ROLZ

Tavern - Curiosity: Dice Distribution

2

Curiosity: Dice Distribution

chadwik66 · 16. Aug 2015 04:02 by SilverBen · 5

Hi Everyone,

We're finding that rolz.org is GREAT for online rolling in our APBA baseball league. Out of curiosity, has anyone posted on the distribution of dice rolls before? I've looked and been unable to find anything that shows large sample sizes, graphs or anything of the like.

I'm kicking around the idea of rolling a ton of dice and doing some analysis, but would love to read up on anything else like this that's already been done.

Thanks!

  • Chad
1 · by udo · 06. Aug 2015 12:01 · rolls: 1 · changed: 07. Aug 2015 16:31
Rolls in this post: #1 1000d6 count=1,2,3,4,5,6 = 3488 1:87 2:213 5:208 3:203 4:204 6:85  

Hey Chad, welcome and thanks for chiming in. Let me start off by saying that rolz.org is not intended for people who roll as part of a serious competition. It's designed for groups of friends engaging in pen&paper or forum-based roleplaying. If you use it for anything else, you're on your own.

The second caveat, and people write to me about this frequently: it is a random number generator. Rolz does not, and will never, generate uniformly distributed results. It generates random results. However, we don't use "quantum randomness" or any of that overblown bullshit. Instead, Rolz uses the same random number generator that operating systems use for cryptography, so in theory things should indeed be as random as is feasible.

You can do your own "study" by rolling things like

1000d6 count=1,2,3,4,5,6 = 3488 1:87 2:213 5:208 3:203 4:204 6:85

over and over again. But the main verdict is and remains: if you use this site for competitions, bitcoin raffles, or similar games, you deserve anything that happens to you. I don't actively screen dice rooms for actual gambling, but I will ban rooms and people who harass me about getting scammed or cause any kind of drama that forces me to police user activity.

That being said, you are welcome to host your league game here, but be aware of the limitations.

1 · by chadwik66 · 06. Aug 2015 15:54

Awesome, thank you! Rolz has been great for our (non-gambling, fun only) league since AIM Express has stopped supporting dice rolls. Thank you for such a great service and quick response!

  • Chad
1 · by chadwik66 · 07. Aug 2015 15:58

As an update to the rolz community...some analysis around distribution of results can be found at http://chadwik66.blogspot.com/2015/08/on-even-distribution-of-dice-rolls.html

1 · by udo · 07. Aug 2015 16:28 · rolls: 2 · changed: 07. Aug 2015 16:41
Rolls in this post: #1 10000d6 count=1,2,3,4,5,6 = 35142 3:1640 6:1697 4:1656 1:1644 2:1681 5:1682   #2 1000d6 count=1,2,3,4,5,6 = 3531 1:160 2:164 3:159 6:163 4:182 5:172  

Thanks for your research, Chad. I do believe though that an expected bucket size of 33 results (1200 samples stretched out over 36 buckets) is probably not going to yield statistically relevant results.

If I do a lot of rolls (here just 1d6 to keep it simple), results should be more uniform:

10000d6 count=1,2,3,4,5,6 = 35142 3:1640 6:1697 4:1656 1:1644 2:1681 5:1682

^ In theory, I would expect the result with two d6 to yield similar results - but you can see the numbers we're getting are pretty close to the expected 1,666.

I'm willing to entertain the possibility that there is something wrong with the random number generator, but I don't see it yet. I would like to say though that my own example in the post above this one does present some skewing away from 1 and 6, however, I can't seem to replicate the effect right now:

1000d6 count=1,2,3,4,5,6 = 3531 1:160 2:164 3:159 6:163 4:182 5:172

Barring anything deeper, I'm going to chalk this up to crazy outliers, but as I said I'm open to further investigation.

Quoting chad:
TL;DR: Use AIM or Google Hangout to roll dice, they seem legit.

To take you up on the general tone of the blog post, I agree this is probably not the site for you. Especially if you can get the same (or better?) from AIM or Google Hangouts, there is no apparent reason why you should use rolz in the first place.

1 · by SilverBen · 16. Aug 2015 04:02

Try anydice.com

I believe it has the graphs and testing mechanisms you are looking for.

21.9
  ↑ top