Genres and story forms

How to build a magic system that doesn’t break your story

August 16, 2026 · The StoryProp team

Most magic systems do not fail on the drawing board. They fail around the middle of the book, when the writer needs the hero in real trouble and the magic, as established, could simply solve it. So the magic forgets itself for a chapter. The reader notices. And the trust that lets a fantasy novel work — the sense that this invented world runs on something — quietly drains out of the story. The rule that prevents this has a name and an author: Sanderson's First Law of Magics, which Brandon Sanderson set out in 2007 in the first of his published Laws of Magic essays. Magic may solve problems only in proportion to how well the reader understands it.

The instinct at that point is to build harder: more lore, more maps, a codex of runes. But a magic system is not a physics paper, and the case against front-loading an invented world — the one made in worldbuilding without the wiki trap — applies doubly to magic. Magic touches plot mechanics directly. It has to be designed like a dramatic engine: around what it cannot do, what it costs, and what it refuses at any price.

Hard vs soft magic: two kinds of promise

Hard magic makes its rules explicit. The reader knows what the magic can do, what it consumes, and where it stops — which means the writer is allowed to resolve plot with it. That is the deal. A solution feels earned when the reader held the pieces in advance; hard magic is a locked-room mystery played fair, satisfying because the answer was available all along.

Allomancy, the system Sanderson built for Mistborn in 2006, is the worked example. Allomancers swallow metals and burn them, one effect per metal: steel Pushes on nearby metal, iron Pulls it, tin sharpens the senses, pewter strengthens the body. The costs are physical and countable — you can only burn what you swallowed, so a long night's fighting is also an inventory problem. The limits are just as concrete: a Push touches metal and nothing else, and it obeys weight in both directions, so Pushing a coin sends the coin flying while Pushing a wall sends you. The reader watches those rules operate for hundreds of pages, which is precisely why the climax of the first book is allowed to turn on a piece of metal the reader was shown early and invited to wonder about.

Soft magic withholds the rules, and Tolkien is the standing demonstration. Gandalf never gets a spec sheet — we are told almost nothing about what a wizard can do, only shown what he chooses to do — and The Lord of the Rings keeps that power away from the load-bearing moments. His stand on the bridge of Khazad-dûm raises the stakes and costs him his life; it does not solve the Fellowship's problem. The Ring is not unmade by wizardry. It is walked to the mountain by two exhausted hobbits and lost there in a scramble. What withholding buys is wonder and dread — the sense that the world is larger than anyone's understanding of it, which no rule sheet can produce. Mystery is not a failure of worldbuilding; it is a design choice with its own legitimate payoff. Sanderson's law is what keeps it honest. Soft magic can create trouble, raise stakes, and inspire awe — an unexplained storm is a legitimate storm. But let it cheaply rescue the hero and the reader feels the author's hand on the scales. Hard magic earns rescues. Soft magic earns weather.

In practice, durable systems tend to be hard at the center and soft at the edges. Whatever your protagonist personally practices should run on knowable rules, because the reader will be asked to follow decisions made with it. Past that circle — the gods, the deep sea, whatever the enemy dug out of the ice — the rules can fade into rumor. Precision where choices happen; mystery where scale lives.

The price of power

A cost is what turns a spell from a button into a decision. Suppose casting burns a memory, and the caster does not get to choose which one. Now every working of magic is a bet placed against the self: the war might be won by a woman who no longer remembers why she wanted to win it. Nobody presses that magic like a light switch. They weigh it — and weighing is drama.

Costs come in more flavors than fatigue. There is fuel, which runs out at the worst possible moment — the Allomancer's vials are exactly this. There is toll: pain, years, the memory-burn above. There is social cost: magic that marks the skin, breaks a law, or opens a debt to something patient. The strongest costs are visible on the page and cumulative across the book, so the reader can watch the bill grow.

Limits are not costs, and a system needs both. A cost is what the magic takes in exchange for what it can do; a limit is what it cannot do at any price. The dead stay dead. Gold can be moved but never made. Nothing scries across running water. Limits carve out the territory where courage, cleverness, and sacrifice still decide events — which is to say, the territory where your actual story happens.

