Random Die rolls not too random

[evilpope]

Im running 5E btw.

I have had several cases where the roll to hit for creatures controlled by me as GM have rolled the same over and over again. I had a scenario where i had a ghoul roll a 17 4 times in a row against a fighter. I then switched to bite and rolled a different number then went back to the same ghoul and rolled its claw attack and again rolled a 17. This does not seem to random to me.

I also had another situation where I had some creatures that had multi-attack. I killed a whole party when the enemies couldn't miss with there attacks. It doesnt seem to random when they die rolls are 18+ almost every time. Not sure if anyone else has come across.

[dulux-oz]

I've had similar situations in the past with physical rolls - while it is unlikely, its not impossible (the chance of winning the lottery is a lot harder / less likely).

As this is the only time that I'm aware of that this has been reported over the entire 10+ years of FG being available I think we can safely put it down to pure chance, don't you?

Cheers

[evilpope]

im going to be running another game in a couple weeks I'll report something then if it happens again.

[dulux-oz]

Sounds like a plan :)

[Zacchaeus]

Welcome evilpope to FG and to the community.

This comes up now and again. It is just random, and although as Dulux-oz says it is unusual it can happen that several dice rolls will come up the same. You can guarantee however that when you want to roll a particular number, say for testing an effect or to see if a template is functioning correctly, said number will remain illusive :D

[dulux-oz]

...You can guarantee however that when you want to roll a particular number, say for testing an effect or to see if a template is functioning correctly, said number will remain illusive :D

Ain't that the truth! :p

[MarianDz]

still you can't count on it. And this is miracle of randomization as in real life :)

[rob2e]

I had a player roll a 1 five times in a row in a face to face (dice on table). Odds are 1 in 3.2 million. Unlikely but it can happen.

[dulux-oz]

I had a player roll a 1 five times in a row in a face to face (dice on table). Odds are 1 in 3.2 million. Unlikely but it can happen.

On what side die?

Five 1s on a d20 is 1 in 3.2M (which is what I assume was the die size), but five 1s on a d4 is only 1 in 1024 - I know, I was demonstrating some different probabilities :)

[rob2e]

On what side die?

Five 1s on a d20 is 1 in 3.2M (which is what I assume was the die size), but five 1s on a d4 is only 1 in 1024 - I know, I was demonstrating some different probabilities :)

It seems you answered your own question. Yes, d20.

[Nickademus]

Real dice are influenced by more than statistics. How you pick up, hold, and roll a die can take some of the randomness out of it. I remember back in my 2nd Ed. days I had a mean d6 curve for rolling 6's for stats.

[DelasGoodfelllow]

I find this to be very true. It's really a problem across the board for computers. They are not inherently random. They do so, typically by using the time string (in part). This ought to be a hint about this and I have complained in the past here about it (a very long time ago). I do find that making 3+ rapid randoms tends to start giving more random results. So, if for instance, if a computer routine was to roll 3 times and take the third only, it might work better. There are other ways to make random better if more effort was put to it. BUT, for some reason, all game devs seem to just take random as random when it ain't. While it is true that random is not as we often envision it (i.e., you will have clusters and repeats and 1k rolls of a d20 doesn't give you 50 rolls of each number), the more rolls you make, the more closely the breakdown generally gets. And, if this was followed and tested, I think dev's might find that computer random isn't near as random as it ought to be. ;-) If only, they'd care... (or perhaps if the "bosses" cared more) ;-)

[Andraax]

(i.e., you will have clusters and repeats and 1k rolls of a d20 doesn't give you 50 rolls of each number)

That doesn't happen in real life with real dice either, so I don't see the complaint. If "random" rolls were evenly distributed they would no longer be random. Dice rolls are independent - they do not rely on rolls either before or after to determine their outcome.

I had a game designer talk about it once, he said that if it worked that way in real life, and a fighter carried 6 missiles and generally hit 1 out of 6 times, then the pilot would be smart to drop 5 of the missiles right after taking off, because then that last one would be a guaranteed hit (since the first five missed).

[DelasGoodfelllow]

I have had several cases where the roll to hit for creatures controlled by me as GM have rolled the same over and over again. I had a scenario where i had a ghoul roll a 17 4 times in a row against a fighter. I then switched to bite and rolled a different number then went back to the same ghoul and rolled its claw attack and again rolled a 17. This does not seem to random to me.

Exactly. Not just in FG, but in many, many games. Dev's all take the random generator for granted. It's not truly random. It's based on the timestamp. The end result is you get a disproportionate amount of repeats. Yes, repeats will occur - the key is "disproportionate." Sometimes, I'll ignore the first roll (or even first 2 sets), then make a 3rd roll set and it'll be better. I wish devs in a great many games would consider modifying their random generator. It might help to do something like what I've just said. Take a random roll based on something other than timestamp and then use that to determine the random roll number to output as the roll. Just create a test random generator that records it stat wise for few hundred. And, try different methods to come up with one that works better. From what I recall, the problem is more often when a human rolls every 1-2 minutes (or rapid fire) than in large numbers - so, have the test generator use a few timed methods.

