarity | Change an item's rarity, adding modifiers as it goes | Normal → magic; magic → rare; normal → rare directly |
| Augment | Add one random modifier, up to the rarity's limit | Adding a modifier to a magic or rare item |
| Remove | Delete a random existing modifier | Annulment |
| Remove and add | Both at once — a targeted reroll of one slot | Chaos |
| Reroll values | Re-randomise the numbers within their existing ranges, keeping the modifiers | Divine |
| Quality, sockets, enchantment | Alter properties outside the explicit modifiers | Whetstones, socket currency, runes |
| Corrupt | An irreversible gamble that also locks the item against further crafting | Vaal |
The limits from lesson 10 govern all of it: a magic item holds two modifiers, one of each half; a rare holds six, three of each. Adding is impossible once the relevant half is full, which is why removal is a real operation rather than a mistake.
2 · Determinism is a separate layer
The operations above are random. A second category of item constrains that randomness — omens, which modify how the next use of a specific currency behaves.
The naming is systematic and worth learning once:
| Term | Meaning |
|---|---|
| Sinistral | Acts on prefixes only |
| Dextral | Acts on suffixes only |
| Greater | Acts twice instead of once |
| Whittling | Acts on the lowest-level modifier rather than a random one |
Combined with a base operation, this produces targeting: removing only a suffix, adding only a prefix, removing specifically the weakest modifier. That converts a coin flip into something closer to a decision.
Random currency is cheap and plentiful; targeted currency is scarce. Spend randomness where the item is worthless and determinism where it is nearly finished.
Two more scaling dimensions exist: currency comes in greater and perfect variants that guarantee a minimum modifier level, and item level caps which tiers can appear at all. A high-tier outcome requires both a high item level base and, often, the better currency variant.
3 · Attempts, not outcomes
The productive question is never "will this work". It is "how many attempts am I willing to spend, and what happens when they run out".
Each attempt is independent — nothing accumulates toward a guaranteed result. So the plan must be stated as a budget and a stopping rule, decided before the first attempt rather than during.
Computing a genuine expected number of attempts requires the modifier pool and its weights for that base at that item level. That data is not reliably published for PoE2. Anyone quoting precise attempt counts is either using datamined tables or guessing.
What follows is therefore relative reasoning — which approach is better than which — rather than absolute numbers. That is enough for every decision this course needs.
3.1 The relative rules
- Fewer open slots means better odds per attempt. Adding a modifier to an item with one open prefix draws from a smaller pool than one with three.
- Preserving progress beats rerolling it. An operation that changes one modifier and keeps the rest is worth far more than one that rerolls everything, even at a worse rate.
- Compound requirements are much rarer than they feel. Wanting two specific modifiers is not twice as hard as wanting one; it is closer to the product of two unlikely things.
- Irreversible operations go last. Corruption locks the item. Anything you might still want to do must happen first.
4 · Two economics that are specific to self-found play
4.1 Crafting on a base you would otherwise vendor is free
If an item's alternative use is being discarded, any outcome is an improvement and the only real cost is the currency.
This inverts the usual anxiety. The expensive mistake is not a failed craft on a worthless base — that has no downside at all. It is a failed craft on an item that was already good, where you destroyed something usable.
Practise on trash. The skill of crafting is built on bases where failure costs nothing, and the currency spent there is not wasted — it is what makes the later attempts competent.
4.2 Unspent currency is worth nothing
In a trade context currency has a market value, so hoarding it is a real position. Self-found, currency cannot be exchanged for anything. Its only possible use is attempts.
A stash of unused currency is therefore not savings. It is attempts you declined to take, on a character that would have benefited from them.
The correct posture is to spend up to the point where you would rather have the currency for a specific known future use. Absent such a use, holding it is a strictly worse option than trying something.
Self-found, which situation carries the largest real risk?
- Crafting further on an item already worth wearing
- Crafting on a base you would otherwise discard
- Spending a lot of currency on failed attempts
- Holding currency until a better opportunity
5 · When to stop
Improvement on a single slot has the same shape as everything else in this course: each addition lands on a growing total, so the returns flatten.
Stop when the next improvement to this slot is worth less than the first improvement to a different one.
That comparison is the whole rule, and it is the audit loop applied to crafting. A slot with four good modifiers is competing against a slot with none, and the empty slot almost always wins — for exactly the reason an empty additive total does.
Three concrete stopping signals:
- The next step requires an irreversible operation. Stop and consider whether the item is already good enough.
- The remaining gain is a better roll of something you already have. That is a small multiplier on one term; another slot may offer a whole missing term.
- You have run out of the half you need. A full suffix bar means further work must displace something — a lesson 10 problem, not a crafting one.
Worked diagnostic
Three items in the stash and a modest amount of currency. Resistances are capped; the weapon is the character's weakest slot by a wide margin.
- 25% increased Movement Speed
- +70 to maximum Life
- +40% to Fire Resistance
- +38% to Cold Resistance
- two open prefixes
- no modifiers — a white base
Where should the currency go, and what is the reasoning for each option?
The boots (A) are the tempting option and the weaker one. They are already good, they have open prefixes, and adding to them feels productive. But every addition lands on a character whose boots are not the problem — and any operation that can remove risks the resistances holding the cap. Real downside, small upside.
The white weapon base (B) is the answer, for three separate reasons.
- The weapon is the starved slot. For an attack character the weapon is layer one — the base every multiplier acts on — so improving it multiplies everything. This is the audit loop's answer before crafting is even considered.
- Item level 82 is a high ceiling. It can roll tiers the level 68 boots never could. Currency spent on a low-item-level base buys a low maximum.
- It is free to fail on. A white base has no value to destroy. Every outcome is an improvement over an unused item, so the only cost is the currency — which, self-found, has no other use.
The general shape: spend randomness on high-ceiling bases that are currently worthless, in the slot the audit loop already identified. Save targeted currency for the item that is nearly finished. The boots are nearly finished and should be left alone until something targeted is available.
What to carry out of this
- Currency reduces to a few operations — upgrade rarity, augment, remove, remove-and-add, reroll values, and irreversible. Names change; operations do not.
- Omens are a determinism layer. Sinistral means prefix, dextral means suffix, greater means twice.
- Randomness on worthless bases; determinism on nearly-finished ones.
- Plan in attempts and a stopping rule, decided before starting. Precise attempt counts are not publicly computable for PoE2 — reason relatively.
- Crafting on a base you would vendor is free. The only real risk is crafting on something already good.
- Self-found, unspent currency is worth nothing. It is attempts you declined to take.
- Stop when the next gain on this slot is smaller than the first gain on another one.