The Honest Anvil

SotA's armor crafting already has the hard part right. Enhancement spends the same resource combat spends, durability, so crafters sell replacement as well as power, and a failed masterwork never destroys the item. What is missing is everything around that core: the odds, the bonus pool, and the durability cost are invisible until after the roll; a failed attempt banks nothing; and a finished suit under-reports itself on the character sheet. This document surveys what the genre learned about armor and crafted buffs, measures our bench against it, and proposes twenty improvements, half of them displays of numbers the game already computes.

Executive summary

Ultima Online ran the whole experiment in public. It shipped a single legible Armor Rating in 1997, replaced it wholesale with a five-resist item lattice in 2003, killed its own crafter economy the same year with item insurance, and then spent two decades bolting determinism back on: imbuing budgets, named reforge bundles, refinement caps. Every game since has walked some version of that road, and by now the genre has converged on one answer: keep the roll, but always sell a guaranteed path beside it, and never let the buff system eat the demand sink that keeps crafters in business. ESO sells its guarantee at twenty times the materials. New World sells it as a consumable ladder. FFXIV gives the floor away and charges for the ceiling. Path of Exile fought its players over determinism for five years and lost.

Shroud is in a rare position on that map: the durability sink is intact, failure never destroys the item, and unsold player goods flow back into the world as loot. Half the games surveyed here have no equivalent. The gaps cluster in three places:

1. Legibility — the station shows one odds number for one attempt and hides the rest: the full success ladder, the bonus pool this piece can roll, the durability budget, the reroll chance, the repair quote (A1, A2, A3, A4, A7, A8).
2. Agency — the roll cannot be aimed, a failure banks nothing, and there is no guaranteed lane at any price. These are the community's own top crafting asks (A10, A11, A12, A13, A14, A15).
3. Cohesion — the worn suit under-reports itself: no set-bonus readout on the sheet, an invisible broken-piece cliff, no per-piece breakdown; and the armor data has holes, from cloaks that ignore their materials to a designed full-suit bonus that sits inactive in the game data (A5, A6, A9, A16, A17, A18, A19, A20).

Twenty proposals follow. Ten are pure client work, displays computed from data the game already holds. Seven are server-backed and flagged with their exact persistence cost. Three form one server-side balance changeset that reconnects designed-but-dormant armor math. No card lowers the price of successful work; the one partial refund of failure costs (A14) is capped and priced in rare materials. The durability sink is load-bearing and this page treats it as such.

This page is a per-tree deep dive in the same series as the inventory, crafting, and alchemy proposals, and re-proposes nothing from them. Three cards the crafting page explicitly set aside for the per-tree dives are picked up here: the guaranteed lane (A12), directed enhancement (A13), and separating the success roll from the exceptional roll (A15). A17 is the data half of the crafting page's bulk-salvage card; the two cross-reference rather than overlap.

Tier B = the balance changeset: three server-side changes that ship together through the balance pipeline with telemetry, after the October launch. A twenty-first idea, re-tempering a finished piece, appears as a directional sketch at the end of Section 3 rather than as a numbered proposal.

Where we are today

The surfaces this page touches are already modern: the crafting station, recipe book, recipe card, and the bonus-effect picker are all on UI Toolkit, and so are all three character sheets, the live paper doll, and the thirteen equipment slots with their full drop-rule matrix. The enhancement systems themselves are real systems, not stubs: two independent five-effect tracks (masterwork at the smithing or tailoring bench, enchantment at the alchemy bench), declining odds per applied effect, a no-repeat rule so a bonus is never offered twice, a mastery-based reroll of the offers, and failure that costs durability but never the item. This page builds on all of that and re-proposes none of it.

