The Enhancement Chain

Everything a crafter can add to a finished item runs through one chain: an exceptional roll at the forge, up to five masterwork effects, up to five enchant effects, one gem socket, and the material bonuses carried up from the components. Armour and weapons use the same odds, the same magnitude formula, the same failure penalty and the same durability floor. This page is that chain. Its two sibling pages, armour and weapons, cover what each kind of gear does with it.

Executive summary

Ultima Online launched in 1997 with a crafting system of two moving parts: a craft was exceptional or it was not, and the smith's name went on it. Everything the genre built afterwards is a reaction to what happened next. In February 2003 the Age of Shadows expansion put random properties on every drop and crafted goods became vendor fodder within weeks. It took UO six years and three separate systems to climb back out, and those three systems are the shape of this document: cap it (Imbuing, 2009), aim it (Runic Reforging, 2012), and show it, which the rest of the genre finished in Last Epoch's Forging Potential in 2021 and Blizzard's Concentration in 2024.

SotA built the ambitious version and then left the instrument panel off. The machinery is genuinely good: two independent rails whose odds scale continuously off crafter proficiency, magnitudes that are live formulas rather than baked constants so a rebalance is retroactive and free, crafter proficiency stored per effect (almost nobody else in the genre does that), a reroll path before you commit, and a failure that never destroys the item. What is missing is everything around that core.

Three gap clusters:

1. The budget nobody can see. Maximum durability is not a condition meter in this game, it is the enhancement budget. Nearly every masterwork and enchant spends it, every failure spends 25 of it permanently, repairs never give any of it back, and the whole chain locks out below 30. A crafted one-handed weapon starts at 165, which is five failures. Ever. Reaching the fifth slot on both rails costs roughly 42 expected failures, so the advertised ten-effect ceiling has never been reachable on a one-hander by anyone, at any skill, with any amount of gold.
2. No agency at the point of decision. The odds, the pool and the cost are all invisible until after the roll. A failed attempt banks nothing toward the next one. There is no way to aim at the effect you want, no way to buy certainty, and nothing to protect the item while you try.
3. No way back. There is no un-enchant, no un-masterwork, no un-socket and no scar removal anywhere in the game. The engine ships a remove verb for the cosmetic layer, so patterns come off; the power layer has none. Every enhanced item is a locked build, and every scar is forever.

Eighteen proposals follow. Seven are client-only displays that need no new server traffic at all, because every effect definition already ships inside the game client and the odds and magnitude functions are shared logic that already runs locally. Three change the economics and are data or shared constants. Eight need server work and each names its cost.

The chain as built

Credit first, because a lot of this is good and none of it should be re-proposed. Enhancement magnitudes are not baked numbers; they are formulas evaluated when the item loads, which means a rebalance reaches every item already in the world at no migration cost. Crafter proficiency is stored with each individual effect, so the smith who rolled it is permanently part of what it is worth. The odds scale continuously off skill rather than stepping at breakpoints. Failure never destroys the item, which puts SotA ahead of a long line of games that learned that lesson the hard way. And the whole effect catalogue is data, so a new bonus type is an asset rather than a code change.

flowchart TD
  M["Refined materials"] --> C["Component parts"]
  C --> R{"Craft roll"}
  R -->|failure| X["Materials lost"]
  R -->|success| W["Item, plus material bonuses carried up"]
  R -->|critical| WE["Item, plus one exceptional effect"]
  W --> MW{"Masterwork, up to 5"}
  WE --> MW
  MW -->|success| MO["One stat or ability effect, plus a durability roll"]
  MW -->|failure| SC["Permanent scar: 25 max durability, 1 item score"]
  MO --> EN{"Enchant, up to 5"}
  SC --> EN
  EN -->|success| EO["One gem effect, plus a durability roll"]
  EN -->|failure| SC
  EO --> S["Socket, exactly 1, one-shot"]
  S --> F["Finished item"]
  F --> RP["Repair restores working durability only"]
  RP --> F
  F --> LK["Locked out below 30 max durability"]

Note the two arrows that diagram does not have: any arrow that removes an effect, and any arrow that puts maximum durability back. Both absences are load-bearing for everything below.

The audit

Measured against the shipped data and the shipped formulas. Where a number is a projection it says so.

Cost and risk: the durability tax

Worked examples

Four calculations, all at Grand Master. Crafting proficiency is the square root of skill over 200, so skill 100 is proficiency 70, and skill 200 (the theoretical cap, which no player is at) is proficiency 100.

W1 · What the slot ladder really costs

Per-attempt success by slots already filled is a base of 75, 40, 10, 0, 0 percent plus proficiency times 0.34, 0.60, 0.50, 0.20, 0.10. Masterwork and enchant use the same curve.

Even at the theoretical skill cap the fourth slot is 20% and the fifth is 10%, still 13.7 expected failures and 342 durability to reach the fifth.

W2 · The floor, not the odds, is the cap

A crafted one-hander starts at 165 maximum durability and refuses enhancement below 31, so it tolerates five failures before it is locked out of both rails forever. A two-handed melee weapon tolerates eight; a bow, seven. Successes push back a little, because nearly every masterwork also rolls a durability bonus averaging about 7.5 and weapon enchants roll larger ones.

