Item Crafting: Forging, Smelting and Rarity
How the Forge turns three materials into equipment, the exact rarity odds for every combination, why rare materials give 6% epic rather than 5%, and when each quality tier unlocks.
Updated 12 October 2026
Special servers let the hero craft equipment. Three materials go into the Forge and one item comes out, with a rarity and a quality tier decided by chance — and the game shows you those chances before you commit. This page explains what the numbers mean, where they come from, and which combinations are worth the materials.
Two scales decide what an item is worth, and they are set independently:
- Rarity — Common, Uncommon, Rare or Epic. This is what the materials control, and it is what makes an item's figures larger.
- Quality tier — I, II or III. This is the item's own grade, and the materials have nothing to do with it.
Everything above Common exists only on special servers, and Epic items can only ever be crafted. No adventure and no auction will ever hand you one.
The one rule behind every forge
Each of the three materials in the Forge contributes its own spread of outcomes, and the result is the average of the three. There is only one spread, and every material uses it:
A material lands on its own rarity 94% of the time, one rarity higher 5%, and two rarities higher 1%.
What makes that worth knowing is the ceiling. Epic is the top, so any upgrade that would go past it simply stops there. That single detail produces the whole table:
| Material you put in | Common | Uncommon | Rare | Epic |
|---|---|---|---|---|
| Common | 94% | 5% | 1% | — |
| Uncommon | — | 94% | 5% | 1% |
| Rare | — | — | 94% | 6% |
| Epic | — | — | — | 100% |
The Rare row is the one that looks wrong at first glance. Why 6% epic, when every other row gives 5% to the next rarity up? Because a Rare material's two-step upgrade has nowhere to go: it would land above Epic, so it lands on Epic instead, and the 5% and the 1% add together. Nothing special is happening to Rare materials — the ladder is just running out of rungs.
Follow that to the end and an Epic material always produces an Epic item. All three of its outcomes are capped at the top.
Working out any combination
Add up what each of the three materials contributes to a rarity, then divide by three:
P(rarity) = [ material 1 + material 2 + material 3 ] ÷ 3
Two Uncommon and one Rare, for example:
| Outcome | Working | Result |
|---|---|---|
| Common | (0 + 0 + 0) ÷ 3 | 0% |
| Uncommon | (94 + 94 + 0) ÷ 3 | 62.7% |
| Rare | (5 + 5 + 94) ÷ 3 | 34.7% |
| Epic | (1 + 1 + 6) ÷ 3 | 2.7% |
The game rounds the displayed percentages to one decimal place, so a column can read 99.9% or 100.1%. The underlying odds always total exactly 100%.
Every combination, worked out
C = Common, U = Uncommon, R = Rare.
| Materials | Common | Uncommon | Rare | Epic |
|---|---|---|---|---|
| C + C + C | 94% | 5% | 1% | 0% |
| C + C + U | 62.7% | 34.7% | 2.3% | 0.3% |
| C + U + U | 31.3% | 64.3% | 3.7% | 0.7% |
| U + U + U | 0% | 94% | 5% | 1% |
| C + C + R | 62.7% | 3.3% | 32% | 2% |
| C + U + R | 31.3% | 33% | 33.3% | 2.3% |
| C + R + R | 31.3% | 1.7% | 63% | 4% |
| U + U + R | 0% | 62.7% | 34.7% | 2.7% |
| U + R + R | 0% | 31.3% | 64.3% | 4.3% |
| R + R + R | 0% | 0% | 94% | 6% |
What this means for how you craft
Mixing drags you down, it does not average you up. A single Common material among two Rares takes the Rare result from 94% to 63%, and hands 31.3% of your outcomes back to Common. The Forge has no mechanism for three weak materials to combine into something strong — a third of your result is simply whichever material you chose least carefully.
Never pad a forge with Commons. If you are short of good materials, waiting is better than diluting. Three Rares give 94% Rare; two Rares and a Common give 63%, and nearly a third of the time you get a Common item for materials that could have made a Rare one.
Epic is a lottery you cannot buy into cheaply. The best epic chance from non-epic materials is 6%, from three Rares — and the difference between three Rares and three Uncommons is only 6% against 1%. Chasing Epic with Rares costs roughly seventeen forges per Epic on average. Get Epic materials and it stops being a lottery at all.