[damned]

The current engine uses a randomisation AND a physics engine. A very small number of people have reported what looked like non random numbers in their games. Some of these have looked very unrandom. Its quite possible that there was a glitch in the game engine that caused it to throw put non random results but in my experience (and Ive rolled a lot of dice on FG - see here https://www.youtube.com/watch?v=zbWs7Lw9aXg) Ive never seen it do anything I wouldnt call random - but that doesnt mean it hasnt happened as a result of a glitch - and of course there are just random numbers that will not appear random because random is random...

[seycyrus]

Exactly. Not just in FG, but in many, many games. ...

Not to sound too harsh, but as a scientist... I want to see the data. How large is the deviation and how does it vary from the statistical clumping that occurs in the real world?

It really sounds like to me that you are picking and choosing the rolls you want because they are not "random enough". That's hardly random either. What's your criteria? The results don't "feel" or "look" random? That's junk science.

[dulux-oz]

I'll echo what seycyrus said.

As a Computer Scientist I've actually deconstructed and re-constructed the various popular random() functions. The one in the lua stack is actually the C random function, which in its the latest iteration is as close to random we can get with current technology - its certainly good enough for the various casinos and similar enterprises to feel comfortable in using.

And as damned has said, when you add in the physical moment of the mouse and releasing the mouse button, FG doesn't get much better at "random" than that.

The problem a lot of people have is recognizing the difference between random events and coincidents - the human brain is wired to see patters very, very well (is that a tiger hiding over there in the reeds). This is one reason that people "see" Jesus in a piece of toast, or images in the clouds (rain clouds, not AWS), and also why its so hard for a computer to do actual facial recognition (as opposed to comparing to a fixed set of existing photographs). I suspect (strongly) that what you're seeing is nothing more than a sequence of numbers that your brain is "wired" to recognize and at the same time dismissing any rolls that don't fit your "pattern".

And to make matters worse, we're only looking at small sample sizes - d4s-d20s.

But here's what you do: record each roll that you make in a game. Keep doing this until you'll got close to a million+ rolls. Then actually do the statistical analysis on the results and if you don't see a pretty uniform distribution (with a small difference or "error" in each one) then I'll believe that there is something wrong - but until we've got a good set of data (as good scientists should) then all we have is conjecture.

[irish_carbomb]

As a DM if my players start rolling the same number over and over, or constantly are rolling very badly I'll stop the game and have everyone roll about 10 attacks. Don't think it actually helps but it makes the players feel better.

Also everyone knows (and it's scientifically proven), that when you start rolling poorly you need to change the color of your dice..

[seycyrus]

For crying out loud, you naysayers crack me up and you sure got accusatory attitudes for people claiming to be all "scientific."...

I'm not the one who chooses to discard a set of rolls because they "don't look random", as if that judgement were possible without statistical analysis.

First of all, your google search doesn't return the results you wish it did. Secondly, I just used random.org which claims to be be better than typical computer algorithms, and set it to generate 1-20. On the 7th and 8th roll I got 11. It happens. It ain't broken.

I don't need to "google random" and read up on a bunch of articles to point out that while the lack of true randomness might be a problem for extremely large complex simulations, it is not a problem for the situations we're faced with rolling dice on a VTT.

Furthermore, there is the very real fact that unless you are gaming with a bunch of robots, the circumstances that dictate what rolls are required when (with what desired result), will be random.

You specifically mention being able to note back to back 1s and 20s (due to their special connotations), but unfortunately that just reinforces the explanation that a logical fallacy is actually at work.

[LordEntrails]

When I buy lottery tickets I like to get the numbers 1, 2, 3, 4, 5, 6.

Drives people crazy thinking I'm stupid. But they are just as likely as any other set. Of course, I only pay a voluntary tax when the pot is big enough to be ridiculous and I'm feeling generous.

[ddavison]

I mentioned this on a twitch stream last night when someone asked about it. The chatlog is saved locally and you can always go back and look at it -- ask the DM to send you a copy. You can look and see for yourself it something really seems amiss. I know that in D&D 5E, you roll with advantage or disadvantage a lot, so that will skew things up or down obviously. Overall, I've never seen it and I chalk it up to the same thing that causes people to ban their dice or change them up. If FG didn't have good or bad runs, then I'd really be concerned that it wasn't working properly as a simulation of what we all see around the table where people feel especially lucky or unlucky for stretches at a time.