I take issue with randomness of pulls, please prove me wrong

Unknown

PeterGibbons316 wrote:

sinnerjfl wrote:

we shouldnt be able to see patterns with "random" results. Is D3/DU loading the dice somehow? It's possible and it can be done both ways, improving or reducing "lucky" pulls.

Your RNG just doesnt feel random and it needs improvements.

The problem isn't that it doesn't feel random, the problem is that you don't understand what random should feel like.

Anyone who has ever pulled 12 assorted tokens and got 12 Storms, and then pulled 12 more assorted tokens and got 12 Juggs, and then 16 more assorted tokens and got 16 Dakens could in fact be intelligent enough to understand what random is supposed to feel like, despite the fact that what they're experiencing feels anything but "random."

It's not exactly rocket science that random often feels a bit less than random for many MPQ players. We could all Google the clinical definition of random and keep reposting it, insult each other needlessly, or accept that to a relatively sane person, experiencing "randomness" wherein one pulls the exact same card 5, 10, 20, or 50 times in a row starts not feeling so random. Whatever people need to do I guess.

DBC

simonsez

OneLastGambit wrote:

I was really just trying to highlight that long odds does not make it impossible.

But at some point, one needs to reject the null hypothesis. If people insist on 100% confidence intervals, no one needs statisticians. Setting the bar at 1 in millions is far, far more conservative than anything ever employed in the real world.

OneLastGambit

simonsez wrote:

OneLastGambit wrote:

I was really just trying to highlight that long odds does not make it impossible.

But at some point, one needs to reject the null hypothesis. If people insist on 100% confidence intervals, no one needs statisticians. Setting the bar at 1 in millions is far, far more conservative than anything ever employed in the real world.

That's a fair point but you can't reject null hypothesis without examining your entire collected data set. You can't pick and choose which data applies and doesn't because it either supports or refutes your accepted hypothesis. Examine the entire collected data set - then a result and conclusion can be drawn. We've said this in a previous thread and both agreed that D3 was never going to do that and we would never get an answer on this either.

simonsez

OneLastGambit wrote:

You can't pick and choose

Who's picking and choosing? Someone maxed a 5* with ungodly speed, posted how he did it, and it was virtually impossible if one were to assume randomness. That's hardly cherry-picking. I didn't ignore millions of other posted token pull results that showed no anomalies.

OneLastGambit

simonsez wrote:

OneLastGambit wrote:

You can't pick and choose

Who's picking and choosing? Someone maxed a 5* with ungodly speed, posted how he did it, and it was virtually impossible if one were to assume randomness. That's hardly cherry-picking. I didn't ignore millions of other posted token pull results that showed no anomalies.

Unless this occurrence of someone doing it where to happen commonly the results suggest it has little statistical significance to the overall data set. The important parts of your post are

"Virtually impossible" and "Anomalies"

I think given this one user (out of 1000's) was the only person to do this I think it's fair to say his results were the anomalous results of this data set (the data set being the collected token pulls of every single player).

and virtually impossible is not impossible. As I said in my earlier post the creation of the earth is as close to mathematically impossible as you can get yet it still happened. As long as there are the chances that something can happen then it will at some point happen on a long enough time line. that dude is just extremely lucky - also I did not see his post but did he provide evidence that he did not simply buy the covers needed (you can still get CP via purchases remember) or that his account was not a sandboxed one? two things which could affect the viability of his post which have not been mentioned anywhere that I have seen, though if it has I would like to see it.

scottee

People realize that results posted on the internet are often embellished, right?

Someone might post that they pulled 5 Moonstone reds, and they actually did. But then someone says, "omg! I just pulled 5 juggs in a row!", when they actually only pulled 3 out of 5. Then someone says it's so rigged that they got 10 storms in a row when it was really 3 out of 4.

Taking people's selective memories that are told on the internet as embellished stories as data for statistical analysis is a pretty bad methodology.

The only data that's valid is if sunstone is writing down every pull over a long period of time and posting all the data, including all the seemingly "normal" runs.

...or if some team had access to the numbers on every token pull ever done...

The Bob The

UncleSmed wrote:

I long for days of more pure randomness to the tokens. Just the other day I was thinking I'd like to open an LT and get a flock of ducks appearing in game for me.
Or to match a 5 of a kind and have pudding fall to replace the tiles.. mmm pudding

You think that's random? Used to be you could open a token and it would send an email or change all your phone's icons.

Mercurywolf

If we are talking about long odds, I am just gonna leave this right here:

http://mentalfloss.com/article/66863/meet-man-struck-lightning-7-times

OneLastGambit

Mercurywolf wrote:

If we are talking about long odds, I am just gonna leave this right here:

http://mentalfloss.com/article/66863/meet-man-struck-lightning-7-times

You don't get much longer odds than the creation of the earth...

10390/1- the longest odds of anything that I have come across

roderic1

I try to keep in mind Tetris. The question has been posed as to whether or not it could be played perpetually given the right mind and conditions, assuming the pieces fall completely randomly. The answer ended up being that it can't be because, given enough time, there would be runs of 'z' and 's' pieces that would yield a well that cannot be cleared. That would be some pretty bad luck. But it is entirely possible in a completely random situation.

However, I believe most Tetris games use a system now where all pieces are shuffled and dealt out 7 at a time before the next 7 are dealt out. This isn't truly random, but it makes the game more enjoyable for the end user.

Unknown

...how does one actually "Prove" that pulls are indeed random?

Like, what sort of data would be needed? IceX posting a complete list of Legendary token pulls? (And even then, would people say it's liiiiies?)

This seems like a hard challenge to complete.

sinnerjfl

Drummerboycroy wrote:

Anyone who has ever pulled 12 assorted tokens and got 12 Storms, and then pulled 12 more assorted tokens and got 12 Juggs, and then 16 more assorted tokens and got 16 Dakens could in fact be intelligent enough to understand what random is supposed to feel like, despite the fact that what they're experiencing feels anything but "random."

