The Honest Gamble

Chaos Magic is the school this game built to break its own rules — the nuke that rolls its element, the teleport that picks its own destination, the spell that attacks the enemy's hand of glyphs, the mastery that charges you resistance for the privilege. It is meant to be the spoiler, the problem child, the school the other eight have to plan around. Thirty days of live casting data say the spoiler has spoiled: it is the least-cast school in the game at a third of one percent of all casts, its two signature combos were cast four times all month, and the one spell everyone does use is a self-buff the other schools borrow. This page argues the school's wildness was never the problem. The problem is that a school built to break rules on purpose has quietly broken a dozen by accident, and the accidents — not the dice — are what players walked away from. The dice stay. Every roll gets printed. Every accident gets fixed.

Executive summary

Every game that ships a chaos school faces the same bet: randomness is either the most memorable thing in the design or the reason nobody plays it, and the difference is never the randomness itself. The tabletop ancestor made the wild mage's surge table a public document players read for fun. The roguelike that perfected the chaos god made every gift and every prank legible, so the stories survived the deaths. The card games that built beloved chaos bombs made the dice a spectacle everyone at the table watches land. The competitive games went further and quietly bounded the streaks — random procs that cannot miss forever, crits that cannot chain forever — because they learned that true randomness feels rigged even when it is fair. And the deck games taught the counter-lesson from the other side: letting a player peek at, cut, or redraw their deck reads as mastery, not cheating. The genre's verdict is consistent everywhere we looked: chaos stays fun when the odds are public and the streaks are bounded, and no game that kept its chaos school healthy ever made it reliable.

Measured against that bar, our Chaos school rolls hidden dice. Its signature nuke secretly resolves one of five weighted outcomes, and roughly one cast in ten does no damage at all — a self-buff or a surprise teleport the tooltip has never mentioned. Its crowd-control spell inoculates its own victim: two castings of Chaotic Confusion make the target immune to confusion for the rest of the fight, which is documented nowhere and inverted in our own internal reference. Its signature utility spell, Juxtapose, does nothing whatsoever without a specialization that sits three prerequisites deep — a trainer will happily sell you a spell that is inert. And the school built to be uncounterable is accidentally uncounterable in the wrong way: no Chaos spell deals chaos-typed damage, so the chaos resistance stat has sat dead in the data for years, the anti-magic warding passive cannot answer the school at all, and the mastery's advertised downside — weakened resistance to the other schools — punishes the specialist without ever applying to the mirror match. The population has voted on all of this. Ninety-six of 1,251 active casters touched the school last month; its specialization ranks 20th of 21 trees; and the school with the most specialization content in the game — 57 distinct mastery hooks, more than any other tree — has the fewest specialists to enjoy it.

The gaps cluster in three places:

1. Dice nobody can read. The five-branch nuke, the blank casts, the secret moon-magic teleport, the confusion vaccine, the inert Juxtapose — none of it is printed on a tooltip, a buff bar, or a document. A school of gambles where nobody can learn the odds is not wild; it is opaque. Tier 1 prints every table, and the Surge Ledger (S31) makes the odds a living HUD surface.
2. A mastery that gates instead of amplifies. Half the school's real effects multiply by the mastery stat directly: zero until specialization, which sits behind an 80-point climb up a side branch. Fire works at level one; Chaos has a level-1-to-79 dead zone. The keystone (S13) converts every gate to a base-plus-amplifier curve — the "mastery amplifies, never gates" rule this school's internal review coined and other schools have since adopted.
3. An identity claimed but never collected. The fire page refused to touch deck manipulation because "Chaos owns the deck" — and Chaos owns one deck spell that only works in player-versus-player. The air and earth pages both deferred their transformation capstones to "the Chaos review's transformation grammar" — which was never published. Five ghost spells sit in the data with shipped descriptions and no implementation, one passive stands where sibling schools have five, and the school's only ward — against chaos itself — does not exist. Tiers 3 and 4 collect the debt: the deck verbs, the gambling verbs, the transformation, and the world content the fiction has been holding in reserve.

The deal that runs through all forty proposals: Tier 1 fixes what is broken or dishonest, in the open, with the defects filed as tracker bugs the day this page went up. Tier 2 un-gates the school and makes it answerable — chaos-typed damage arrives in the same patch as the ward that resists it. Tier 3 pays the deck debt and builds the gambling verbs the identity promises, on the ghost spell slots the data already reserves. Tier 4 is the capstone and the world: the transformation two other schools are waiting on, and the Obsidian fiction finally spent. And one refusal, stated once: nothing on this page makes Chaos reliable. No card removes a roll, narrows a spread to nothing, or converts a gamble into a rotation. The dice stay wild; they just stop being loaded against the people who chose them.