Here is the character sheet as it stands, with the gaps numbered:

• 1Three aggregate numbers, no breakdown. The game computes each worn piece's contribution to those totals every time they change; the sheet shows only the sums. Proposal A9.
• 2The set bonus lives only on item tooltips. Nothing on the sheet says which pieces count toward it, that the torso gates the whole bonus, or that one broken piece drops the count. Proposal A5.
• 3The broken cliff is invisible. A piece at zero durability contributes no protection at all, and the only signal is a small status icon. Full protection at 1 durability, zero at 0, and the sheet never says which state you are in. Proposal A6.
• 4The tooltip under-reports the work. Masterworks and enchantments appear in display-sorted order, which is not the order they were applied in, and that order sets their power. There is no count header and the item's quality score is shown nowhere. Proposal A8.

And the station, mid-masterwork:

On success, a second window offers three bonuses drawn by the server:

• 5One odds number for the next attempt only. The full ladder is computed by the game and shown nowhere. At a representative skill it runs roughly 96%, 78%, 42%, 13%, 6% for the first through fifth effect, and nobody can price a five-effect commission without it. Proposal A1.
• 6The pool is knowable but never displayed. The exact list of bonuses this piece can roll, filtered by armor type, body part, station, and skill, is knowable from data the game already holds, and is shown only three at a time, after the roll. Proposal A7.
• 7The budget is invisible until it is spent. Most effects cost 10 to 28 maximum durability, a failure costs 25, and below 30 maximum durability the piece closes to further work until its maximum durability is restored, with major repairs returning less each time. The only message about any of this arrives too late: "Durability must be more than 30 for masterworking." Proposal A2.
• 8Revise Effects rolls blind. The reroll succeeds at a rate that halves per revision, and the button shows no percentage. Proposal A3.
• 9Order matters and nothing says so. The durability enchantment gives more when applied later; sequencing is real strategy and the UI offers no way to plan it. Proposal A10.
• 10No memory, no aim, no guarantee. A failed attempt banks nothing (A11), the roll cannot be steered toward defense or vitality (A13), and there is no guaranteed lane at any price (A12). These three are the server-backed heart of this page.

The machinery every proposal plugs into

One shared rulebook drives the whole bench. The station's display reads the same rules the server uses to resolve an attempt, and today those rules feed exactly one summary line. The station validates the recipe, tool, and the 30-durability floor; a successful roll draws three offers from the pool for this piece's armor type and body part; Revise Effects is a server-gated reroll with halving odds; the chosen effect is committed with a snapshot of the crafter's skill, and the finished piece carries its full record from then on. Every proposal in Section 3 is a new reader of this pipeline; none of them forks it.

flowchart LR
  A["Station: enhancement recipe,
target piece on the table"] --> B{"Server validates recipe,
tool, materials, and the
30 durability floor"}
  B -->|roll fails| C["Failure mark on the item
minus 25 maximum durability"]
  B -->|roll succeeds| D["Server draws 3 offers
from the pool for this
armor type and body part"]
  D --> E{"Revise Effects:
server-gated reroll,
odds halve each time"}
  E --> D
  D --> F["Player selects one"]
  F --> G["Server commits the effect
with a skill snapshot, spends
materials and durability"]
  G --> H["Piece returns to inventory"]
  H --> I["Worn: effects add to the
character, and combat is
resolved on the server"]
  style D stroke:#c9a55a,stroke-width:2px
  style G stroke:#c9a55a,stroke-width:2px

Lessons: Ultima Online to modern

UO 1997 had one armor number. Armor Rating was legible at a glance, plate was the ceiling, and because gear broke and was lost, a grandmaster blacksmith's exceptional plate was a consumable with permanent demand. It also had no build space at all: everyone wore plate, and armor was a solved problem within a week. The 2003 Age of Shadows expansion fixed that by replacing AR with five resistance channels and a general magic-property system, and armor became a genuine optimization puzzle. It also, in the same stroke, made UO an item game, buried suit math in spreadsheets, and shipped item insurance, which ended item loss and with it the crafter's structural demand. The community's verdict on insurance has hardened over twenty years: nothing felt rare, everyone wore identical suits they never lost, and the blacksmith became a hobbyist. Shroud kept the sink UO deleted. That single fact is worth more than any feature in this catalog, and every proposal here is written to leave it alone.