W3 · What chasing the fourth slot costs in gold

A masterwork attempt consumes five silver ingots; an enchant attempt consumes five gold ingots and ten mandrake root. At merchant-buy prices that is 500 gold and 1,040 gold per attempt respectively.

The practical three-and-three build is about 6,860 gold and three expected scars. Everything past it is 33,000 gold spent on an outcome the durability floor will not permit.

W4 · What three masterwork slots of pure damage are worth

Duplicate effects are excluded from the offered set, so the best damage stack is one of each damage effect. At Grand Master the slot multipliers are 1.56, 1.61 and 1.66.

Three masterwork slots of damage, at the practical ceiling of the system, are worth less than a twelfth of what changing which weapon you crafted is worth. That is the crafted-versus-drop problem in one table, and it is precisely the problem Ultima Online spent 2003 to 2009 solving.

W5 · The crown repair ladder, priced

A crafted longsword at 165 maximum durability, repaired with crowns from empty each time, restores 165, then 110, 82, 66, 55, 47, 41, 36, 33, 30. Ten repairs cost 2,500 gold crowns and buy 625 points of working durability. A brand new longsword is 165 points for three ingots.

Lessons: Ultima Online to modern

Ultima Online is not one comparison among many here. It is the direct ancestor, and the only MMO that has run this whole argument in public across three decades.

1997 to 2002. Crafted goods had two properties that mattered: exceptional or not, and whose name was on it. The maker's mark was gated behind both Grand Master skill and an exceptional roll, so a signature certified two facts at once. That is what turned it into a brand, and named smiths ran real businesses on it for years.

February 2003. Age of Shadows put random properties at random intensities on every drop and crafted goods became vendor trash. This is the most instructive event in the genre for our purposes, and the reason the sibling pages both end with a card about crafted gear competing with drops.

2009, cap it. Imbuing let crafters put chosen properties on an item, bounded two ways at once: a hard five-property limit and a 500-point weighted intensity budget, where a strong property costs more of the budget than a weak one. Unravelling, shipped alongside it, fed junk drops into the reagents imbuing consumed. Together they moved crafters from competing with the loot table to being downstream of it, which is the structural insight most games miss.

2012, aim it. Runic Reforging let a smith spend more tool charges to narrow which properties could roll, through named option pools. Path of Exile shipped the same idea as a crafting bench in 2013, with exactly one guaranteed player-chosen modifier as a deterministic floor under a random system, then as essences and fossils. When Path of Exile tried to walk determinism back in 2021 it produced the loudest community reaction in that game's history.

2021 to 2025, show it. Last Epoch replaced its item-fracture chance with Forging Potential, a visible budget printed on the item that every craft depletes, where every craft succeeds until the budget runs out. Blizzard replaced World of Warcraft's random Inspiration with Concentration in 2024, stating the goal as a more deterministic way to reach the same outcome, giving crafters agency and, notably, "a resource that you can use without having to give your customers a full statistical breakdown." Diablo IV shipped tempering with a limited charge count that could brick an item, took the loudest itemization complaint in its history, and walked it back with a scroll that restores charges.

The paragraph that strip adds up to: enhancement began as a coin flip with a signature, fragmented into an affix arms race that made crafters irrelevant, and was rescued by three ideas in order, which were to cap it, to aim it, and to show it. The last five years are a near-unanimous industry retreat from irreversible, invisible, low-probability failure. The surviving consensus is not "no risk." It is that the budget may run out and the ingredients may be consumed, but the action does not fail and the item does not die on a coin flip.

Where SotA sits

Read the two highlighted columns against the rest. The pattern is consistent and it is not flattering in one direction: SotA is at or near the front of the genre on plumbing and at the back on disclosure and recovery. Per-effect crafter proficiency is something almost nobody else stores. Formula-driven magnitudes that re-evaluate on load are genuinely unusual. And then the player is shown none of it, and given no way back from a bad roll.

Two rows deserve a note. Gear repairable indefinitely stays empty in the proposed column on purpose: terminal gear is a deliberate design position here and nothing on this page reverses it. What is proposed is that terminality become legible, survivable and productive when it arrives. And weapon speed is carried here only for completeness; it belongs to the weapons page.

The eighteen proposals

Each card carries its tier, its rough size, whether it needs server work, the genre precedent, and implementation notes. Cards that came from the armour dive and cards that came from the weapons dive are not marked differently, because the whole point of this page is that the bench does not know which one you put on it.

Show the system · E1 to E7

The economics · E8 to E10

These three change what the chain costs rather than what it shows. They are data or shared constants, so they are cheap to ship and expensive to get wrong, and they should be modelled together: the failure penalty, the slot curve and the carried-forward bonus are three terms in one equation.

Agency · E11 to E18

Each names its cost. Anything that changes what is stored on an item lands with its migration-ledger entry in the same change, and anything that adds a new operation rides a regenerated call table and a deployment preflight. All of these sit behind the change freeze.

