# Why your d20 only rolls low.. explained !

Started by Fkewl, Oct 31 2008 01:18 PM

35 replies to this topic

### #3

Posted 31 October 2008 - 02:02 PM

I didn't get all the way through the second one, so I don't know if he put up any actual numbers, but I'm not convinced. Yeah, a die with one rounded edge will roll strangely, but what if they're all pretty well rounded? I imagine that the effects, for the purposesof most gamers, are too small to notice unless someone is deliberately taking advantage of them.

Has anyone ever tested their dice for randomness? How many times would you have to roll a die to verify that it is sufficiently fair? A thousand times?

Has anyone ever tested their dice for randomness? How many times would you have to roll a die to verify that it is sufficiently fair? A thousand times?

### #8

Posted 02 November 2008 - 12:52 AM

Yeah, I'm not sure that I buy that the level of randomness in most dice is quite as low as he makes out. But the faster stop of his dice when rolled side-by-side with other dice appeals to me, since my dice tend to roll all the way across the table and onto the floor.

### #9

Posted 03 November 2008 - 11:28 AM

Sure, some of it's marketing hype to bolster is claims

but for me, it's the 10 (or 20?) dice all piled up with 20's touching and the other with 1's touching and both piles didn't have the same height.

I assume manufacturers now have a better grip on to create them and shouldn't have the as big of an error, but we never now. Buy cheap, get cheap.. that's always gonna be.

Now is it cheaper just to buy enough dice so you have a good one, or just buy a good set but for more moola ?

I'm seriously thinking of buying his dice for it's durability mostly, and some boasting too

but for me, it's the 10 (or 20?) dice all piled up with 20's touching and the other with 1's touching and both piles didn't have the same height.

I assume manufacturers now have a better grip on to create them and shouldn't have the as big of an error, but we never now. Buy cheap, get cheap.. that's always gonna be.

Now is it cheaper just to buy enough dice so you have a good one, or just buy a good set but for more moola ?

I'm seriously thinking of buying his dice for it's durability mostly, and some boasting too

### #10

Posted 03 November 2008 - 06:28 PM

I have a Gamescience die that players long ago dubbed the 'vorpal die'. Much like a good Gamescience die It has sharp edges and a nub...and rolls 9,10,19 and 20 far too often (chi square test confirmed with 1% significance).

Warning: Statistical Content (please ignore if you know statistics)

The chi-squared test is, in this context, basically a test of equal probabilities. The number of rolls depends on how unfair a die is(and how many sides it has). To test the vorpal die I rolled 1000 times but it was so biased that I could have done it in well under half of that. Lesser imperfections could easily take more rolls.

If you have msft office then excel can conduct the test (chitest) on collected data and the answer is the likelihood that the die is fair. Statistics generally feels that 5% and 1% are good levels of proof. If the likelihood of the die being fair is below 5% then they would say that there is some confidence that the dice is unfair.

If your answer is not that low then you can keep collecting more observations or give up. However if you need many observations to detect unfairness that suggests that the die is not overly unfair.

Warning: Statistical Content (please ignore if you know statistics)

The chi-squared test is, in this context, basically a test of equal probabilities. The number of rolls depends on how unfair a die is(and how many sides it has). To test the vorpal die I rolled 1000 times but it was so biased that I could have done it in well under half of that. Lesser imperfections could easily take more rolls.

If you have msft office then excel can conduct the test (chitest) on collected data and the answer is the likelihood that the die is fair. Statistics generally feels that 5% and 1% are good levels of proof. If the likelihood of the die being fair is below 5% then they would say that there is some confidence that the dice is unfair.

If your answer is not that low then you can keep collecting more observations or give up. However if you need many observations to detect unfairness that suggests that the die is not overly unfair.

### #14

Posted 05 November 2008 - 05:37 PM

I might add, if you want TRUELY odds based dice (like dice should be, to be fair), check out the dice of game science. They are scientifically proven and guarenteed to roll correctly (meaning, to roll true odds). I will see if I can find a link for them.

### #19

Posted 08 November 2008 - 02:31 AM

I didn't run any chi square tests, (I've happily forgotten all of my high school statistics) but I'm pleased to report that my new gamescience d20, on an initial test of five rolls, rolled 20, 20, 16, 7, 20.

Very auspicious.

Su. SUspicious.

Meh, it just depends on your point of view.

### #20

Posted 27 November 2008 - 08:14 AM

It isn't possible to check a die with 5 rolls. You could get 5 20's, but that doesn't mean the die is unfair.

Best thing to use is a random number generator on a computer. This isn't truly random as it is an algorithm based on an initial number seed, but if you make the initial number seed a number based on the current time on your computer clock then it is about as truly and fairly random as you are ever going to get.

Saying that it would be nice to a small portable randomizer that does not use algorithms, but instead uses quantum probability. Then that would be truly random on a profound cosmic scale.

Best thing to use is a random number generator on a computer. This isn't truly random as it is an algorithm based on an initial number seed, but if you make the initial number seed a number based on the current time on your computer clock then it is about as truly and fairly random as you are ever going to get.

Saying that it would be nice to a small portable randomizer that does not use algorithms, but instead uses quantum probability. Then that would be truly random on a profound cosmic scale.

#### 0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users