Is your random number generator broken?

firethorne

Because I think it is.

Jarvind

(Seventeen posts angrily/condescendingly explaining what "random" means)

cyineedsn

Heehee, but yeah that sucks, pulled 13+ different 2* mags covers except for the one I needed

firethorne

I mean, I actually don't mind, since that was lightning round match rewards, not draws. But, all Moonstone, all red. To roll the same number 5 times in a row on a 75 sided dice... I'm legitimately wondering if something in the inner workings of super lightning round got a bit hinkey.

veny

Heh, 2*s... Want to see picture of my roster with 1/2/9 Kate Bishop? Or rather 8/10/2 Prof X? (he would be like 8/15/2 if i put him in roster three times instead of just twice )

smkspy

So that's where all my moonstone reds are going lol. I've been sitting on a 12 cover moonstone for nearly two weeks and cannot for the life of me pull that 13th red while all my black and purple covers are dying on the vine.

vinsensual

I'd love to be getting 2* covers for PVP rewards again. All I've gotten today was a bunch of 1*s.

Dragon_Nexus

veny wrote:

Heh, 2*s... Want to see picture of my roster with 1/2/9 Kate Bishop? Or rather 8/10/2 Prof X? (he would be like 8/15/2 if i put him in roster three times instead of just twice )

It was embarrasing how long it took me to champ X-Force Wolverine. He was something like 1/3/17. Even now when I get a random cover for him, it's usually yellow.
By the time I hit 5 yellow for him, he was 0/1/5. Took me ages to get covers...

elwhiteninja

yes, i am about to sell my 14th purple cover and have yet to get a but everyone has a unicorn

Quebbster

I just got four two-star cover drops from a Lightning round. Every single one was CStorm blue. No other covers dropped.
I'm not normally one to assume something is wrong, but this did seem odd.

carrion_pigeons

Jarvind wrote:

(Seventeen posts angrily/condescendingly explaining what "random" means)

Do you think you know?

Because I would just love to hear what you think it means. Having an actual discussion where people learn things seems to be so much better than plugging your ears and saying lalala whenever someone tries to explain something.

Freelancer

I experienced something similar in the Shield Sim last season.

I had 7 or 8 back to back.

AlphaNik

carrion pigeons wrote:

Jarvind wrote:

(Seventeen posts angrily/condescendingly explaining what "random" means)

Do you think you know?

Because I would just love to hear what you think it means. Having an actual discussion where people learn things seems to be so much better than plugging your ears and saying lalala whenever someone tries to explain something.

Random. What I think it means in this context: you pull one element from a finite set of n elements and every element have the same probability to be the chosen one.

Piro_plock

AlphaNik wrote:

carrion pigeons wrote:

Jarvind wrote:

(Seventeen posts angrily/condescendingly explaining what "random" means)

Do you think you know?

Because I would just love to hear what you think it means. Having an actual discussion where people learn things seems to be so much better than plugging your ears and saying lalala whenever someone tries to explain something.

Random. What I think it means in this context: you pull one element from a finite set of n elements and every element have the same probability to be the chosen one.

Okay. So how will you define probability?

Nick441234

Thats funny cause I'm sitting on 5 red Moonstone's as we speak. I think we've hijacked everyones red Moonstone's!

I'm also sitting on 3 red Human Torch's too so it seems red is my colour right now.

AlphaNik

Piro_plock wrote:

AlphaNik wrote:

carrion pigeons wrote:

Jarvind wrote:

(Seventeen posts angrily/condescendingly explaining what "random" means)

Do you think you know?

Because I would just love to hear what you think it means. Having an actual discussion where people learn things seems to be so much better than plugging your ears and saying lalala whenever someone tries to explain something.

Random. What I think it means in this context: you pull one element from a finite set of n elements and every element have the same probability to be the chosen one.

Okay. So how will you define probability?

In this context I'd say probability is a number from 0 to 1 that define how likely is that I pull a given element from a finite set.

0 -> Impossible
1 -> Certain

I have a glass with 2 little red balls in it. Probability I pull a red one ? 1. Probability I pull a green one? 0.
I have a glass with 1 litlle red ball and one little green one. Probability I pull a red one ? 0,5. Probability I pull a green one? 0,5. In this case I have an uniform distribution.

If I put the ball back in the glass after every pull things get a little more complicated and this is just the case with every token that is not in vault.


Agreed?

Piro_plock

AlphaNik wrote:

Piro_plock wrote:

Okay. So how will you define probability?

In this context I'd say probability is a number from 0 to 1 that define how likely is that I pull a given element from a finite set.

0 -> Impossible
1 -> Certain

I have a glass with 2 little red balls in it. Probability I pull a red one ? 1. Probability I pull a green one? 0.
I have a glass with 1 litlle red ball and one little green one. Probability I pull a red one ? 0,5. Probability I pull a green one? 0,5. In this case I have an uniform distribution.

If I put the ball back in the glass after every pull things get a little more complicated and this is just the case with every token that is not in vault.


Agreed?

Your definition is (more or less) correct. But let's take another model, for example a 6-sided die. On a fair die, each face has 1/6 chance of appearing. The problem is: how do you check if the probability is exactly what you assumed?

