The Forge Calculator – Crafting Odds and Ore Values

average multiplier = Σ(count × multiplier) ÷ Σ(count) · trait on when count ÷ total ≥ 10%

The starter pile. Twenty Iron (0.35×) with five Gold (0.65×). Weighted sum = 20 × 0.35 + 5 × 0.65 = 10.25, over 25 ores — an average of 0.41×, and a raw pile worth 20 × $5.25 + 5 × $19.50 = $202.50 before it ever melts.

The dilution lesson. One Arcane Crystal — the rarest roll in this table at 1 in 100,000, with a 7.5× multiplier — buried in nine Stone: (7.5 + 9 × 0.2) ÷ 10 = 0.93×. The crystal keeps exactly 10.0% of the pile, and the pile still crafts like slightly-posh rock. Rarity does not leak into the average; only share does.

The trait edge. One Demonite among nine Iron is 1 ÷ 10 = 10.0% of the crucible, so its burn trait is live. Add a single extra Iron and the same Demonite becomes 1 ÷ 11 = 9.1% — present in the item, absent from its effects.

The long grind. Darkryte rolls at 1 in 5,555. Mine exactly 5,555 rocks and the chance of having found at least one is 1 − (1 − 1/5,555)^5,555 = 63.2% — not a certainty. A coin-flip 50% shot takes 3,851 rocks; a 90% shot takes 12,790.

The default: Gold does the lifting. Twenty Iron at 0.35× and five Gold at 0.65× give a weighted sum of 20 × 0.35 + 5 × 0.65 = 10.25, divided by 25 ores for an average of 0.41×, and the raw pile would sell for $202.50 (20 × $5.25 + 5 × $19.50). Gold is 20% of the pile but contributes 3.25 of the 10.25 weighted sum — nearly a third of the average from a fifth of the ores — which is why the page’s advice is to grow a pile with mid-tier ores first and reach for the rarest ones only when they can hold at least a tenth of the mix. Every common ore stirred in pulls the average back toward its own low multiplier, and the average, not the best ore, is what the forged item inherits.

The trait that switches on. Replace the five Gold with five Obsidian (2.35×, $35.25 each, trait +30% Vitality) and the pile becomes 20 × 0.35 + 5 × 2.35 = 18.75, an average of 0.75×, with a raw sell value of $281.25 — and the trait switches on, because Obsidian is 20.0% of the mix, past the 10% line. The average nearly doubles and a trait appears, from a single substitution, because Obsidian is both a far stronger multiplier and present in enough quantity to count. This is the pile the default is gesturing at when it says mid-tier ores first: a pile built around one strong ore at a real share beats a pile sprinkled with rarities at token shares.

The trait that stays off. Keep the Obsidian but cut it to a single ore — 24 Iron and 1 Obsidian — and the average falls to 0.43× (24 × 0.35 + 1 × 2.35 = 10.75, divided by 25), the sell value falls to $161.25, and the trait stays off: Obsidian is 4.0% of the mix, under the 10% line, so it still counts toward the average but its trait does nothing. One rare ore drowned in filler is just expensive filler — the page’s own phrase, and the arithmetic proves it: the single Obsidian lifts the average from 0.35× to 0.43× (+0.08×) while contributing a trait that never fires. If the trait is the reason for the ore, the ore needs a tenth of the pile; if the average is the reason, the ore needs to be strong enough to move it.

The grind behind the pile. A pile is only as good as the ores you can actually mine, and the grind calculator prices that in luck, not effort. Iron rolls at 1 in 5 per rock: 5 rocks give a 67.2% chance of at least one Iron, with 4 rocks for a 50% shot, 11 for 90% and 14 for 95% — and after 5 rocks there is still a 32.8% chance Iron has not shown up at all. Gold rolls at 1 in 16: 5 rocks give only a 27.6% chance, 20 rocks give 72.5%, and the budget rows stretch to 11 rocks for 50%, 36 for 90% and 47 for 95%. Expected rocks — about 5 for Iron, about 16 for Gold — are averages, not promises, and the 50/90/95% rows are the honest way to budget a grind, because “expected” is the number you beat half the time and miss half the time.

