Dice

If there's one thing Shapeways is suited to, it's speciality dice.
There's a whole section for them, and they often appear on the front page.

So when I started to design my own I was slightly surprised to find no special interest thread. I just don't know why not, because I'm sure there are things we could productively talk about.

For myself, I'm particularly concerned about the fairness of the dice. How are people testing for bias? When print orientation matters, do you find that they get positioned appropriately?


I will try to start and get the discussion rolling, with what I do for bias testing.

Basically the process I have at the moment is to buy a single die, and roll it 100*n times, where n is the number of sides.
I try to roll it in a manner which will highlight any issues - for example a roller should roll and come to a stop without hitting anything. And I try to define miss-rolls strictly and declare them without looking at the number (for metal dice the sound is a good indicator).
Then a chi-square test indicates whether there is any evidence that the die is not fair, and I add a comment to that effect to the listing.

But there are some things I don't like about this:
[list type=square]
[*] Only a single instance of the design is tested.
[*] Being impartial on the rolls is really hard. I think I'm managing, but who would trust that?
[*] The test doesn't have as much power as I'd like.
[*] Sometimes (one time in 20) the test will reject even a perfectly fair die.
[*] I'm not sure my report actually makes sense to people[/list]
Overall I'm a bit concerned that reporting the data may put customers off rather than reassure them.

Obviously I could buy more dice and test them for many more throws, but even what I'm doing is a job of work, and that wouldn't solve the other issues.
And I should also say that I appreciate that this is also a deal more work for a D20 than a D6.


Links:
chi-square test (basic description with probability table)
More on die fairness testing (Useful introduction, and also has something on a test with more power, suitable for larger dice.)
On the power of the chi-square test (How discrimination relates to number of trials, and how a negative result doesn't prove fairness.)

When it comes to bias testing, I've decided it's probably not worth it. As long as the die is designed fairly, which in not hard because we're essentially making dice out of math when we 3d model, and prints accurately (which can be inspected visually upon receiving test prints), we have no reason to suspect that our dice would be biased.

I fall back on this idea: A good, random roll of a badly biased die is far more random than a poor roll of an unbiased die. Even if the variance in the product is a few milligrams here and there, and even if that randomly works out to a slight bias in theory, how you throw the die will have far more influence on the result and its fairness than any physical factor of the die itself.

We could design dice unfairly, and that would probably yield a bias that matters. We could also modify dice that are designed fairly to be unfair after printing in some cases, but that's just cheating so it's beyond the scope of this discussion. The only other major factor that could introduce bias is large flaws with the print. Overall, shapeways' printers are pretty awesome at not producing these flaws. I haven't found orientation to matter at all. The only real issues pop up with delicate parts that can be unevenly worn during the polishing process or printed with flaws due to the level of detail necessary. It is thus best to keep dice relatively robust and simple, at least at any points that will contact the table.

Because of all of this, a pretty general rule holds true: Provided that the die is designed mathematically to be fair, the only way that it will print significantly unfair is if the printers can't handle the design. If the printers can't handle the design, it will look terrible anyway, so generally any good/visually appealing product won't have a problem.

Also, I don't really want to roll every dice set I design a cumulative 7000 times.

If customers are put off by the lack of a statistical test, that first bit of wisdom is usually what I tell them: How you throw it matters far more than if it's slightly biased, and it's not more than VERY slightly biased. You probably have a D20 in your collection already that's slightly unevenly weighted, but it probably doesn't show bias.

I don't think the quality of 3D printed dice are anywhere close to product that is supplied to the gaming industry, where even the paint filling the pips is part of the balancing equation. Any bowing or density changes within the 3D print will affect balance which means to get fair dice you probably need to adjust them after manufacture. But that also means you need ways of accurately testing balance in some sort of gizmo, and that gizmo may not work on different types of dice. So I think it's difficult to make any sort of fairness claims for balance other than it being inherent in the original design file and not in the actual printed product.

>So I think it's difficult to make any sort of fairness claims for balance other than it being inherent in the original design file and not in the actual printed product.

Yeah, that's pretty much how I feel. Most of the time, it's not really testable in a way that would satisfy a true pedant.

And actually, interestingly enough, the gaming industry gets it wrong quite a lot, especially on opaque dice: https://www.youtube.com/watch?v=VI3N4Qg-JZM

If anything, ours are likely more uniform in internal density due to the nature of the sls process.

