Scope
A theme is a starting point: a skill that looks interesting, a damage type with an appealing feel, a playstyle worth trying. It is not a plan, and the gap between the two is where most self-built characters fail — not through a wrong decision, but through a decision made without noticing it was one.
A skeleton is the small set of choices that constrain everything else. Get them consistent and the remaining hundreds of choices are cheap and correctable. Get them inconsistent and no amount of good gear repairs it.
This lesson is about identifying which choices those are. It does not contain a character, and it will not tell you what to pick.
1 · Sort by reversibility, not importance
The instinct is to make the most important decisions first. The more useful ordering is by how expensive each is to undo, because a decision you can revisit costs nothing to get wrong.
| Decision | Cost to change |
|---|---|
| Class | Not reversible — requires a new character |
| Ascendancy | Expensive and content-gated: requires a Trial item of sufficient level and enough gold |
| Passive allocation | Gold, with the cost per point rising as you level |
| Gear bases | Whatever it costs to find or craft a replacement — in SSF, potentially a long time |
| Skill and support gems | Effectively free to swap |
Respec mechanics per Mobalytics: passives refund for gold at the Hooded One or at Doryani in the hideout, with cost scaling by level. Ascendancy respec additionally requires a Djinn Barya or Inscribed Ultimatum of appropriate area level.
Decide the irreversible things deliberately. Decide everything else by trying it.
This inverts the usual anxiety. Support gems feel like they matter enormously, and they do — but they are free to change, so agonising over them early is wasted effort. The ascendancy you picked in twenty seconds because the icon looked good is the one that will still be with you at level 90.
1.1 Reversibility is not fixed
Two mechanics soften choices that would otherwise be hard commitments.
Weapon set specialisation. The passive tree supports allocating points to a specific weapon set rather than globally, and skills can be configured to swap to a particular set when used. A character can therefore run two weapon-based configurations rather than being locked to one.
Note the friction: refunding weapon-set points requires removing one from each set to reclaim a single normal point (Maxroll). Cheaper than rerolling, not free.
Gold. Passive respec cost scales with level, which means the same change is cheap at level 30 and painful at level 90. The window for cheap experimentation is early and it closes gradually. This argues for testing uncertain ideas well before the character is finished, not after.
2 · The load-bearing decisions
Four choices constrain most of what follows. Each is load-bearing because reversing it invalidates work done elsewhere, not merely because it is impactful.
2.1 The base source
Attack, spell, or minion — the question from lesson 3. This is the most constraining decision of the four because it determines which half of your gear can contribute at all.
A character built around weapon damage that switches to minions does not lose some of its investment. It loses the weapon, the flat added damage on rings and gloves, the attack speed, the accuracy, and the tree regions that fed all of it.
2.2 The weapon type
For attack builds, weapon type gates which skills are usable at all — many skills state a required weapon — and the tree carries weapon-specific clusters. For spell and minion builds this decision is much lighter, since the weapon is a stat stick rather than a damage source.
Weapon set specialisation makes this the softest of the four, but the softening is not free.
2.3 The primary damage type
Damage type determines which additive total your investment feeds, which tree regions pay, which gear affixes are relevant, and which enemy-side tools apply — penetration and exposure are per-type.
Note what changing it does and does not cost. Your base often survives a damage-type change, since conversion and gain can redirect it. Your scaling does not: an additive total built for one type contributes nothing to another.
2.4 The defence layer
Life, energy shield, armour, evasion, block — the choice determines which gear bases you need, which tree region you path through, and which recovery mechanisms are available. It is expensive to change because it is expressed almost entirely through gear bases and tree geography, both of which take time to replace.
Changed at level 70, which of these invalidates the most other decisions?