The rest of UO's arc is a two-decade search for the right amount of determinism. Runic tools (1999) paid crafters in lottery tickets: burn charges and hope for the property you wanted. The standing player complaint about them reads exactly like Shroud's forum today: too random, too wasteful. Imbuing (2009) was the correction and the best crafting system UO ever shipped: pick the property, spend a budget, failure never destroys the item, and a whole loot tier gets melted down into crafting ingredients. Runic Re-Forging (Publish 74) added the pattern this page borrows for A13: pick a named bundle, and the roll stays a roll but lands only inside the family you chose. And enhancement, the one UO system where failure could destroy the item, is the cautionary tale: players simply refused to use it until a protection consumable existed.

What each game contributes

World of Warcraft ran the agency experiment twice. Reforging (2010) let players convert 40% of one secondary stat into another not already on the item; it was removed four years later because the optimal reforge was computable, so every drop became homework with a right answer. Titanforging, the random upgrade lottery on finished gear, was deleted with the stated principle "let loot be loot." What survived is the pattern Shroud should note: sockets stayed random, but a purchasable consumable adds one deterministically. Transmog settled a different question for the whole genre: appearance and stats should never share a budget, and Shroud's patterns already agree. The read: never add a layer with a computable right answer, and sell determinism as an item, not a menu toggle.

Elder Scrolls Online is the closest thing to a proven answer for crafted armor. Traits are chosen, not rolled, unlocked by researching sample items over real time; crafted sets live at physical stations out in the world, so crafting has geography; and quality improvement is the one gamble, with a posted exit: one Tempering Alloy is a 5% upgrade chance, twenty is 100%. A crafter can therefore quote an exact item at an exact price, which makes crafting a service business. Transmute Crystals then let players re-trait dropped gear at a real cost. The read: ESO's twenty-tempers rule is the exact shape of A12, and it is also what Shroud's own forum asked for in as many words.

Final Fantasy XIV splits every piece's buff capacity into guaranteed slots and gamble slots. Melding into the first slots always succeeds and the player picks the materia; overmelding past them is published-odds gambling that only the obsessed pursue. Playing in your gear generates the buff currency itself. The read: Shroud's five-enchant track with declining odds is structurally FFXIV without the guaranteed floor; A11 and A12 build that floor two different ways.

Guild Wars 2 has zero randomness in its buff layer: runes and sigils are named, purchasable, deterministic components, and legendary gear eventually swaps stats freely. It is the least-complained-about gearing in the genre, and its crafting economy is also the shallowest: with no loss and no decay, buff components are bought once and the market is a puddle. The read: determinism is pleasant, and pure determinism with no sink is a dead crafting profession. Shroud's durability sink is what lets it import the pleasant half safely.

Albion Online makes crafting the whole game: nearly every item is player-made, item power is one legible scalar, enchantment materials are visible in the craft UI before committing, and deep specialization gives a master crafter a real market identity. Full-loot PvP supplies the demand. The read: Shroud's unsold-goods-into-loot loop is a gentler version of Albion's demand engine and deserves the same pride of place; and specialization payoffs should be published numbers, not folklore.

New World walked, under player pressure and in public, from a pure drop lottery to a ladder of purchasable certainty: a Craft Mod guarantees one perk, a Timeless Shard adds a guaranteed attribute, a Golden Scarab guarantees an attribute and two perks, and Gypsum turned random endgame progression into pick-your-activity progression. The read: the consumable ladder is the most directly transferable shape for A13's catalysts, with one hard rule attached: none of it can ever be sold for real money.

Black Desert Online is the cautionary tale on both ends. Enhancement can downgrade or destroy gear, protection is monetized through the cash shop, and the community's standing thread asks plainly for an enhancement system without punishing failures. But inside that system sits the single best idea in this whole survey: failstacks, a pity counter that converts every failure into a better next attempt, turning tilt into anticipation. The read: Shroud already refuses BDO's cruelty, since failure never destroys the item; A11 imports BDO's one kindness.