The answer is: you can't. You can never be 100% sure about any experimentally measured probabilities... You can only approximate by performing many, many throws. Perform an experiment: do 50-100 consecutive throws and note the results. You'll notice a lot of clusters, streaks etc. That's the nature of randomness. A small sample won't tell you much about the underlying probabilities.

With token pulls you have a much bigger pool of elementary events, so you can't judge the RNG fairness without a huge statistical sample.

firethorne

Tin foil hat conspiracy time. My only real reason for pause on this is not the nature of random numbers themselves, but the nature of Demi's infrastructure around random number generation in this game. It doesn't work the same way as it does in many other games. And by that, I mean the random element is not generated at the exact time the random element is needed. This much has actually been confirmed by interactions with customer support.

I opened a legendary token on my phone, got home, and while playing on my iPad, noticed the draw from my phone had not saved. My CP was there, the draw was not. I contacted support, asking if I could get that draw back. They said there was no need, and to simply draw again, and I'd get the same thing. Sure enough, I did.

That means the draw would have to pre-exists as a random element of my information in their database, and gets regenerated when "saved." This makes a lot of sense. Its purpose would be to prevent save scumming of their token draws (opening a token on one device, then force closing before it saved, and opening again on a different device). That was a problem at one time that was fixed.

Knowing that sort of business exists, the programmer wonders if to save some time, other rewards in the game aren't linked in to a bit of the same subsystems. And, with the added stress of writing multiple super lightning round brackets, the thread writing my new random seed didn't get a bit bogged down, and the thread reading it was able to pull the same value a few times before the thread to write a new one caught up.

Anyway, end of crackpot corner.


Edit: Announcement about the old save scumming fiasco, for reference.
https://www.d3go.com/forums/viewtopic.php?f=13&t=36681

Crnch73

firethorne wrote:

Tin foil hat conspiracy time. My only real reason for pause on this is not the nature of random numbers themselves, but the nature of Demi's infrastructure around random number generation in this game. It doesn't work the same way as it does in many other games. And by that, I mean the random element is not generated at the exact time the random element is needed. This much has actually been confirmed by interactions with customer support.

I opened a legendary token on my phone, got home, and while playing on my iPad, noticed the draw from my phone had not saved. My CP was there, the draw was not. I contacted support, asking if I could get that draw back. They said there was no need, and to simply draw again, and I'd get the same thing. Sure enough, I did.

That means the draw would have to pre-exists as a random element of my information in their database, and gets regenerated when "saved." This makes a lot of sense. Its purpose would be to prevent save scumming of their token draws (opening a token on one device, then force closing before it saved, and opening again on a different device). That was a problem at one time that was fixed.

Knowing that sort of business exists, the programmer wonders if to save some time, other rewards in the game aren't linked in to a bit of the same subsystems. And, with the added stress of writing multiple super lightning round brackets, the thread writing my new random seed didn't get a bit bogged down, and the thread reading it was able to pull the same value a few times before the thread to write a new one caught up.

Anyway, end of crackpot corner.

I agree, this happens to me often. I always end up pulling the same character the second time after the game hadn't saved, which does indeed tell me that it's not so random... which is mildly upsetting only because it means that any frustrating dupe cover pull feels extra frustrating because it was somewhat predetermined.

Also, I am up to 21 Ghost Riders red covers, zero black covers yet... so my RNG is mean. Not a big deal because he isn't a hugely awesome character, but that black cover would be nice to max out... even if only to send as team ups for boss battles.

AlphaNik

Piro_plock wrote:

AlphaNik wrote:

Piro_plock wrote:

Okay. So how will you define probability?

In this context I'd say probability is a number from 0 to 1 that define how likely is that I pull a given element from a finite set.

0 -> Impossible
1 -> Certain

I have a glass with 2 little red balls in it. Probability I pull a red one ? 1. Probability I pull a green one? 0.
I have a glass with 1 litlle red ball and one little green one. Probability I pull a red one ? 0,5. Probability I pull a green one? 0,5. In this case I have an uniform distribution.

If I put the ball back in the glass after every pull things get a little more complicated and this is just the case with every token that is not in vault.


Agreed?

Your definition is (more or less) correct. But let's take another model, for example a 6-sided die. On a fair die, each face has 1/6 chance of appearing. The problem is: how do you check if the probability is exactly what you assumed?

The answer is: you can't. You can never be 100% sure about any experimentally measured probabilities... You can only approximate by performing many, many throws. Perform an experiment: do 50-100 consecutive throws and note the results. You'll notice a lot of clusters, streaks etc. That's the nature of randomness. A small sample won't tell you much about the underlying probabilities.

With token pulls you have a much bigger pool of elementary events, so you can't judge the RNG fairness without a huge statistical sample.

You're right. in fact I don't believe MPQ's RNG (where are you Adrian Cronauer) to be unfair or broken. I'm pretty sure in the long (or very long run) it will conform to an uniform ditribution. I just have some subjective and maybe unjustified suspects that it's a bit too prone to generate stream. I do think it's possible to determine if this is the case even with a relativelty small data sample and the appropriate mathematical instruments, like a t-test or somethink like that (this it's me really scratching the bottom of my high school memories).

Freelancer

Without knowing the specific chance of each reward in the PvP loot pool, we'll never be able to accurately calculate the probabilty of a specific outcome.

5 Red Moonstones, while still perfectly random, has such a very small probability of happening.

X X X X = Should have bought a lottery ticket?