- The primary damage type the character scales
- The support gems socketed into the skill
- The specific pair of boots being worn
- The flask setup carried into each map
3 · The decisions are coupled
The four are not independent picks from four menus. Each constrains the others, which is why a skeleton is a consistent set rather than a list.
| Choosing this… | …constrains this |
|---|---|
| A base source | Which weapon matters, and whether weapon type is a real decision at all |
| A weapon type | Attribute requirements, which pull you toward one region of the tree |
| A tree region | Which defensive clusters are nearby, and therefore cheap to reach |
| A defence layer | Which gear bases you need, and which attributes those bases require |
The loop closes on attributes and tree geography. A choice requiring one attribute puts you in one part of the tree; the defences cheaply reachable from there require compatible gear bases; those bases require the same attribute. Consistent skeletons are cheap to build because everything reinforces everything else. Inconsistent ones spend their passive points travelling.
An inconsistent skeleton rarely announces itself as a mistake. It presents as a character that is simply behind — slightly short on damage, slightly short on life, always one item away. The cause is that a large fraction of the passive points went into reaching things rather than into being things.
4 · What to leave open
Deciding early is not a virtue. Every commitment made before it is forced narrows the space of things you can discover, and discovery is the point of building your own character.
| Leave open | Why |
|---|---|
| Support gems | Free to change; the correct set depends on numbers you do not have yet |
| The specific skill | Within one archetype and damage type, skills are largely interchangeable at the skeleton level |
| Gear affixes | Determined by what drops; planning them in SSF is planning around luck |
| Ailment and enemy-side strategy | Depends on what your damage type makes available, which the skeleton already decided |
| Precise passive routing | Only the destination regions need to be decided early |
The useful question about any decision is not "what does this make good?" but "what does this make impossible?"
A choice that closes nothing is not a skeleton decision, however large its numbers. A choice that closes many doors belongs in the skeleton even if its immediate effect is small.
5 · The order of operations
The procedure, given a theme:
- State the theme in one sentence. A skill, a damage type, a playstyle — whatever the actual attraction is.
- Derive the base source. Read the tags. This is forced by the theme, not chosen freely.
- List what the base source already forced. Often more than expected — an attack theme has already half-decided the weapon.
- Choose the remaining load-bearing decisions consistently with each other, checking attributes and tree geography as you go.
- Stop. Everything else is discovered by playing.
Step three is the one most often skipped, and it is where the coupling from section 3 becomes visible. A theme frequently determines more than it appears to, and noticing that early is free.
Worked diagnostic
A player states their plan for a new character:
Two questions. Which decisions were made in the wrong order? And which single undecided item is the most urgent?
Wrong order: support gems and the chase amulet were decided early despite being the two cheapest things on the list to change. Supports swap for free; a specific rare in SSF may never drop, and planning around it is planning around luck. Both are effort spent where reversibility is highest.
The ascendancy was decided correctly — it is the most expensive item on the list to reverse, so deciding it deliberately and early is right, whatever it was chosen for.
Most urgent undecided item: the defence layer. It is expensive to change because it lives in gear bases and tree geography, and it couples to attributes, which in turn shape the whole tree route. Leaving it open means every passive point allocated in the meantime is provisional.
The weapon is not urgent — and noticing why is the point of step three. The theme already answered it. A minion base source means the weapon is a stat stick, so weapon type is not a load-bearing decision for this character at all. It looked like an open question and was actually closed the moment "minions" was said.
One thing nobody listed: "fire themed" is a damage-type commitment, and for a minion character it needs checking rather than assuming. As lesson 3 established, minion damage scales only from minion-scoped modifiers — so "fire themed" only means something if fire scaling can be routed to the minions. That is a question to answer before the tree is built, not after.
What to carry out of this
- Sort decisions by reversibility, not importance. Cheap decisions can be made by trying; expensive ones must be made deliberately.
- Four decisions are usually load-bearing: base source, weapon type, primary damage type, defence layer.
- They are coupled. A skeleton is a consistent set, and consistency is what makes it cheap to build.
- Ask "what does this make impossible?" rather than "what does this make good?"
- A theme forces more decisions than it appears to. Listing what is already closed is free and is usually skipped.
- The window for cheap experimentation closes gradually as respec costs rise. Test uncertain ideas early.
The procedure applied to a theme of your own: state it in one sentence, derive the base source from the tags, then list everything that choice has already decided. The list is usually longer than expected, and the items on it are no longer open questions.
What remains after that subtraction is the actual decision set.