The Anatomy of a Balanced DND Die

Every tabletop player has owned a die they quietly suspected. It always seemed to roll low, or clung to the same three faces, and eventually it got retired to the bottom of the bag. Usually the suspicion is superstition — but not always. A die is a small physics experiment, and every stage of its manufacture either preserves fairness or leaks it away. Here is the anatomy of a balanced DND die as we see it from the YSDICE factory floor: what geometry, core density, numerals, edges and finishing each contribute, and how to tell a fair die from a merely pretty one.

1. What "balanced" really means

A balanced die is one whose centre of mass sits at its geometric centre. That is the entire definition. When the two centres coincide, no face is heavier than any other, and the outcome of a roll is decided by the throw — not by the die. When they drift apart, even fractionally, the heavy side wants to face down and the die develops a bias toward the faces opposite it.

What makes this hard is that nothing in manufacturing is naturally symmetrical. Material flows unevenly into molds, bubbles rise before curing finishes, paint adds mass to one face and not another, and polishing removes material wherever the operator presses hardest. Balance is not a property a die has by default; it is the sum of a dozen small disciplines, each of which can be done carefully or carelessly. The rest of this article walks through them in the order they happen.

2. Geometry: why the D20 is the diva

The standard seven-piece set spans five geometries — tetrahedron, cube, octahedron, pentagonal trapezohedra, dodecahedron and icosahedron — and they do not tolerate error equally. A D6 is forgiving: six big faces, large angular differences between resting positions, so a small mass offset rarely changes an outcome. A D20 is the opposite. Twenty faces means each face "owns" only a small slice of the probability space, and the energy difference between settling on one face versus its neighbour is tiny. The same manufacturing flaw that would be invisible on a D6 becomes a measurable bias on a D20.

Geometry also dictates where trouble hides. Vertices are where mold halves meet and where flash — excess material at the parting line — forms. On a D4 the sharp vertices carry the die's identity; on a D20 a single unpolished sprue mark at one vertex is enough to nudge statistics over thousands of rolls. This is why we judge any new mold by its D20 first: if the icosahedron comes out clean and true, the rest of the family follows.

3. The core: density, voids and bubbles

Everything starts inside. A die can have perfect faces and crisp numerals and still be biased, because the flaw is buried where no eye can see it: an air pocket from casting, a bubble that rose during resin curing, or a dense swirl of pigment that settled to one side before the pour set. The centre of mass follows the material, and the material does not care what the surface looks like.

A die can only be as fair as its least-controlled process. The surface you can polish; the core you have to get right the first time.— a rule our casting team repeats to every new operator

Each material fights the void problem its own way. In die-cast zinc alloy, molten metal is injected under pressure, which collapses air pockets that would survive gravity casting — but gate placement and cooling rate still decide whether the casting is uniform. In hand-poured epoxy resin, the enemy is the bubble: pouring technique and patience during degassing determine whether the blank cures solid or hides a microscopic balloon under one face. Inclusions — glitter, mica flakes, a liquid-core capsule — raise the difficulty further, because anything suspended inside a die must be either neutrally distributed or centred, and that has to be engineered into the pour, not hoped for.

4. Numerals: where fairness meets paint

Numerals are subtractive and additive at once: a cavity is engraved into the face, then paint fills it back in. Both halves of that equation touch balance. The "20" removes more material than the "1", and the paint that fills it weighs something too. On a well-designed die the numeral depth is shallow and consistent, so the removed volume is negligible against the die's mass; on a badly designed one, deep gouged numerals on light material genuinely shift the centre of mass toward the low faces.

The fill matters as much as the cavity. We hand-fill numerals with baked enamel, Pantone-matched to the client's specification — the baking hardens the enamel so it stays flush and permanent in its recess instead of wearing proud or flaking out. Flush is the operative word: paint standing above the face changes how the die tumbles on that face; paint sunk too deep collects grime and reads badly across the table. And legibility is part of the anatomy too — a 6 and a 9 distinguished only by an underscore, placed consistently, oriented the same way on every die in the set. A die you have to squint at is a die that slows the table down, however fair it is.