Each row is a pile of 25 ores unless the row says otherwise; multipliers and sell values are the page’s own data, and the trait line is 10% of the pile.

The 25-Iron row is arithmetic from the page’s own data (25 × 0.35 = 8.75; 25 × $5.25 = $131.25), included as the baseline every other row is measured against. Read the table down the average column and the story is plain: the trait-on pile is the only one that escapes the 0.4× band, and it does so because one strong ore holds a fifth of the pile, not because the pile contains a rare name.

What the category table actually says. For the default pile of 25 ores, the weapons table reads: Dagger at its best-known count (3 ores, 100%); Straight Sword at its best-known count (6 ores, 86%, first appearing at 4 ores with 14%); Gauntlet at its best-known count (9 ores, 65%, first at 7 with 20%); Katana at its best-known count (12 ores, 72%, first at 9 with 10%); Great Sword at its best-known count (16 ores, 69%, first at 12 with 3%); Great Axe at its best-known count (22 ores, 67%, first at 16 with 1%); and Colossal Sword in range between its published points (21 ores at 2%, best-known 50 ores at 70%). The pattern across every category is the same: a long, flat start — single-digit chances at the first-appearance count — and then a steep climb to a best-known count in the high sixties to eighties percent. A pile between the two points is “in range”, and the honest statement about it is that its exact chance is not published: testers publish the chance at the two anchor counts, and chances at counts between them are not published per ore count. Inside the game, the crucible shows the exact live percentage for the pile you loaded, so the anchors are for planning and the crucible is for confirming.

Why the anchors are trusted, and where they stop. The anchors come from ProGameGuides’ tested crafting table, updated December 2025, cross-checked in two guides — the page says so, and the cross-check matters because category chances are the kind of figure that gets copied from wiki to wiki until nobody remembers who tested it. The page’s discipline is visible in what it does not do: it does not interpolate between the anchors into a smooth curve, because no tester published the intermediate points, and a curve fitted to two points would be invented data wearing a tested label. The same discipline appears in the ore data: 42 ores with multipliers from 0.20× (Stone) to 7.50× (Arcane Crystal), and the one known disagreement stated rather than resolved — Eye Ore’s sell price is $37.50 on the wiki against $60 on TechWiser, and the page uses the wiki figure everywhere, including the calculator, so its pile values are consistent even where the sources disagree.

The traited ores, and the tenth that wakes them. Most ores carry no trait, and the ones that do are worth listing because the 10% line decides whether they matter: Poopite (Bad Smell, a poison aura at low health), Obsidian (+30% Vitality), Rivalite (+20% Critical Strike Chance), Uranium (armor aura dealing area damage around you), Mythril (+15% Vitality), Eye Ore (−10% Health, +15% Damage), Fireite and Demonite (burn damage on hit), Magmaite (explosions on hit), Lightite (speed boost on armor) and Darkryte (Shadow dodge). Every one of them follows the same rule proven in the variation table: at or above a tenth of the pile the trait switches on, below it the ore still counts toward the average and nothing else. A pile built for a trait is therefore a pile built around a share, not a name — Darkryte at 6.30× is the second-strongest multiplier on the page, but a single Darkryte in a 25-ore pile is 4% of the mix, and its Shadow dodge stays asleep.

The one idea, applied to a real pile. The page’s summary line is that a forge is a weighted average wearing a furnace costume, and the default pile is the proof: Gold, the best ore in the pile, is 20% of it, yet the average is 0.41× — closer to Iron’s 0.35× than to Gold’s 0.65×, because the average is weighted by count and Iron outnumbers Gold four to one. The most expensive mistake in the game is “best ore in a big pile of junk”, and the calculator exists to make that mistake visible before a single ore is at risk: build the pile on paper, check every traited ore against the 10% line, read the category anchors for the count you actually have, and only then commit the ores. The forge never adds multipliers together, and neither should your plan.