It's not exactly rocket science that random often feels a bit less than random for many MPQ players. We could all Google the clinical definition of random and keep reposting it, insult each other needlessly, or accept that to a relatively sane person, experiencing "randomness" wherein one pulls the exact same card 5, 10, 20, or 50 times in a row starts not feeling so random. Whatever people need to do I guess.

DBC

This, thank you for explaining it better than I could. I know that random means random, therefore anything should be possible but there are also odds involved. Sometimes things that happen in this game that are really really weird and shouldnt happen often but yet they still do.

There are many methods of randoming for computers, the one that D3/DU is using seems to yield too many anomalies too frequently. Sure, you can say that's what random is. But it's not.

We see stuff happen that shouldnt happen on a frequent basis. Look at cascades...

XandorXerxes

There are two issues being discussed in the thread:

1) Are tokens random
2) Is the RNG appropriately random.

The OP's question (and the 5* player case) reference the first issue. If the second issue was a problem (and I believe it is) that doesn't explain the OP's issue - that the two covers he pulled were two of the few covers he has maxed. Unless the RNG derives randomness from a player's roster (which would be incredibly horrible at generating randomness, given how infrequently rosters change compared to what's normally used for randomness), I don't see how the two issues would be correlated here.

If the RNG were the issue, I'd expect OP to have pulled two of the same character or something similar that would indicate that something in the number generation isn't sufficiently generating randomness. For the tokens to target 4* and 5* characters that are already maxed, that would have to be intentionally programmed in or something in generating the token pulls would have to be taking roster into account. I really don't see that happening, and if it were I can't imagine they'd ever admit it. Then again, maybe these covers work just like mission rewards. At least 4* covers are worth more than 20 ISO?

simonsez

scottee wrote:

People realize that results posted on the internet are often embellished, right?

But in this case, we had a dev offer implicit confirmation.

If something has a probability of 1 in 687 million, I'm very comfortable saying that the occurrence of that event refutes the notion of randomness. Otherwise, you can be this guy

rossmon

Random means anything is possible. What I find highly improbable is that the devs spent the time and money to figure a way to code it so that you are more likely to get what you don't want, and not get what you want. BTW, statistically, the odds of getting the same token that you last pulled are the same as the odds for that token.

scottee

The devs have evidence that the tokens distribute evenly. So what everyone who says it's rigged is implying is that not only is it coded to be random, but that it distributes unevenly to some people (like giving 10 moonstones in a row) and then it distributes unevenly in the reverse in order to make the numbers look correct (giving the rest of the playerbase 10 LESS Moonstones).

You're saying it's making some people unlucky, and others lucky, in indirect proportions so that the numbers look even in the end.

The amount of coding effort this would take for absolutely zero gain to the devs is pretty absurd.

If anyone's been a part of the online poker scene in depth, you know that plenty have people have accused the card distribution of being rigged. If the cards were distributed unevenly, the community would have been able to empirically track it and expose it. They did track it, and they found that the cards were being distributed just fine. They did find cheating where superusers could see cards on some sites, but none in terms of card distribution. There's zero incentive to make an unfair RNG, and all of the incentive in the world to use normal to the programming world RNG's that as close to fair as possible.


What we have instead are forums full of people who don't have empirical evidence, but are sure it's uneven because chatting with a handful of people. Getting a couple dozen people to agree with you on the internet doesn't make the claims true.

simonsez

rossmon wrote:

What I find highly improbable is that the devs spent the time and money to figure a way to code it so that you are more likely to get what you don't want, and not get what you want.

I agree, that'd be silly. But if they've skewed the odds table to favor older covers, indirectly, they've accomplished this.

DrStrange-616

We all fall pray to apophenia sometimes, seeing patterns where none exist. It does feel like they have their thumb on the random scale sometimes, but I doubt it's true. Depending on your level of vitriol, Hanlon's razor may or may not apply.

That said, being at the mercy of so much randomocity is one of the more frustrating and unsatisfying parts of this game.

simonsez

scottee wrote:

There's zero incentive to make an unfair RNG

Come on, you have more imagination than that. If the RNG is spitting out MrF and Bag Lady instead of cykes, rhulks and iceman, that's incentive to spend HP to buy the covers you need. And if the "RNG" gives a player four 5*s out of just two 40 packs, that incentivizes other people to try their luck too. There's plenty of incentive.

scottee

simonsez wrote:

scottee wrote:

There's zero incentive to make an unfair RNG

Come on, you have more imagination than that. If the RNG is spitting out MrF and Bag Lady instead of cykes, rhulks and iceman, that's incentive to spend HP to buy the covers you need. And if the "RNG" gives a player four 5*s out of just two 40 packs, that incentivizes other people to try their luck too. There's plenty of incentive.

That line was referring to online poker.

But essentially you're saying that a few people being extra lucky incentivizes spending, and lots of people being extra unlucky incentivizes spending. That's ALREADY accomplished by just using a regular RNG. Like the lottery, casinos, crane arcade games, little machines you put a quarter in to get a random toy, sweepstakes, playing the stock market,

You guys are all being psychologically played by a regular RNG. I know it feels good to think you're smarter than that, but there's no conspiracy. The human biases are so evident in every post like this. No one's willing to track with empirical data long enough. Instead, we fall victim to selective memory, confirmation bias, seeing patterns where they don't exist, embellishment, and voluntary response bias.

Yet this is exactly what companies that use RNG rely on. They used to give a guaranteed feature cover in the 10 pack, but they found people buy more when you take away the guarantee, even with the exact same expected value. That's psychology, not conspiracy programming.