The school we have

Chaos ships ten castable spells, one innate, and three disabled rows waiting in the data. There is real design here — more deliberate, stranger design than almost any other school: a nuke that rolls its element so no single resistance answers it, a decoy clone that copies your face and armor, a daemon at the end of the tree, a spell that attacks the enemy's hand of glyphs, self-buffs that charge health and attunement as their price, and a mastery that openly costs you magic resistance. The school's grammar — power with a posted price — is already in the shipped data: the chaos enchant gem grants bolt damage and subtracts thirty durability in the same breath. What never shipped is the half of the design that makes a gamble playable: the odds, the payoffs, the passives, the counter, and the sound.

Confusion (root) ...................... [2]
├─ Chaos Bolt ....................... [1]
│  ├─ Tabula Rasa ................... [8]
│  │  └─ Summon Daemon .............. [6]
│  └─ Chaotic Step .................. [5]
├─ Juxtapose ........................ [3]
├─ Chaotic Clone ....................
│  └─ Chaotic Shield ................
└─ Torpid Torment ...................
   └─ Chaotic Feedback .............. [4]
      └─ Chaos Magic Specialization . [3]
[disabled: Control Demon, Chaotic Corruption,
 Obsidian Onslaught ................... [7]]

• 1The nuke with hidden dice. Chaos Bolt resolves one of five secret weighted branches: fire, cold, lightning, a damage-free self-buff (8.2% of casts), and a rare teleport-to-target that is literally a Moon school spell (1.6%). One cast in ten deals no damage; no tooltip, document, or buff icon says so. Cards S1, S21, S31.
• 2The self-defeating control. Confusion's hidden rider grants its victim stacking confusion resistance — two casts and the target is immune for the rest of the window. The school's own reference documents it as a debuff. Card S3.
• 3The inert signature. Juxtapose's four outcomes all require specialization to fire at all: untrained, the spell does nothing. The specialization sits behind an 80-point climb up a different branch. It is also the ingredient in nine of the game's eleven chaos-touching combos — the school's real engine, sold as a dud. Cards S13, S14.
• 4The one spell that escaped. Chaotic Feedback — faster casts and draws, paid for in discard health and life attunement — is 61% of the school's casts and is used by more players than any other chaos spell, mostly by casters who never touch the rest of the tree. The school's most successful export is its buff. Card S20.
• 5The step that shrinks. Chaotic Step's random teleport promises more distance with specialization; the data raises only the minimum, until the floor can exceed the ceiling. Card S5.
• 6The daemon's strange ledger. The summon levels twice as fast as every sibling summon in the game, multiplies its whole health pool by mastery where siblings multiply a bonus, moves at a crawl, and carries a fire attunement grant driven by a health stat — a paste error two columns wide. Cards S9, S33.
• 7The ghost school inside the school. Five unbuilt spells sit in the data with shipped player-facing descriptions — a chaos storm, a shield that "rarely amplifies" damage, a shapeshift into a random creature, and two anti-fizzle passives — plus a disabled second nuke that is a byte-level copy of the first with a borrowed icon. The school's future was designed, described, and never built. Cards S22, S23, S28, S29.
• 8The unclaimed crown. One passive where Fire has five; no ward, because Chaos has no opposite — and no ward against chaos exists either; chaos resistance is a stat nothing in the game can trigger; and the anti-magic warding passive explicitly resists every school except this one. The school nobody can counter is also the school with nothing to invest in. Cards S15–S17.

What the players do with it

Thirty days of live casting data, ending the day this page was written. The school's ten spells were cast 31,896 times by 96 distinct players — 0.38% of the 8.3 million skill casts recorded, and less than half the share of the next-smallest school measured. For scale: the Sun school's humble torch spell alone was cast more times than this entire school.

The combo data is starker. Juxtapose anchors nine cross-school combos; the only one players actually use is Evaporate, the fire-cleanse (2,659 casts by 9 players — more than eight of the school's own ten spells). The school's two pure-chaos combos, Chaos Bomb and Disorderly Displacement, were cast four times, combined, all month. The specialization census tells the same story over a longer window: 99 players have ever trained the specialization, 66 finished it, 32 of them were active in the last year — 20th of 21 trees, in the school with the most specialization content in the game.