I saw an interesting show on TV about manufacturing dice but I can't remember what show. I think they were showing manufacture at this place:

http://www.midwestgamesupply.com/dice_manufacturing.htm

Just having faces that are different sizes can cause problems let alone weight problems. When you're dealing with microns for dimensions it's definitely a precision part. Getting a comparable 3D printed die seems to point towards good measurement techniques and some post process machining to get it right. It would be interesting to see if that company could start with a block of 3D printed plastic, run it through their process, and see if they end up with something decent. I think in the real world the precision of manufacture is intended to eliminate the need for actual testing by rolling dice umpteen times. Still it's an interesting problem. Maybe there's a novel 3D printed balance tester that can be designed to measure dice!

Those are casino dice, MrNibbles. Other dice, what you get with boardgames, etc, are injected into molds. And in many cases, paint applied roughly and then cleaned, as well as the edges smoothed, in tumblers. The injection is not of the best quality, so they deform when cooling, or even have internal bubbles (even transparent are sold with them). OTOH, SLS has directional properties, for example, and orientation is not selectable. Polishing adds the same issues than as with injected ones.

If you want threads about dice, look in the archives. They were a common product at the start, but saturation, convenience (you want a material that does not break or breaks the table, that does not bounce like mad or just stops at landing, etc) and price probably made them fade. The designers that did many of them are not very active latelly. Some products even jumped into industrial manufacturing.

I also did some just for the fun of it and to give to friends even if they can be kind of crappy right out of the box without post-processing or painting. They are certainly not casino quality dice! But the initial post was primarily about bias, and that relates specifically to the precision of manufacture and properties of materials used. So yes, not so great with 3D printing available today.

As far as measuring bias I see that some people float them in a saltwater solution to see if they are weight biased. That can be difficult with commercial dice (board game or casino grade) but it might work better with 3D printed ones, at least for some light weight materials like SF.

I got a die I designed - practically as simple a form as is possible - which had a significant print issue. Perhaps half a millimeter wasn't printed (so maybe 4-to-6 0.1mm steel layers), at an angle to one of the faces. So the die had an extra flat surface. This meant that for one value the die would come to rest at about 15 degrees to that desired. It took me some time to work out what the problem was and convince myself it was definitely a printing error rather than my own mistake in the model or an issue with resolution (it's definitely multi-layer not sub-layer positioning).
I didn't complain; maybe I should have. It would be hard to get photos.
At least it seems like a rare occurance.
But I think all possibility of this could be avoided by printing orientation, in this instance at least.

Other than that, I have some sympathy towards the view that these arn't casino-quality dice and can be considered at least as good or better than most home-gaming dice they'll be replacing.

I actually don't know where to find the archives systematically. Occasionally I see a pdf from shapeways forums in a google search, but I can't find them internally.

This saddens me, because I'm just forming up the listings on my own dice.
Are they all really so done and dusted?

>I didn't complain; maybe I should have.

Actually, yeah, if it's not a flaw in your design, it may be a (relatively rare) printer/software glitch. Definitely worth bringing up with CS, even if just to let them know it might be happening.

>Other than that, I have some sympathy towards the view that these arn't casino-quality dice and can be considered at least as good or better than most home-gaming dice they'll be replacing.

This, a thousand times. I don't think a single one of us would ever try to claim that what we're selling is casino-level quality in terms of fairness. We would be fools to make that claim.

I wonder if anyone actually has thought to put that much precision into manufacturing of dice denominations other than D6. Then again, there would be diminishing returns. The difficulty of even intentionally rigging a die for a single result is directly proportional to the number of sides (as long as the number layouts are standard).

That casino die place fabricates to a tolerance of 0.0002 inches, or about 0.005 millimeters!

https://www.midwestgamesupply.com/dice_specifications.htm


My beef with printed plastic dice (like SF plastics) is that they seem much too light even though they should have more accuracy than stainless versions and be easier to adjust with sanding or other post processing. I had toyed with an idea to segment the insides into 6 sections to allow filling with something heavier, like tiny metal spheres and resin or something. I never actually got around to filling them so maybe it's high time I start doing some experimentation. I'm still not sure what materials to use...



Even though the "wave" of dice has swept through Shapeways long ago that doesn't mean there's nothing left to try in the future as new printers are introduced or designers come up with new ideas.