Dice's dirty secret (how fair are they?)

Sorry if this is a hardy perennial and I have bad search-fu.

I have dice. I have dice in a multitude of shapes [1] and colours. But something that’s been bothering me for a while [2] is… are they any good?

Some manufacturers of polyhedral dice will boast of their precision, but they’re long on information about proxy outcomes (if you stack 20 of our d20s in a tube, they always come out to the same height, so they must be all the same) and manufacturing processes, short on actual facts about testing.

A few obsessives have come up with arrangements normally involving cups that shake the dice and OCR of the results, so there’s some interesting data about individual types.

Particularly I have grave suspicions about the d20, which is unfortunate because both 3e and 5e D&D care strongly not just about the average value but the chance of hitting particular values. I suspect, although this is sheer guesswork, it would be better to roll a normal d10 and a high/low die.

Casinos and backgammon fiends seem to have very good dice, but vexingly don’t want to make them in any other sizes.

We’ve been playing on roll20 during the plague, and well, computers can generate perfectly good random numbers on demand. Hooray! But IDK I want to go back to rolling bits of plastic on the table and onto the floor.

Trouble is, computers take up a lot of space on the actual gaming table, and lack a certain tactility.

First thought: perhaps there is a kind of dice which are more definitely known to be as fair as practically possible?

Second thought: a physical object with a computer in it. How about, IDK, a Raspberry Pi Zero with a speaker and a little screen attached to it, and a mechanical key numberpad attached with buttons saying things like ‘2d4’ and ‘1d12’. Hit the button and the screen displays the outcome while the speaker yells it out.

Bit of a fiddly mass of objects, and perhaps you want one or two switches (eg if I hit “3d6” do I want to be told “15” first or “6, 4, 5”?).

The next thing I discover is that there are “button box controllers”, USB devices mostly for flight and driving games which are what they sound like, a box with a bunch of chunky buttons and switches on top. A DIY button box controller is a practical project, and in particular something that gets suggested a lot is reusing old hard drive caddies because they’re hardy metal boxes which you own already. We get the tactility because we’re using actual buttons, and also now we can buy the buttons from electronics shops for pennies.

And now, of course, the Pi can live in the box, probably with a battery pack because it’s going to be pretty obnoxious having something on the gaming table with a power cable attached.

However… everyone’s got a smartphone, and they’re probably on the gaming table already. Existing dice rolling programs for smartphones aren’t really what are wanted (they try and offer you all the options, which is all very well for Roll20 when you’ve got an entire computer screen to look at and it’s all you’re meant to be looking at); what I want is something where I load “dungeonquest” and it shows me a screen with “d6, d10, d12, 4d10 and tell me the individual results not the total”, reads the results out [3], and there’s just a little UI element in the corner of the screen to go back to one of these crowded roll-arbitary-dice interfaces. This lacks tactility, but it’s a familiar device and a hell of a lot less work.

So… what do you think? If I want to take your dice away, what should I replace them with?

[1] Six.
[2] Thirty-five years.
[3] I have no idea if this would be obnoxious or helpful in practice but it seems worth giving it a go.

1 Like

Welcome back!

One thing you can do: take a glass of water, and dissolve salt in it until it’s dense enough to float the die. Then record the way up that it floats, many many times. Then you just do a chi-squared test to see if there’s a bias.

Personally I like actual physical dice and I just use 'em.

Electronic dice rollers that I’m aware of go back to the Dragonbone from the 1980s.

They’re a bit antique now.

If I have to use my phone for dice rolling, Dice Roller uses an actual physics engine (the same one that underpins TTS) to simulate actual dice.

You can set up groups of six preset rolls, and have as many stored groups as you like. There are lots of preset dice, with varying levels of readability. It’ll even use the device’s motion sensor if you like. The only thing it won’t do is read them out loud. (Not shown: FUDGE dice.)


I’ve seen the water test. I’m willing to believe it weeds out some bad dice, but I’m less convinced all the dice left are good (hm, is that true if I do the proper probability work afterwards, and would this take a prohibitive number of rolls?). I like dice too, but I’m kind of worried they’re doing me a disservice.