How a chaos spell happens

Combat here is a synchronized simulation: for most rolls the server deals a seed and every client plays the identical result forward. Chaos Bolt's branch roll is the exception that defines the school — the casting player's machine rolls it locally, outside the seeded simulation, and then tells everyone which branch happened:

flowchart LR
  A["Chaos Bolt cast"] --> B{"Weighted roll
on the caster's machine
outside the seeded sim"}
  B -->|"weight 4 · 32.8%"| C["Fire bolt + burn"]
  B -->|"weight 3 · 24.6%"| D["Cold bolt + slow"]
  B -->|"weight 4 · 32.8%"| E["Lightning bolt + stun chance"]
  B -->|"weight 1 · 8.2%"| F["Self-buff only
zero damage"]
  B -->|"weight 0.2 · 1.6%"| G["Night Step
a Moon spell:
teleport to target"]
  C --> H["Chosen branch
replicated to all clients"]
  D --> H
  E --> H
  F --> H
  G --> H
  style B stroke:#7c4a9e,stroke-width:2.5px
  style F stroke:#a33d2e,stroke-width:2px
  style G stroke:#5e6fa3,stroke-width:2px

Lessons: wild magic from Ultima onward

Chaos magic is the genre's oldest deliberate malfunction. Four decades of games have shipped a school, a god, a class, or a card built on "this spell does something random," and the graveyard is full of the ones that hid the dice. The survivors share two design moves so consistently it reads like a law — publish the table, bound the streak — and one thing none of them ever did: make the chaos reliable.

Ultima, the ancestor who refused

Start where this game's lineage starts. Ultima Online — whose designers went on to build this game — shipped no wild-magic school at all, and the refusal was a design position. Every UO spell did exactly what it said; the randomness lived in whether the cast succeeded, and even that was escapable — forty skill points above a spell's bar and you never fizzled again. The one place UO let dice touch a player's fortune, the famous Luck stat, was pure upside: it could only improve loot, never worsen it, and once the community reverse-engineered its formula, luck suits became a beloved gear meta. The lesson our Chaos school inherits is double-edged. A school of random outcomes is a deliberate departure from the lineage — legitimate, but chosen, and the ancestor's terms still apply: randomness players accept is randomness they can learn, bend with skill, and read about on a wiki that gets the numbers right.

The wild mage's forty-year argument

The tabletop lineage is where wild magic was invented, and it ran the full experiment. The 1991 wild mage rolled on a public d100 surge table — a hundred outcomes, printed in the book, that players read for entertainment — and its signature spell was a deliberate surge: name the spell you want, pull the lever, and your level shifts the roll in your favor. Twenty-three years later, the 2014 wild magic sorcerer hid the trigger behind a 5% chance the table master had to approve, and the loudest complaint in the class's history followed: not "the chaos hurt me" but "the chaos never happens." The community's standard houserule escalates the surge die every quiet cast until a surge fires — players retrofitting a pity timer onto their own subclass. The 2024 revision capitulated in full: the trigger became player-owned, a feature lets you surge on purpose, and the mastery ladder is distribution control — roll the table twice and pick, then once a day, choose. Nobody, in forty years, houseruled the chaos down. People who choose the chaos school are angrier when nothing happens than when something terrible does.

The gods and the price

Two systems define the outer edge. Dungeon Crawl Stone Soup's chaos god Xom is pure, opaque, unbiddable randomness — no piety bar, just a mood you cannot read acting on a schedule you cannot see — and the community's verdict is precise: Xom is the game's designated challenge run, celebrated for stories and never for power. Opaque chaos survives only where choosing it is the challenge. Warhammer's winds of magic sit at the other edge: the caster chooses how many dice to stake on every spell, the jackpot and the catastrophe share a single trigger — the same double six that makes a spell undispellable also detonates the miscast table — and the one result players still curse decades later is the one that destroyed permanent character progression rather than dealing acute damage. Its video-game adaptation displays the miscast chance on the tooltip and lets its chaos faction, fittingly, manipulate the winds themselves. Between them, the two systems draw the school's cost rules: stakes chosen, prices posted, damage acute and self-contained — and at the top of the tree, the chaos specialist's fantasy is leverage over the randomness system itself.