Path of Exile has the deepest crafting in the genre and spent five years fighting its own players about determinism. Harvest league shipped targeted crafts; the developers nerfed them to protect the thrill of the lottery; the backlash became one of the loudest in ARPG history; and the dispute resolved almost entirely in the players' favor, with recombinators and Perfect Essences shipping exactly the determinism once refused. The read: "keep it random so it stays exciting" is a losing long-term position, on the record, across two games. Lose it on purpose, early, and design the guaranteed lane yourself.

Capability matrix

● native · ◔ partial or third-party · ○ absent. *Steering is marked absent for systems that are already fully deterministic; there is nothing left to steer. "SotA today" = the migration branch, 2026-07.

The seven proposals

These are the armour-specific cards. Everything about the shared enhancement bench moved to the enhancement chain page, which is where the odds ladder, the durability budget, the pool browser, the effect tooltip, the planner, the practice bonus, certain work, focused attempts and the restoration recipe now live.

The worn suit · A1 to A3

Three client-only displays. The suit you are wearing under-reports itself in three separate ways, and every input needed to fix that is already on the client.

Data completeness · A4

One armour-math ship · A5 to A7

Three server-side changes that ship together through the balance pipeline with telemetry, after the October go-live. They move the same numbers, so they move them once rather than three times.

Considered and rejected

• Everything on the shared enhancement bench. Item insurance, Crown-store determinism, destroy-on-fail variants, the full intensity-budget rework and the 5+5 cap question are all argued on the enhancement chain page, because they apply identically to weapons.
• A real third armour weight class. The medium-armour gap is real: medium pieces currently get no set bonus at all. Folded into A5's cleanup rather than made into a new system.
• Per-element resistance channels on armour. The gem and attunement layer already carries elemental defence deterministically; a second channel would double-count it.
• New base character stats for armour bonuses. The stat enumeration is append-only by ordinal and an unknown identifier in a formula resolves to a silent zero. Real risk, no player gain.
• A smithing minigame, and any further UI migration work: the station, picker, sheets, doll and tooltips are already on UI Toolkit.

The composed surface

One surface, not seven widgets. The character sheet owns the worn-suit story: which pieces count toward the set, what is broken, and what each piece is actually contributing. The enhancement decision lives at the bench on the enhancement chain page, and this page deliberately does not duplicate it.

The character sheet, composed

• 1The armor-set block with its broken-piece line and next-piece delta is A1; the "lost to broken gear" line is A2.
• 2The hover breakdown, with pieces lighting up on the doll, is A3; its rows sum to the total above it.
• 3Item tooltips everywhere on this window carry the full enhancement record, the enhancement page's effect tooltip.

Sequencing

Recommended phase 1: A1, A2, A3. The shared bench work is sequenced on its own page and the two can proceed in parallel; the only ordering constraint between them is that A5's set-bonus display leans on the enhancement page's effect tooltip.

Appendix

What each display reads

Constraints honoured

• Shared logic is extended, never forked; displays call the shared helpers rather than restating the maths.
• Effect formulas are pre-parsed token arrays and an unknown identifier resolves to a silent zero, so any new effect asset needs editor-tool authoring plus a drift tripwire test.
• The balance changeset ships with the armour formulas' first test coverage; today's only armour test is a cosmetic asset scan.

Sources

UOGuide and the official Ultima Online wiki, Blizzard's professions posts, the Warcraft wiki, Last Epoch's crafting posts, Diablo IV patch notes, Final Fantasy XIV materia documentation, ESO transmutation, Path of Exile's crafting bench, New World and Albion crafting documentation. Claims that could not be confirmed from a primary source were cut rather than softened.

Comments and votes below are per proposal. The shared bench that armour and weapons both sit on is argued at the enhancement chain.

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.