How random is quantum roll

How random is quantum roll? 81 1s 68 2s 64 3s 72 4s 63 5s 53 6s 68 7s 54 8s 65 9s 60 10s 62 11s 53 12s 64 13s 50 14s 63 15s 69 16s 57 17s 71 18s 59 19s 23 20s This is the results of tallying up rolls for a game I am running. Every session seems pretty inundated with ones, and 20 the highest number shows up at least half as much as the next number up there, and only 1/4 the number of times that one does

I won't deny it. We also tend to get long streaks of low numbers. Some of the players I was with had an absolutely terrible session because they failed to make even the easiest of skill checks. This doesn't happen with high numbers. Though my bigger concern is the occasional lag I get on the roll results.

Here are aggregate overall stats of currently 392,559 rolls (which is a lot more rolls than you tallied), <a href="https://app.roll20.net/home/quantum" rel="nofollow">https://app.roll20.net/home/quantum</a> Here is more information about how random it is & how it is generated, <a href="https://wiki.roll20.net/QuantumRoll" rel="nofollow">https://wiki.roll20.net/QuantumRoll</a> Keep in mind there can be swings (like you tallied) in any smaller sample-size set like a few hundred rolls. Note: This thread is not a bug, so I will move this thread into the General on-topic section of the forum.

It's counter intuitive but it would actually be more unusual if your result were evenly spread. I wouldn't even say the 1 is particularly ambiguous (I don't think it'd be particularly noticeable if it were anything other than the 1) but anything close to random on a smallish sample size will have some peaks and troughs - I would imagine it's just an unfortunate coincidence that the most significant trough on your sample is the 20.

Human brains are very bad at understanding randomness and probability. Our brains are great for calculus (even if you thought the class in school was difficult, your brain does that kind of math automatically when you try to catch a ball) and great for spotting patterns... but we also spot patterns that aren't there. It's better to see a stick and run away from the nonexistent snake than it is to be bitten by the snake you didn't see. You've tallied up 1,218 rolls. While it's a better sample size than the ones used by many lay people who challenge randomness in a given venue, a good sample size for pulling bits from the RNG is going to need to be ten or twenty times that large, preferably more.

I looked into this very issue of true randomness when working on our small moderated chat for World of Darkness and was planning on implementing a free version of this using a 3rd party server. It's cool to see Roll20 recognizing that players/customers would want something like this and then implementing it in such a professional manner.

In a nutshell; Very.

Question: how often does the system have streaks? That's the problem our games seem to be having. It doesn't really matter to me if its fair over hundreds thousands of rolls, but that it's fair over the few hundred I might see in a given session. I work on video games, and to make things 'feel' more fair we sometimes use a system where it rebalances chances based on previous results. Make it impossible that a character with 80% accuracy to miss three times in a row, for example. I don't really think that 'my' system fits in table top gaming, but I've had two of my six campaigns on Roll20 end in streaks of crits. I use 3d6 and totally removed fumble rules when I dm online because of those games.

It does not support or have any mechanism for streaks. It is as random as random gets. Small samples sizes will show anecdotal evidence for unusual behaviour which you may attribute a cause to. Also the above result you mention means 4 of 6 campaigns did not end in streaks of crits. The same mechanic that produced a level distribution Mark G. has depicted above governs both 100,000,000 rolls and 3 rolls. It's all the same engine. However, if I roll 100,000,000 times and got 20 each time we can reasonably deduce the system is broken. If I roll 3 natural 20s in a row the odds are only 1 in 8000. Roll20 as a platform has measured 850,000+ rolls by now, so if someone somewhere had NOT seen 3 natural 20s in a row we could equally conclude something was wrong. Hope that helps!

Random is good. Streaks are good. They happen.

I'd love to see a chart on distribution of streaks from the roller, just out of curiosity. I'm sure that its been tested for streaks, but I do feel that that sort of data is as useful from the user perspective as the 'fair across half a million rolls' chart is. My last campaign ending streak was 20, 20, 20, 19, 20. I've played for 2700ish hours. lets assume I roll five times every minute and discount any time I may have spent prepping. I have thus rolled ~800,000 times. In the situation, it wouldn't have actually mattered if it was 19 or 20, so lets open it up and call it 1/10. That streak had a ~1 in 100,000 chance in happening, and thus is totally believeable that it would happen. 8 whole times, in fact. Still feels terrible that it did, which is the real reason this sort of thread keeps popping up and why I mentioned the streak killer. Also why I mentioned that I use my own streak killer in the form of 3d6. Though, it did generate a story. The paladin who took those crits got off easy, since he died before his soul could be turned into a wraith.

@ Duodecimus Here is a streak-killer code, Instead of rolling /r d20 Just roll /r 11 For example if you recently rolled a 1, and you need a nice 14 to balance it out and make sure you don't get a streak with another 1, /r 14

I believe that is known as fudging the rolls and is one of the reasons some DMs roll things in secret.

