Computer Dice rollers

[palleomortis]

Call me an idiot, but I never really liked the idea of a computer "randomly" rolling for you. It seems to me, that the outcome would be one of two things.

1) The rolls would have been pre-generated based on, or copied from, real dice rolls. Wich, to me, seems un-natural to have a premade list of rolls, for the fact that it would simply either repeat itself, or be completely non-sensical.

2) The computer that would "randomly" generate your rolls would be set to a programe dealing with statistics, perhaps from time to time giving you a seemingly random generation, but in the long run only providing numbers that match the probablility of their outcome, making it either TOO random (That is to say that things such as the ever joyful two natural 20's in a row would become exceedingly rare as they don't match with the true probability), or simply a near predictable set of outcomes, that is to say that if you rolled a natural 20, you could count on not hitting another the next roll.



So is this just a fear of technology or hatred of it, or is there somthing there? How can you explain to me or show me how a program truly is random?

 

[kenobi65]

If it makes you feel any better, odds are that your dice aren't perfectly random, either. Inperfections in the manufacturing process, materials, molds, etc. likely mean that there are slight imperfections in your dice -- slight enough that, in most cases, it doesn't materially affect how often a "1" versus a "20" come up.

Someone more versed in computers than I will need to speak about how random number generators work, but I think it's closer to your second example than your first. I do recall reading that some random-number generators are better (i.e., more truly random) than others.

So, the real question is, is the amount of "non-randomness" in a random-number generator more biasing than the "non-randomness" that comes from using a not-perfectly-random die?

 

[Thornir Alekeg]

The formulae used in computerized random number generators are usually good enough that the fact that it isn't truly random is irrelevant. Here is a link to a simple article about how they work: Link

 

[TheLe]

Call me an idiot, but I never really liked the idea of a computer "randomly" rolling for you. It seems to me, that the outcome would be one of two things.

1) The rolls would have been pre-generated based on, or copied from, real dice rolls. Wich, to me, seems un-natural to have a premade list of rolls, for the fact that it would simply either repeat itself, or be completely non-sensical.

2) The computer that would "randomly" generate your rolls would be set to a programe dealing with statistics, perhaps from time to time giving you a seemingly random generation, but in the long run only providing numbers that match the probablility of their outcome, making it either TOO random (That is to say that things such as the ever joyful two natural 20's in a row would become exceedingly rare as they don't match with the true probability), or simply a near predictable set of outcomes, that is to say that if you rolled a natural 20, you could count on not hitting another the next roll.



So is this just a fear of technology or hatred of it, or is there somthing there? How can you explain to me or show me how a program truly is random?

I'd be willing to put it to a test. Would you be willing to roll 1d10, one hundred times and mark down the results? I'd be happy to write a generator to do the exact same thing... then we can compare.

`Le

 

[Gundark]

Call me an idiot,

You're an idiot.



Hey you said to.

 

[kenobi65]

I'd be willing to put it to a test. Would you be willing to roll 1d10, one hundred times and mark down the results? I'd be happy to write a generator to do the exact same thing... then we can compare.

To generate a big enough sample, you may want to do a few more. 1d10, rolled a hundred times, would only generate, on average, 10 occurances for each number...and, getting 8 or 13 for a particular number will look "skewed", but in fact, just be the result of random chance (and not having a big enough sample of results).

Try rolling it 500 or 1000 times.

 

[Slife]

To generate a big enough sample, you may want to do a few more. 1d10, rolled a hundred times, would only generate, on average, 10 occurances for each number...and, getting 8 or 13 for a particular number will look "skewed", but in fact, just be the result of random chance (and not having a big enough sample of results).

Try rolling it 500 or 1000 times.

Nah, for a truly decent sample you should roll it over nine thousand times.

 

[TheLe]

To generate a big enough sample, you may want to do a few more. 1d10, rolled a hundred times, would only generate, on average, 10 occurances for each number...and, getting 8 or 13 for a particular number will look "skewed", but in fact, just be the result of random chance (and not having a big enough sample of results).

Try rolling it 500 or 1000 times.

Well, writing the program is easy -- I can set it to as many iterations as I want. I was trying to make it easier for people rolling.

However, I think this would still be a great test.

How about we get 10 people to roll 1d10 fifty times each? Shouldn't that be feasible?

If you like, we can go with a 1d20 instead.


`Le

 

[TheLe]

How about I put my time where my mouth is?

Here you go:

http://www.TheLe.com/temp/RandomNumberGenerator.asp

Note that I reseed before each randomization attempt.

As far as I can tell, the numbers seem to work out well.

~Le

 

[kenobi65]

How about we get 10 people to roll 1d10 fifty times each? Shouldn't that be feasible?

Except that now you're comparing one random-number generator against the aggregate of 10 different d10s, and 10 different rolling styles.