5. Edges, faces and the five-grit truth

Edges decide how a die stops. Sharp edges bite the mat and halt the roll decisively — which is why sharp-edge dice feel "casino-like" — while rounded edges let the die tumble longer and settle gradually. Neither is fairer in principle; what matters is consistency. Twenty faces need twenty identical edge treatments, because one rounder-than-its-siblings edge behaves like a tiny ramp every time the die crosses it.

This is where hand finishing earns its keep. Our sharp-edge resin dice are hand-poured and then hand-sanded through five successive grits, each pass finer than the last, flattening every face truly flat and bringing the surface to optical clarity without rounding over the very edges that define the product. Machine tumbling — the fast way to polish rounded dice in volume — must be timed so that every die in the drum loses the same whisper of material. Over-tumble one batch and the whole run's dimensions drift; under-tumble and mold seams survive as functional (and visible) defects.

6. Four materials, four routes to fair

We run six production lines in-house at YSDICE, and each material reaches balance by a different road. Metal relies on pressure and mold precision; sharp-edge resin relies on patient hands; acrylic relies on the repeatability of injection molding, which is what makes it the economic choice at volume; gemstone relies entirely on the carver, because stone cannot be molded at all — every blank is ground, faceted and engraved by hand from solid mineral. Knowing which road your material takes tells you which questions to ask about it.

One structural note that surprises buyers: a solid die and a deliberately hollow one can both be fair. Our hollow metal dice work because the wall is engineered to uniform thickness — symmetry, not solidity, is what fairness actually requires.

7. Proving it: float tests and our two-stage inspection

The classic home check is the salt-water float test: dissolve enough salt to float a resin or acrylic die, spin it gently several times, and watch which face surfaces. Random faces mean a homogeneous die; the same face rising again and again means the opposite side is heavy. It is a genuinely useful test — with two caveats. It only works on materials light enough to float, so metal and stone are exempt, and it detects internal offset, not edge or face defects, so it is a screen rather than a verdict. The fuller picture comes from long roll logs, which is tedium best left to reviewers with camera rigs — and from stopping defects at the source.

At the factory, that source-control is statistical. Every order leaves under a two-stage inspection — once at the workshop that made them, once again before dispatch — with a photo report: a defined sample of the run is pulled and inspected — dimensions, faces, numerals, finish — and photographed, so you review the evidence before the goods ship rather than after they land. Behind that sit SGS material reports and EN71 / ASTM toy-safety testing, and third-party audits are welcome if your project needs independent eyes. Balance is a process property; the inspection regime exists to prove the process held.

8. Size and weight: how mass shapes the roll

Balance decides where a die stops; mass decides how it gets there. A standard 16mm resin or acrylic die weighs a few grams and skitters — long tumbles, plenty of bounces, an outcome thoroughly randomised by the journey. The same 16mm geometry in zinc alloy weighs roughly four times as much, and the roll changes character completely: one or two authoritative bounces, a short slide, done. Neither behaviour is fairer. A heavy die is not "more random" — the physics does not care — but it is less sensitive to a soft or uneven throwing surface, because its momentum flattens out the influence of table texture. What the extra mass absolutely demands is a dice tray. We tell every buyer of metal sets the same thing: the die will win any argument with a bare wooden table, and the table will keep the dents as evidence.

Scale amplifies everything else in this article. Mass grows with the cube of the dimension, so a 33mm D20 carries roughly eight times the material of a 16mm one, and a 50mm showpiece more than thirty times. An internal void that was statistically invisible in a 16mm die becomes a genuine bias in a 50mm one, simply because there is more wrong mass sitting further from the centre. That is why our oversize resin D20s and D100s are poured slower, rest longer before demolding, and get a dedicated inspection pass for internal clarity — a big die shows off its interior, and its interior has more room to hide problems. Hollow metal dice run the same logic in reverse: by removing the core entirely and holding the wall to a uniform engineered thickness, a 50mm-class die stays light enough to roll on a normal table while keeping its centre of mass exactly where geometry says it should be. Size, in other words, is not just a style decision. It quietly raises the manufacturing bar, and a factory quoting oversize work should be able to tell you precisely what it does differently at 33mm and above.

9. Three failures we caught — and what they teach

Process talk stays abstract until something fails, so here are three catches from our own floor — the kind of thing an inspection regime exists to find before a customer does.