What the competitive games proved

The multiplayer generation added the engineering. Dota 2 runs practically every proc in the game on a pseudo-random ramp: an advertised 25% chance actually starts near 8.5% and climbs with every miss until, by the twelfth attempt, it is guaranteed — the long-run average is exactly 25%, but streaks are bounded by arithmetic, not luck. The community documented the constants themselves, and that transparency is the acceptance. The same game's Chaos Knight — the designated variance hero — carries the genre's most elegant chaos spell: his bolt rolls damage and stun duration inversely linked, so a low damage roll means a long stun and both tails pay. World of Warcraft proved the same law from two directions: its proc systems multiply the odds after dry spells so that, in the designers' words, "extremely unlucky streaks should never happen" — while its hidden loot pity timer bred community trackers and resentment until it was surfaced. And Polymorph, the game's beloved twenty-year spoiler spell, survives because the victim always has four outs and the sheep heals while it grazes. Hearthstone ran the chaos-bomb experiment: Yogg-Saron, the ten-mana prayer that casts a random spell for every spell you cast that game, became the most controversial card its studio ever made — loved as an elective, expensive, slightly-bad spectacle, nerfed within six months of becoming tournament-correct, and then domesticated over eight years into threshold gates, enumerated wheels, and finally chosen outcomes. Unbounded chaos is lovable exactly as long as it is opt-in, expensive, and slightly bad.

The deck games and the honest ledger

The card games supplied the vocabulary this school needs most, because our combat is a deck. Slay the Spire's scry — look at the top of your deck, discard what you don't want — and its draw-pile-revealing relic turned hidden draw randomness into open information, and the community treats draw manipulation as the strongest climbing skill in the game. Its most instructive relic gives two extra cards a turn while randomizing all costs on a known distribution: players architect entire decks around chaos they can compute, and rate it among the best relics in the game. Balatro prints the odds directly on its gambling jokers — "1 in 4" on the card face — ships a joker that doubles those printed odds, and its community's response to five misses in a row is a meme, not a rigging accusation. Path of Exile added two more entries to the dictionary: chaos damage that bypasses a standard defense layer — instantly menacing, and answered in the same design by an opt-in immunity keystone that became one of the game's iconic build choices — and the two-word randomness modifiers Lucky and Unlucky: roll twice, keep the better; roll twice, keep the worse. Public, bounded, computable — chaos as a language players can build sentences in.

The two cautionary tales

Guild Wars 2's Mesmer carries a trait line literally named Chaos, whose signature armor grants a random boon and inflicts a random condition when struck. It is the safe way to do chaos — every face of the die pays something — and it is also the forgettable way: the outcomes are small, the player has no lever over them, and a decade of patches has sweetened its cooldown three times without ever making the dice the reason anyone picks the line. Random garnish is not an identity. Final Fantasy XIV's Astrologian is the graver warning from the opposite side: a healer sold for a decade on random card draws with manipulation verbs — redraw, hold, convert — whose randomness was progressively simplified and then, in 2024, removed entirely. The result was open mourning on the official forums: the randomness was the reason its players played it, and its removal read as class deletion. A randomness identity, once shipped, is a covenant with the players who chose it. That covenant is this page's refusal, stated in Section 0: we will not make Chaos reliable — and it cuts both ways, because it also means we owe the school dice worth choosing.

The capability matrix

Rows are the capabilities the genre converged on; the two highlighted columns are this game today and this game if the forty proposals ship. A filled circle is a first-class shipped capability, a half circle is partial or vestigial, an empty circle is absent.

Read down the "SotA today" column and the diagnosis writes itself: the school already has the genre's best cost grammar — every chaos power pays a posted price, which most of these games never shipped — and almost none of the supporting structure that makes the gamble playable. The proposals column is not a redesign. It is the rest of the genre's homework, done.

The forty proposals

Tier 1 fixes what is broken or dishonest, mostly in data, with the defects filed as tracker bugs the day this page went up. Tier 2 un-gates the school and makes it answerable: every spell works untrained, chaos damage becomes a real type, and the ward against it ships in the same patch. Tier 3 collects the identity the other pages reserved for this school — the deck verbs, the gambling verbs, the ghost spells with shipped descriptions — and Tier 4 is the capstone and the world. Every tuning number is an illustrative starting point: final values ship through the combat balance program's test-shard changesets, and for this school specifically, no damage number ships at all until the measurement tool is fixed and the catalog re-measured (S11).