The cheapest upgrade is the one you already have. Three materials of one rarity give a 6% chance (5% + 1%) of coming out above that rarity. It is small, but it is free — you were going to forge anyway.
Quality tiers, and when they unlock
Quality is the second scale, shown as I, II and III beside the rarity chances. It does not interact with rarity at all: changing your materials moves the rarity odds and leaves the tier odds untouched.
What does change the tier odds is how far through the server you are. Tiers unlock in sequence, and a tier that has not unlocked cannot be rolled:
| When | Tier I | Tier II | Tier III |
|---|---|---|---|
| Start of the server | 100% | — | — |
| After ⅓ of the server | 50% | 50% | — |
| After ⅔ of the server | 33% | 33% | 34% |
Early on, every item you craft is Tier I, however good your materials. This is worth planning around: materials keep, and a Rare material forged in the first third of the server can only ever become a Tier I item, while the same material held back becomes a one-in-three shot at Tier III later.
Tier III takes the extra percentage point in the final stage, which is simply where the rounding lands.
Refining: the one guaranteed upgrade
Forging is a gamble. Refining is not.
Two identical items plus one material of the same rarity produce that same item, one rarity higher — guaranteed. No roll, no 94%, no disappointment. Two Common Helmets of Awareness and a Common material give an Uncommon Helmet of Awareness, every time.
The catch is the word identical: two helmets will not do, it must be two of the same helmet. That is a much harder thing to collect than three loose materials, which is why refining is the exception rather than the routine. Epic cannot be refined — there is nothing above it to reach.
This changes how you should treat duplicates. A second copy of an item you already have is not junk to be smelted; it is half of a guaranteed upgrade. Smelt the things you have one of, and keep the pairs.
Smelting: turning equipment back into materials
Smelting is where materials come from. It destroys an item and returns material of that item's own rarity — a Common item yields Common material, so smelting never upgrades anything by itself.
A finished smelt returns two materials. You can take one immediately and give up the second, which is the choice the button offers while the timer runs. Taking the early one halves your return, so it is worth doing only when that single material completes a forge you are making right now.
Smelting costs time rather than resources, and the time is banded by world speed rather than divided by it:
| World speed | Smelting time |
|---|---|
| x1 – x2 | 8 hours |
| x3 – x5 | 4 hours |
| x10 | 2 hours |
A 2x world and a 1x world smelt at the same rate; so do a 3x and a 5x. If you are choosing where to play and crafting matters to you, the jump from 2x to 3x halves your smelting time while a jump from 3x to 5x changes nothing.
Spending 2 Gold halves whatever time is left.
Because the limit is the clock and not your resources, the real constraint on crafting is how early you start. Equipment you are sure you will never wear is worth more in the furnace on day one than in your inventory on day thirty — and at 8 hours a smelt, a day of hesitation is a material you do not have.
What comes out of the Forge
Rarity and quality are the two things the Forge shows you. The item itself is not one of them: the type is picked uniformly at random across right hand, left hand, body, helmet and boots, and then uniformly at random within that type. You are forging a rarity, not a particular piece of equipment.
Reforge spends Gold to generate a second candidate so you can pick between the two, at a price that rises with each reforge on the same item. It is the only influence you have over what you get.
How these figures were worked out
The percentages on this page were read directly from the game's own Forge display across nine material combinations, and the rule above reproduces every one of them exactly — not approximately. The same rule then predicts the one combination not observed, three Commons, at 94/5/1/0.
The Epic row is the single figure here that is extrapolated rather than observed. No forge with Epic materials was recorded, so "Epic always produces Epic" follows from the ladder-and-ceiling rule holding for the three rarities that were. It is the natural reading, and it has not been seen directly.
The smelting bands and the refining rule come from a previous Annual Special and are not re-confirmed by the 2026 screenshots. Both 2026 readings — 8 hours at 2x and 4 hours at 5x — sit exactly on the old bands with every figure a third of what it was, which is what established that smelting is banded by speed rather than divided by it. The x10 figure of 2 hours follows from that ratio and has not been seen directly.