The humid week. One summer run of sharp-edge resin cured through a stretch of heavy Dongguan humidity, and a fraction of the batch trapped micro-bubbles that the pour itself never showed. Routine float-testing in QC caught the tell — the same face surfacing again and again on a handful of pieces — and the sub-batch was quarantined and re-poured. The lesson for buyers: ask whether resin QC includes a homogeneity check at all. Bubbles do not announce themselves on the surface, and a factory that only inspects faces will ship them without ever knowing.

The long tumble. An acrylic run once spent too long in the polishing drum — a timing slip, nothing exotic. Every die came out beautifully smooth and roughly 0.15mm under dimension, with numeral edges softened just past our retained reference. AQL sampling flagged the dimensional drift, the run was rejected internally, and the mold went back into service for a replacement batch. The lesson: a golden sample — a signed-off reference die kept from your approved pre-production run — is the cheapest quality tool that exists. Insist your factory keeps one for your project; at YSDICE the archived mold and the golden sample live together, which is how a reorder placed a year later still matches the original run.

The vertex that came back. Metal dice are polished at the parting line before plating, but plating itself deposits metal, and it deposits preferentially on edges and points. One batch of D20s grew a barely-perceptible ridge at a single vertex during a heavy electroplating cycle. Pre-plating inspection had passed; post-plating inspection caught it; the affected pieces were re-polished and re-plated. The lesson generalises: inspection has to happen after the last process that adds or removes material, not before it. A QC photo taken at the wrong stage of the line proves very little.

10. What fairness costs on a quote — and what it shouldn't

Everything above sounds expensive, so let us put numbers on it. Balance discipline is baked into process time, not itemised as a surcharge — you will never see "degassing: $40" on a quotation. What you will see is the structure that pays for it. A full-custom design starts with a steel mold at $600–$2,000 one-time depending on complexity, with the fee refunded progressively as reorder volume accumulates. Full-custom minimums sit at 100 sets for metal, sharp-edge and rounded resin, and 500 sets for injection-moulded acrylic, because that is the volume at which a carefully-run line stops losing money on setup. Light customisation — your colours and a laser-engraved logo on proven stock molds — starts at just 10 sets with no tooling fee, and it is quietly the safest first rung precisely because of this article: a stock mold has already produced thousands of balanced dice, so its geometry is beyond argument before your order even begins.

Timelines carry the discipline too. Mold cutting takes 20–25 days; custom samples take 3–5 days after that; mass production runs 10–30 days by material — metal and acrylic 20–25, sharp-edge 15–20, rounded resin 25–30. Those spreads are not padding. Sharp-edge finishes faster than rounded resin because the five-grit hand-sanding replaces long tumbling cycles; rounded resin is slowest because cure and polish cannot be hurried without exactly the dimensional drift described in the previous section. When a competing quote beats these numbers dramatically, ask which stage got shortened — the honest answers are usually degassing time, sanding grits, or the inspection itself.

For the record, the numbers here are our own, not industry averages: YSDICE, also known as Dongguan Yushun Hardware Co., Ltd., is a registered dice manufacturer in Chang'an, Dongguan, and these are the figures we quote buyers every week. A fair die costs pennies more per unit than a careless one. A biased run costs a relaunch, a review crisis, or a Kickstarter apology update — which is why the cheap quote is so often the expensive one.

11. The checklist to send your factory

You do not need to become a metrologist to buy fair dice. You need to ask the questions that force the process into the open — because every discipline in this article leaves a paper trail if it actually happened.

On that last point: our samples take 3–5 days for custom designs, or within a week for stock designs with the cost refunded on your first order — deliberately fast, because a sample in your hand answers more questions than any specification sheet. A balanced die is not a marketing claim. It is geometry honoured, mass controlled, edges finished evenly and the whole thing inspected before it ships. Ask for each layer, and the fair dice find you.

Written by Queenie Founder & English-speaking project manager · Dongguan Yushun Hardware Co., Ltd. (YSDICE)

Queenie has cast, poured and shipped custom dice from Chang'an, Dongguan since 2018 — and still answers every project inquiry herself. Questions about your run? Write to queenie@ysdice.com or message us on WhatsApp.