Tier 1 · Honest dice — bugs, blanks, and the paperwork

Tier 2 · The mastery amplifies — un-gating the school and arming its counter

Tier 3 · Chaos owns the deck — the ghosts, the verbs, the gambles

Tier 4 · The sky that fell — the capstone and the world

Considered and rejected

• Making Chaos reliable. The page's standing refusal. No de-randomized bolt, no "pick your element" mode, no certainty anywhere. The genre's covenant runs the other way, and the school's thirty-two specialists signed it.
• A base-damage nerf from the old audit numbers. Drafted in July 2026, withdrawn the same day as a measurement artifact (S11). Quoting those numbers for tuning would repeat the mistake this page exists to correct.
• Daemon rebellion. The drafted "your daemon periodically turns on you" downside stays cut: highest grief potential in the program for a mechanic most players would read as a betrayal by their own build. The daemon negotiates (S33); it never betrays.
• Shipping the taming crossover spell. The school's third disabled row is a pet-buff spell that scales on taming and bard mastery and targets tamed creatures — a taming skill filed under chaos. It belongs to the taming program and is handed there with a note, including the stale mastery reference it still carries from an old fix.
• The Ebon Ledger. A joinable Ebon Dawn infiltrator arc surfacing the hidden Oracle endings — the one true epic in the fiction stock, and too big to ride this program. Future initiative, on its own budget.
• An Obsidian crafting school. Early drafts floated one; the game's data briefly grew a ghost "Obsidian" mastery by accident and it was formally deleted this month. The relics craft within existing schools.
• A fizzle-and-backfire overhaul. Two ghost passives promise fizzle and backfire manipulation, and a dead fizzle stat exists (S7) — but the fizzle system is a game-wide surface owned by the balance program, and a chaos-only fork of it would be exactly the kind of accident this page catalogs. The ghosts stay parked, named, reserved.
• An EV-positive luck stat. Pure-upside luck (the Ultima model) is loot design, not combat design; in combat it becomes mandatory. Improbability (S32) pays its premium for a real cost instead.
• Vendor-purchasable obsidian chips. The 2018 ask; answered differently — supply through the school's own scenes and weather (S38) keeps the reagent a reason to go outside.
• Hostile swap into hazards. The S27 swap validates destinations by the same rules as every teleport; "juxtapose them into the lava" is a griefing verb, not a chaos verb, and it stays impossible by construction.

Forty cards compose into one moment, repeated all night: the chaos mage reads the table, places the bet, watches the dice, and pays or collects in public. Here is that moment as the screen shows it — the Surge Ledger live above the hotbar, a storm on the ground, a transformation about to send its bill.

Read the anatomy top to bottom and it is the whole page: printed odds (S1, S31) over house-dealt dice (S21) with bounded streaks (S30); a storm that gambles on everyone while its caster stands warded at the center (S23, S16); and a capstone form whose price is printed while you are still enjoying it (S34). Nothing in the window is reliable. Everything in it is honest.

Sequencing

Five waves. The first two are data and honesty; nothing tunes until the yardstick is fixed, and nothing new ships until the dice are pinned.

flowchart LR
  S10["S10 pins"] --> S13["S13 un-gate"]
  S11["S11 yardstick"] --> S17["S17 passives"]
  S11 --> S22["S22 Onslaught"]
  S7["S7 dead stats"] --> S17
  S13 --> S14["S14 Juxtapose"]
  S13 --> S34["S34 Transformation"]
  S15["S15 chaos-typed"] === S16["S16 Eye of the Storm"]
  S16 --> S23["S23 Bedlam"]
  S21["S21 house dice"] --> S30["S30 streak governor"]
  S30 --> S31["S31 Surge Ledger"]
  S37["S37 Four Unseen"] --> S39["S39 Destiny Pool"]
  S14 --> S27["S27 hostile swap
(spike-gated)"]
  style S13 stroke:#a3771c,stroke-width:2.5px
  style S31 stroke:#7c4a9e,stroke-width:2.5px
  style S15 stroke:#4d7a3a,stroke-width:2px
  style S16 stroke:#4d7a3a,stroke-width:2px

Cross-page contracts

Appendix

A1 · Where the numbers come from

Usage data is a thirty-day window of live-server skill-cast telemetry ending 2026-08-14: 8,313,849 casts by 1,251 distinct casters, filtered to actual glyph executions. Chaos school: 31,896 casts, 96 casters. Per-spell counts and player counts are in Section 1; combo outputs were measured by name, except the four elemental-summon inversions, whose outputs share names with the base summon spells and cannot be attributed — stated here so nobody mistakes the combo table for complete. Rate limiting on the telemetry channel means every count is a floor, uniformly. Specialization population comes from the 2026-08-14 census of the live player database (120,337 characters): 99 players ever learned the chaos mastery, 66 reached specialization 80, 32 learned-it players active in the last year. The 2021 dev ranking (20th of 21) is from the public specialization-popularity thread. One measurement honesty note: Chaos Bolt's rare teleport branch does not appear in cast telemetry as a Night Step cast — branch outcomes are not logged, so the blank rate is computed from the weight table, not observed. And one number this page refuses to print: the old audit's damage ranking for this school, withdrawn as a tool artifact (S11).