Considered and rejected

• Item insurance or blessing. UO shipped it and it is widely held to have done more damage to the crafter economy than any single mechanic in that game's history, because gear stopped being consumed. The whole demand sink here depends on items eventually leaving the world.
• Crown-store determinism. Selling guaranteed outcomes or scar removal for real money turns the entire risk model into a storefront. E11 and E12 are deliberately material sinks, and neither should ever be purchasable.
• Item destruction on a failed enhancement. Every game that shipped this has since added a mitigation item. The current 25-point scar is already the humane version; do not escalate it.
• A full intensity-budget rework on UO's imbuing model, where each effect costs a share of a weighted total rather than one slot. Genuinely the more elegant design and a real answer to "always stack the five strongest". Rejected for now because it replaces the chain rather than extending it, and because UO's own cost is that players plan items in third-party calculators. Revisit only if E6 proves the in-game arithmetic can carry it.
• Path of Exile style re-roll currency applied repeatedly until the effects are right. Wrong shape: that economy needs disposable items, and gear here is scarce, signed and permanently marked by every attempt. E13 and E15 give a bounded, priced version of the same freedom without the gambling loop.
• New base character stats for enhancement bonuses. The stat enumeration is append-only by ordinal and an unknown identifier in a formula resolves to a silent zero, so the risk is real and the player gain is nil. Use the stats that exist.
• Raising the five-and-five caps. Both sibling dives independently concluded the existing caps are already unreachable. Adding a sixth slot to a chain whose fourth is decorative is not a feature.
• A crafting minigame. Out of lane, and it would make E11's guaranteed lane incoherent.
• Turning on region-locked enchant pools. The machinery is authored and every pool currently sits at no region. Switching it on would create travel-gated crafting overnight. Documented as available; not proposed.

The composed bench

The anti-pattern is eighteen features becoming eighteen widgets. Everything client-side above composes into one bench panel, and the rule is that a crafter should be able to answer "what are my odds, what can I get, what will it cost, and what happens if it fails" without leaving the station.

Zone anatomy

Two numbers in that mockup do not exist in any form today, and they are the ones the whole page is arguing for: 4 failures left, and on failure: 23 budget, no slot lost.

Sequencing

flowchart LR
  E1["E1 odds ladder"] --> E6["E6 planner"]
  E2["E2 budget"] --> E6
  E3["E3 lifetime"] --> E2
  E4["E4 effect tooltip"] --> E2
  E4 --> E15["E15 re-tempering"]
  E5["E5 pool browser"] --> E6
  E5 --> E13["E13 directed rolls"]
  E7["E7 reroll odds"] --> E1
  E8["E8 survivable scar"] --> E9["E9 slot cliff"]
  E9 --> E10["E10 failure pays forward"]
  E10 --> E11["E11 certain work"]
  E10 --> E12["E12 safeguard"]
  E2 --> E14["E14 scar restoration"]
  E4 --> E14
  E15 --> E16["E16 reversible socket"]

All of Tier 3 sits behind the change freeze. Recommended phase 1: E1, E2, E3, E7.

Appendix

Why Tier 1 needs no server work

Constraints honoured

• Shared logic is extended, never forked. Every odds and magnitude figure displayed comes from the same shared functions the actual roll uses, so a shown number and a rolled number cannot drift apart.
• Displays never restate the maths. The one place the game already restates a formula in the UI is a known wart; this page adds no more.
• Effect formulas are pre-parsed token arrays, and an unknown identifier resolves to a silent zero. Any new effect asset (E13's category families) needs authoring through the editor tooling plus a drift tripwire test, or it will ship quietly doing nothing.
• Retunes stay retroactive. E8's proportional scar works because items store an effect's identity, not the number it evaluated to. No re-crafting, no migration.
• Stored-shape changes carry their ledger entry in the same commit. E10 and E14 add keys; both say so on their card.
• Balance ownership is respected. Nothing here changes a skill's numbers.

Test surface

Most of this page is unusually testable, because the interesting parts are pure functions. Existing fixtures pin the exceptional chance, the craft chance, the effect pools, salvage and repair. New coverage this page implies: odds-ladder goldens, a pool-filter matrix, durability-gate and scar accounting, roll-distribution goldens for E10 and E18, the planner and budget models, a non-English fixture for E4's grouping, and a drift tripwire per new effect asset.

Sources

Genre claims are drawn from UOGuide and the official Ultima Online wiki (exceptional crafting, imbuing's five-property and 500-intensity caps, Runic Reforging, Powder of Fortifying's charge ladder and its imbued-item restriction, durability); UO Publish 74; Blizzard's professions posts on Concentration in The War Within; the Warcraft wiki on normalization and sockets; Last Epoch's Eternal Legends post on Forging Potential; Diablo IV patch notes on tempering and restoration; Final Fantasy XIV materia documentation; ESO transmutation; Path of Exile's crafting bench and the 2021 Harvest coverage; Albion's Crafting Focus; and Star Wars Galaxies experimentation documentation. Age of Shadows is dated February 2003. Claims that could not be confirmed from a primary source were cut rather than softened.

Comments and votes below are per proposal. If you only read one card, read E2: it changes no mechanic at all, and it is the difference between a system players think is unfair and a system players think is hard.

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.