Physically simulating the roll of dice seems, ah, a curious approach, although I can see what’s going on there may satisfy the desire for tactility. I think if I was going to do that I’d start by randomising the die’s orientation so an RNG can do the real work.

1 Like

Actually fair numbers don’t seem fair to humans, and vice versa. Humans are broken. Upgrade now!


That I know, but ofc we’ve been getting fair numbers from roll20 for a year, and I suspect also numbers from a moderately biased die will look “non-random” to humans in that way.

1 Like

I enjoyed this article from Forbes on the subject. (and the links contained within.)

I’m very much of the opinion if you think a dice isn’t rolling correctly, then it isn’t rolling correctly.


Even distribution is boring. Dice bring random, not fairness.

Also: any system that relies on the result of a single die for output randomness is oppressive, in my book. Dice are best used in sets of 3+ (more the sides, the more dice you need to be rolling, at about a 1:2 ratio)


(Plus rolling a fistful of dice is more satisfying)


It’s hard to get away from this point.

When I roll some dice, I feel like I did a thing. It’s a tiny little event in the game. There’s usually some slight element of ritual to it.

When something/one else tells me some numbers I don’t feel like I was involved.


I think that’s what the shake-your-smartphone programs are trying to address, as is the idea of a physical box you lean over and press the ‘3d6’ button on (and the desire to make the buttons as chunky and satisfying as possible).

1 Like

It’s more expensive to hurl your smartphone across the room in disgust, though (although potentially a real windfall for the next person to move the fridge).


Are there different materials used in our favourite polyhedrons, that are more suited than others? I have resin, clear (ish) plastic, and metal dice. And am eyeing up some gemstone ones. Ive heard of tests where molded (resin or plastic) dice have been cut up and found to have voids/bubbles in them that introduce bias to the rolls (for good or ill). I’m sure I’ve owned such dice in the past and have had to consign them to the dice dungeon! Are more reliably cast materials (metal) or naturally solid ones (stone) less biased?

Hi @DangerousDave , welcome!

I’m not sure how much of this is related to Kickstarter style marketing. How have people found metal dice?

I find the really weighty metal dice put dents in the soft pine tabletops that flat-pack furniture places sell. One for people with really solid tables.

Some of my clear plastic dice you can see the bubbles so you know they’re not going to be accurate.

Although arguably a clear die you can see the bubbles in is better than an opaque one… :-/

Casino dice are clear so that you can see the lack of bubbles (or lead shot), and sharp-cornered – but as @denisbloodnok points out, only available in d6. (Play GURPS!) They’re also quite cheap as second hand, too worn for casino use but probably well within the tolerances for RPGs. I had a GM once who liked them, but we all found them very uncomfortable to handle.

I like metal dice for the hand feel, but as Tom says, they can damage tables – even rounded dice.

Many dice are made with “urea”, usually a urea-formaldehyde resin. I think that’s the cheaper option compared with petroleum plastic, but it’s been a while since I was reading about this.

If I were going to the trouble of building a machine for this, I’d use an explicitly noisy hardware source: avalanche breakdown in a Zener diode, a low-grade radioisotope, that sort of thing. Then do the standard “turn grotty randomness into useful random numbers” stuff that you can steal from the Linux kernel – indeed, given a Raspberry Pi, you might as well use the Linux kernel, feed it with your noise, and let it do the mangling.

I got my first set stainless steel dice many years ago (decade or more) from https://crystalcaste.com/. They have since changed the style of their dice to something I don’t particularly care for. I also have a set of anodized aluminum dice but the coloring is worn off in places (very minor but enough for me to notice).

If you use a dice tray, they don’t pose a hazard to gaming tables. However, I have found annoying scratches in them when I stored them in a dice bag by themselves.

My travelling dice set has copper-plated steel, plain steel, and haematite dice. Most are quite small, which limits their ability to damage tables. There are three large d6, which were made with very rounded corners, and don’t seem to do damage.

A friend has a sharp-edged iron d6, about 15mm on an edge. He made it himself as an engineering student. That does a lot of damage.

1 Like