I agree. On the one hand, there is random with streaks. Then on the other hand, there is fudging. Another idea to eliminate streaks and ensure even distribution is go in a straight rotation. Example, your first attack is a 1. Second attack is a 2. And so on. After you get your attack that's a 20, you go back around to 1 again.

Brian said: Human brains are very bad at understanding randomness and probability. Our brains are great for calculus (even if you thought the class in school was difficult, your brain does that kind of math automatically when you try to catch a ball) and great for spotting patterns... but we also spot patterns that aren't there. It's better to see a stick and run away from the nonexistent snake than it is to be bitten by the snake you didn't see. You've tallied up 1,218 rolls. While it's a better sample size than the ones used by many lay people who challenge randomness in a given venue, a good sample size for pulling bits from the RNG is going to need to be ten or twenty times that large, preferably more. +1

Gold said: I agree. On the one hand, there is random with streaks. Then on the other hand, there is fudging. Another idea to eliminate streaks and ensure even distribution is go in a straight rotation. Example, your first attack is a 1. Second attack is a 2. And so on. After you get your attack that's a 20, you go back around to 1 again. Actually, you could very easily make a deck with 20 cards in it from 1-20, and draw them. That would give you perfect distribution and an easy way to keep track. If you want to allow the possibility of streaks of a given length, duplicate the cards to the streak length of your choice. There is a similar deck for Catan.

We need a forum section just for this.... I say that somewhat joking and somewhat not joking. We have a wiki section... that too often does not get read on this topic. We have always have some really good replies that address this that get lost over time (my favorite ones are when Brian explains human failings quite accurately. The charm in those is his tone that makes the reader think he is speaking as if he is NOT part of the human race.... which could also be quite actuate. &lt;- a joke Biran, humans do that from time to time.) And it's understandable for new people to want to talk about this, however it could be contained and collect. Maybe a FAQ or 2FAQ ;) We all know this topic will come up, over and over and over… And that is OK, just saying bundle the conversation somewhere. Just a random thought.

A wiki page explaining this might be a nice idea...

There is the Wiki documentation page about this, (already linked earlier in the thread) <a href="https://wiki.roll20.net/QuantumRoll" rel="nofollow">https://wiki.roll20.net/QuantumRoll</a> If any additional FAQ on this topic, it could be added to that page.

Duodecimus said: . . . My last campaign ending streak was 20, 20, 20, 19, 20. I've played for 2700ish hours. lets assume I roll five times every minute and discount any time I may have spent prepping. I have thus rolled ~800,000 times. In the situation, it wouldn't have actually mattered if it was 19 or 20, so lets open it up and call it 1/10. That streak had a ~1 in 100,000 chance in happening, and thus is totally believeable that it would happen. 8 whole times, in fact. . . . This is why so many people think that there are problems with random generators; probability is hard. 1 in 100,000 is the odds of getting that one specific (19 or 20 five times in a row) streak. If you expand that 'memorable' streaks and include 1 or 2 five times in a row that goes up to 1 in 50,000. If you expand that to 'any set of two numbers that are only 1 point apart' the odds rise to 19 in 100,000 or almost 1 in 5,000. The real thing, however, is this; the 1 in 100,000 chance is based on a somewhat arbitrary rule that is designed to fit the conditions while giving a very low probability that it actually happened. We had four rolls out of five that were 20's. Normally the odds of that are around 1 in 33,684, and that's because we are only considering 20's. The odds of rolling four of the same nonspecific number is about 1 in 1,684. However, since these were 20's we get to kick it up to the lower probability of 1 in 33,684. Then because the non-20 was really, really close to the 20 we modify the rule a little bit more to come up with our 1 in 100,000 rule. In short, what's done is that people look at rolls and then retroactively apply a rules set that is 'aesthetically pleasing' but which produces the lowest probability possible. It really is nothing more than our pattern recognition spotting patterns that aren't really there (or that are just occurring through happenstance if you want to argue that yes, that specific roll constitutes a pattern) the same way we see shapes in clouds.

The Aaron said: Actually, you could very easily make a deck with 20 cards in it from 1-20, and draw them. That would give you perfect distribution and an easy way to keep track. If you want to allow the possibility of streaks of a given length, duplicate the cards to the streak length of your choice. There is a similar deck for Catan. Again, to show how difficult probability can be to understand; this will give you an overall even distribution but it won't eliminate streaks. With a single deck of 20 cards you are still susceptible to getting the same number twice in a row and you could conceivably have a streak of 10 cards in a row all with values of 5 or less, just as examples.