Rules are plot fuel, not fences

Omnipotence kills stories, and not because it is unrealistic — everything in fantasy is unrealistic. It is that a character who can do anything cannot credibly want anything, and want is the engine of narrative. Constraint does the opposite work. Every rule is a wall, and characters pressed against walls produce plot. A magic system is best understood not as decoration but as a machine for generating predicaments.

Take a healer who cannot erase wounds, only move them into a willing body. The rule is one sentence long, and it detonates on contact with people: who volunteers, who gets asked, who is not told the whole truth about what they are agreeing to carry. A mother answers differently than a soldier; a king answers differently than a stranger. None of that was plotted. The constraint produced it.

So test every rule by the temptation it creates. What would a desperate person do at its edge? Whatever answer makes you uneasy, someone in your world has already tried it — and that someone may be your villain. Villains, in a rule-governed world, are mostly people who read the rules more ruthlessly than everyone else.

Exceptions are where systems live or die. An exception the reader can trace back to a planted seed is a twist; an exception invented on the spot to rescue a cornered plot is a hole. If book three needs the impossible to happen once, the honest move is to go back and seed the crack early — a rumor, an anomaly, a rule that was always phrased a little too carefully.

How to create a magic system while you draft

You do not need a finished system before chapter one. You need a minimum viable version: one cost, one limit, one visible consequence, all demonstrated in an early scene. That is enough to draft against. The rest can be discovered the way your reader will discover it — by watching the magic under pressure.

The discipline that makes this safe: decide reactively, record permanently. When a scene asks a question the system has not answered — can wards overlap, does the burn scale with distance — answer it in the scene however the drama demands, then write the answer down as a rule you are now bound by. And before you commit, ask the one question that matters: am I willing to be held to this for the whole saga? Unstated is revisable. Stated is contract.

Keeping magic consistent across a saga

Long fantasy has a specific decay mode: escalation inflation. Each book the set pieces grow, and each book the costs quietly shrink, because the writer needs bigger magic and the old prices have become inconvenient. The most visible case in modern fantasy is the Time-Turner. Introduced in Prisoner of Azkaban as a device that lets a schoolgirl double-book her timetable and then save a condemned hippogriff and an innocent man, it left every later emergency shadowed by a question the series could not keep answering: why not simply go back? Two books later, in Order of the Phoenix, the Ministry's entire stock is smashed on the page and written out of the world. Retiring a power in front of the reader is the honest fix, and it is still expensive — far cheaper never to mint the power at all.

Readers of a saga have long, collective memories, and there can be an enormous number of them. Rebecca Yarros's Empyrean series had sold close to 6.5 million copies before Onyx Storm was published, and Onyx Storm then moved about 1.3 million US print copies in its first week on sale — the fastest start for an adult fiction title in Circana BookScan's twenty years of tracking. Fantasy at large is a category of that order: 24.1 million print units sold in the US in 2025, on Circana's count. A limit stated in book one binds book three, and the moment the binding slips, the magic stops being a force of the world and becomes a convenience of the author.

The working answer is a ledger: every rule, where it was stated, every exception, and where it was seeded. Check it before drafting a magic scene; update it after. It is the same discipline that keeps a long story consistent in any genre. When you find yourself wanting to break a rule, there are three honest options — pay the stated price, go back and seed a revision earlier in the timeline, or want something else.

This is bookkeeping, and it is the part a drafting tool should carry — which is exactly what StoryProp does as an AI writing tool for fantasy novels. A project's settled decisions are held as records — canon, characters, confirmed rules — and the agent reads them before it writes, so the limit you stated two books ago is still in force when tonight's scene gets drafted. When a new passage contradicts the record, the contradiction is surfaced for you to rule on, not silently corrected. What you agree to gets written down; what you don't stays a suggestion.

The paradox at the heart of the form is that the more your magic cannot do, the more your story can. Rules are not fences around wonder. They are what make an act of magic cost something, and cost is what makes it matter. Build a system that says no often enough, and the one time it says yes will land like thunder.

A magic system is a promise kept over hundreds of pages, and promises need a written record. StoryProp holds your rules, costs, and exceptions as canon the agent consults before drafting, so the price you set in chapter two still binds in book three.

← All posts