Random Rolling numbers maybe not so random?

I've tested this a few times and basically if i were to say roll 100d100/100 100 times the average should lay some where around 50-57 with 1 major exception getting an average of 42.81 or something to that effect i did not see a major variance of the average number rolled which at least imo should shift higher or lower im not saying this is a bug because i'm no mathematician... So in a general sense are these calculations correct?

Your sample size is waaaaaaaaaaaay too small I'm afraid, but never fear, here's data from a hundred thousand rolls: <a href="https://app.roll20.net/home/quantum" rel="nofollow">https://app.roll20.net/home/quantum</a> Hope that helps allay your worries!

no got to about the same conclusion :P lulz anyways yeah i figured the rolls were way too even there should be no one definitive average if your trying to compute a truley random number if the "if it ain't bouncing it ain't random enough" to quote a guy i know

100d100 is not a statistically significant set to derive a statistical average from. You would need to roll thousands (or even millions) of rolls to show that the dice roller is not rolling average. I rolled 10 sets of 100d100/100 and then averaged the sets. I got 50.862 which is pretty close to the average of 50.5. This shows that 100d100 is not a statistically significant set. You need to roll far far more dice. Even 1000dice is too few, a single group of 100d100 can throw the results off, at 900 the average was 50.53, an extra 100d100 changed the results be 0.35 which is a large change in statistics. Until the overall set size is large enough where a single group of rolls does not significantly change the results then the set size is too small. Keep rolling dice until the average no longer changes and you have found the average. In addition to the quantum roller upgrade our base dice roller was put through major paces about a year ago: <a href="https://app.roll20.net/forum/post/63487/from-the-s" rel="nofollow">https://app.roll20.net/forum/post/63487/from-the-s</a>...

Waaaaait are you saying you don't believe it is random enough because the AVERAGE stays in the 48-57 range? "If it ain't bouncing it ain't random" does not apply to something averaged across a million rolls. That's sort of the point of "average". I would suggest taking a look at the meaning of mean, average and median on Wikipedia if you haven't already done so as this might help clarify things.

From what I understand about computing a random number is that ideally, the averages of the computed numbers should generally shift quite frequently. also then explain to me if it isn't sufficient to test the algorithm then why did I make generally the same exact find (btw the shift should greatly increase if you use a larger frequency of numbers). in any case my question hasn't been answered if the numbers are shifting greatly then is any random roll not really random. you can pretty much guess what number your more likely to get should have made my question abundantly more clear lol and I apologize.

Then again statistically speaking you should on average get a number in the center of the frequency of numbers you are using this further proves my point when playing a rpg you should try and shoot for an above average stat and skill with all of your characteristics. This making your character OP I guess which actually makes the bard class the most powerful class in pathfinder imo.

Ok Average Intelligence. This thread is drifting away from stated question about the roll20 dice engine being random enough. We are not here to discuss game system mechanics as you posted. Gauss and Alan H both pointed out that your random sample was not big enough to conclusively state one way or another. They also pointed you to specific pages that will answer your question. If you are going to pull in specific game systems to proclaim what class and stat is best then please continue your discussion at Wotc forums or a redit subforum. Otherwise I will be forced to move this thread to off-topic.

I was asking about the computed random number generation the original question is about exactly what the title is about which my question is still on topic as it pertains to how we perceive these results ie not asking if it is working as intended but rather if the algorithm can calculate outside of it's averaged number base ie rolling an average of 8 on a 20 sided die a million times or rolling an average of 12 a million times. but feel free to move it i apologize again based on my reasoning i thought it might be worth bringing up in the general discussion forum.

We encourage discussions about the mechanics of roll20 itself but when the discussion about game systems and classes being over powered that type of discussion belongs to a more general rpg forum such as at pazio, wotc, redit, etc... Please keep it focused on the computed random number generation and we are all happy.

Average Intelligence said: From what I understand about computing a random number is that ideally, the averages of the computed numbers should generally shift quite frequently. If Im understanding what your saying, then its the exact opposite of that. With truly random numbers within a range, the average over enormous iterations should not shift that much at all and always be within +/- 1 of what the mid number is (for example a d20 should be mean average of within 10.0 and 11 as middle is 10.5 (20+1 / 2, 19+2 / 2..etc), more ideally only a few decimals from 10.5 either way, the more you roll the tighter that variation should become). Say random 1-100, rolled a million times, should give roughly the same average as a second set of 1-100 a million times. Then a 3rd set rolled 2 million times should still fall within that same mean range as average. (which should always be roughly around 50ish). I think when people see patterns they are confusing actual chance of individual results, with our human desire to see patterns based on past results for current randomness. You roll a d4 5000 times and get 5000 1's You still technically have a 1 in 4 chance of a 501st 1. Like flipping a coin, say you got 400 heads in a row, whats the chances of getting another head on next flip? 50/50. sure the chances of getting that MANY heads in a row predicted ahead of time is compounded up to a very high chance of NOT getting 401 in a row. but the flip itself isn't dependant on past results. its still 50/50. Like I cant remember where I saw it(i think it was vsauce), but have a human write a string of random 10000 numbers on a board, and a truly random number next to it, and any mathematician can point out the truly random one just by looking not knowing which is which. Because the human one will try to 'appear' random by mixing up the numbers..no one would have part of it say as "..12544444446.." or "...123456..." as we don't consider than many 4's in a row to 'look' random, or that 123456 is random. while the 10000 truly randomly generated ones will have sections like that. (or in other words, whole patterns of parts of it, that someone wouldn't consider random at all) edit: This sums up what I was trying to say far better than me. it got a bit of fluff unrelated to this discussion..but good bits that do. a good point is around 1:25 <a href="https://www.youtube.com/watch?v=9rIy0xY99a0#t=85" rel="nofollow">https://www.youtube.com/watch?v=9rIy0xY99a0#t=85</a>

Average Intelligence said: From what I understand about computing a random number is that ideally, the averages of the computed numbers should generally shift quite frequently. also then explain to me if it isn't sufficient to test the algorithm then why did I make generally the same exact find (btw the shift should greatly increase if you use a larger frequency of numbers). Humans are really bad at understanding randomness and really bad at understanding statistics. Part of the reason for that is because humans are really good at spotting patterns... even when there is no pattern there. (Think spotting animal shapes in the clouds.) Random number generation is complex subject in computer science, and a lot of really smart people have put a lot of time into it. I'm not sure what you mean by "shift" but it does appear that you are having some difficulty with some of the concepts surrounding randomness. If you truly want to test Roll20's randomness, I would recommend ENT . ENT will give you the entropy of the data set, the chi-squared distribution, the arithmetic mean, a Monte Carlo approximation of Pi using the values of the dataset (and the error of that approximation), and the serial correlation coefficient (how much each byte depends on the previous byte). A good RNG will produce something highly entropic, exceeds the Chi-squared distribution neither frequently (&gt;90%) nor infrequently (&lt;10%), has an arithmetic mean close to 127.5 for byte processing or 0.5 for bit processing, has a low error for the approximation of Pi, and has a correlation coefficient close to 0. (Note that the input for ENT needs to be either bytes or bits, not text.) As a closing note, I feel it's important to point out that any roll which is accompanied by the little atom icon is using a physical source of randomness.