Yeah, I was restricting my streak measurements to a single shuffle of the cards, but I take your point. =D

I actually tend to notice patterns usually tend to be campaign specific, and last most of the campaign. I've had characters that practically sneezed 20s, and even when that character left, my followup character also sneezed 20s. Similarly I have had campaigns which i couldnt do a thing right with 1s popping up everwhere, and even after my pc died, the next character had more of the same. *shrug*

My final say before i withdraw from this thread: The issue with simulated 'random' results over the real time, die rolling 'random' is the ability to question the outcome. When rolling a die you can attempt to influence its result by rolling in a certain way, flicking the wrist, using a lucky shaker or praying to &lt;insert deity here&gt; and if the outcome is consistently undesirable the only conclusion is that 'it just isn't your night'. The flip side is that if the outcome is consistently favourable then maybe 'its your lucky night!'. When that same result and fluctuation in results, potential patterns or streak is simulated by a dice engine the automatic response is to question the validity of the engine rather than just assume the same 'luck' applies to both your real rolls and your virtual rolls (unless you are doing high dice rolling to the power of 2 + 1 in which case streaks can certainly apply... don't ask!). The net result is that if a human can blame something for their failings then they will, its human nature. When the fault is our own and undeniably so we'll accept it or blame unproven external factors. Personally, i blame The Aaron.

Ziechael said: Personally, i blame The Aaron. +1

I feel like I'm something of a proven external factor at this point, and thus unblamable... at lest in this regard! =D

Eliminating streaks, or especially trying to guarantee 1 through 20 are each rolled at least once every 20 rolls, would actually promote a sort of card counting among players. If you have a streak eliminating algorithm and two 20s just came up, the player might change what his next action is, figuring the next roll will be lower to even out the distribution. that is the last thing you want. Streaks happen, accept them, they are not good nor bad.

I agree with Ziechael I think some of its a mental thing. With real dice there is no way you can really chalk it up to "my dice aren't being random" its either you think you got a lucky streak, or an unlucky one. Being its virtual dice, its a lot easier to point out and feel doubt the results are truly random when you get a pattern you notice. Like the vsauce video says about "what is random" Humans are pattern finders, we will find patterns in total randomness. And True randomness can, and will, produce patterns. We are pattern seekers, its how we progressed to make sense of the world while evolving. I also think its a bit of the "you tend to remember the best and the worst" human trait. I have yet to see someone complain about "does 8's seem to come up a lot for others on d20s" or "anyone notice the pattern '5,3,14,7' seems to repeat a lot?" its because we forget the mundane, and focus on the extremes, which make them stand out more as being noticeable, therefor skewing your perception of the ratios of rolls. The best example I ever heard about 'human random' vs 'true random': Have a computer generate a 1000 digit random number, and a person. A mathematician or statistician can walk in glance and tell which the person made the majority of the time. Because they will try to make it look random. No one will have, say, the sequence 1234567 in their number, or 55555, or 202020, which the computer generated one can, and usually always does have such 'non random to us' sets.

I actually know several people who don't like virtual dice, because they actually think they can influence dice. They will point to when they rolled a crit at the opportune time as proof, They apparently also try to roll fumbles at other times, people remember success and failure, all of the stuff in between just fades.

al e. said: I actually know several people who don't like virtual dice, because they actually think they can influence dice. They will point to when they rolled a crit at the opportune time as proof, They apparently also try to roll fumbles at other times, Wow...now THAT really IS "Fantasy" gaming

William R. said: I agree with Ziechael I think some of its a mental thing. . . It is largely a mental thing. The truth of the matter is that physical dice aren't really all that random. If you buy just about any 20 sided die that is made and subject it to a chi-squared goodness of fit over 1000 rolls you will almost always show a strong likelihood of bias. It's not something which you will probably be able to identify just looking at the raw data (not correctly, anyway), not unless the die is seriously flawed, but the math will bear it out. There's just too many imperfections that will creep into the production of the dice. One axis will be .1mm shorter than another (and actually, that would be pretty good accuracy. I've measured dice with calipers and the deviation is often much higher), the two resins used to create the swirl pattern aren't absolutely identical in density, even the amount of plastic removed for the numbers will be different on the different faces and will ever so slightly alter the center of gravity. Your average PRNG on the other hand can pretty typically do hundreds of thousands of 'rolls' before it begins to show any biasing. There is also often an issue with what is referred to as 'the period'. Because PRNGs are based on mathematical formulas they can eventually 'loop' and you will start the whole series of randomly generated numbers again, which is something dice won't do. However the typical period of 'bad' PRNG are in the billions, so the odds of you ever rolling enough dice in practical use is non-existent (you would have to be rolling once every second for 136 years). So the long and the short of it is that no, a PRNG isn't absolutely random, but for all practical purposes it is probably substantially more random than the dice we already use. At that point it becomes sort of like the joke about the two guys and the bear. I don't have to outrun the bear. I only have to outrun you.