A2 · Constraints honored

• The server holds no positions and runs no spell logic; every proposal ships as data, client code, or content — except S27, which is honestly labeled as the page's one new-network ask and spike-gated.
• Shared game data changes (new stats, reserved-slot activations) follow the standing two-tree mirror discipline; five proposals deliberately ride slots reserved in the data for years, precisely to avoid new shared surface.
• All new strings wait on the standing localization-pipeline check every school page has carried.
• Icon work batches into one atlas pass; the scrying panel (S39) is flagged as the program's largest UI risk and scheduled last.
• Mastery formulas stay evaluated on the authoritative side per the combat-sync rules; every card that touches them carries the live two-client verification requirement, because that bug class is invisible to editor tests.

A3 · The test surface

S10 is the harness: works-untrained pins for every formula at mastery zero and grandmaster, weighted-table expected-value pins, skill-sheet structural audits, string coverage. S11 adds the corrected expected-value model to the balance audit tool with a regression pin. Every table-rolling card (S22, S23, S26, S28, S30, S34) lands with distribution pins — long-run frequencies within tolerance of the printed odds, using the deterministic random source built for exactly this. The live checklist that cannot be automated: two-client agreement on every un-gated formula, the wipe, the swap if it survives its spike, the capstone's bill, and the daemon's band — filed as tester tasks per wave, per the series' standing practice.

A4 · Errata to our own documents

Research for this page found the school's internal reference wrong in four load-bearing places, corrected alongside this page: Juxtapose is not a teleport and never was (its four outcomes are heal, harm, and two stat swings — the "paired displacement" description matched nothing in the data); Chaos Bolt has five branches, not three (the self-buff and teleport faces were undocumented); Confusion's rider grants the target resistance, not a lingering impairment — the documented direction was backwards; and the school's second silent-cast list was incomplete. One erratum exports: the life page's "one chaos glyph anchors eight combos" undercounts — the verified number is nine. This is what Tier 1 means by honesty passes: the documents lie to us too.

A5 · Sources

Community threads cited on cards, all public: the 2018 chaos specialization feedback thread (Sakarisei, devourerofmemes, redfish, Quenton, Curt, Black Tortoise); the 2018 chaos shield thread (Gurney2, Barugon, neoandrew); the 2020 mage-builds thread (Lazlo, Violet Ronso, Joe Zhudarak); the 2021 summon comparison and daemon-scaling threads (Chiasma, thesometimeslurker, Cora Cuz'avich, Razimus, DeadnGone); the 2022 Chaotic Feedback counter thread (FBohler, Coswald_Dirthmire, Xee, Anpu); the 2025 Tabula Rasa investigation (Krovakin, Zealot); the 2026 chaos-resistance thread (Wishdragon, Rinaldi, Barugon, Adam Crow); the obsidian-chip farming thread (Nevyn Waldail, Nib, Cordelayne, Jason_M, Floors, Ataniiq, Elwyn); the 2021 specialization-popularity thread (Anpu, Powersurge45); and Amber Raine's 2018 release-notes comment. Genre systems studied for Section 2: AD&D's wild mage and D&D 5e's two wild magic revisions, Dungeon Crawl Stone Soup's Xom, Warhammer Fantasy's winds and miscasts with Total War's adaptation, Ultima Online's magery and Luck, World of Warcraft's Polymorph and proc mathematics, Guild Wars 2's Mesmer chaos line, Hearthstone's Yogg-Saron arc and pity timers, Dota 2's pseudo-random distribution and Chaos Knight, Final Fantasy XIV's Astrologian, Path of Exile's chaos damage and gambling economies, Slay the Spire's scry, and Balatro's printed odds.

General comments

No account needed. Every proposal above also has its own Discussion box for talking about that one alone; this thread is for the plan as a whole. Comments are plain text, held to basic decency, and may be trimmed.