something fishy - rng is not rng

"

mark1030 wrote:

"

johnKeys wrote:

"

Gobla wrote:

RNG is streaky by it's very nature. An evenly spread result is the very definition of not-random.

true RNG is random by nature.
which means given a set of outcomes of size n, the probability for it to be a streak equals the probability for it to be evenly distributed, equals the probability for it to be anything in between.

tendency to create streaks is a property of a PRNG, and the higher the probability of a streak, compared to other possible outcomes - the bigger sign it is that this PRNG is poorly implemented.

I think you may need to brush up a bit on your statistics and probability course (if you took one). I don't remember much from mine, but I remember that random and evenly distributed did not mean the same thing. Bell curves and such.

and you need to brush up on your reading skills.

what I said, translates roughly to:
give me a list of results. the probability for that list to be a streak equals the probability for it to not be a streak.
and both equal the probability for those results to somehow be evenly distributed: x numbers in (t,y) range, x numbers in (y, y+(y-t)) range, and so on (yes I know I forgot a +-1 in there somewhere. sorry about that) for some arbitrary x, t, and y.

that's what "random" means: you can't predict or analyse random results - only calculate probability for those results to happen.

PRNG isn't random. otherwise it wouldn't be "P".
but it should be close enough to being random, unless shitty implementation makes it no where near random.

"

johnKeys wrote:

"

mark1030 wrote:

"

johnKeys wrote:

true RNG is random by nature.
which means given a set of outcomes of size n, the probability for it to be a streak equals the probability for it to be evenly distributed, equals the probability for it to be anything in between.

tendency to create streaks is a property of a PRNG, and the higher the probability of a streak, compared to other possible outcomes - the bigger sign it is that this PRNG is poorly implemented.

I think you may need to brush up a bit on your statistics and probability course (if you took one). I don't remember much from mine, but I remember that random and evenly distributed did not mean the same thing. Bell curves and such.

and you need to brush up on your reading skills.

what I said, translates roughly to:
give me a list of results. the probability for that list to be a streak equals the probability for it to not be a streak.
and both equal the probability for those results to somehow be evenly distributed: x numbers in (t,y) range, x numbers in (y, y+(y-t)) range, and so on (yes I know I forgot a +-1 in there somewhere. sorry about that) for some arbitrary x, t, and y.

that's what "random" means: you can't predict or analyse random results - only calculate probability for those results to happen.

PRNG isn't random. otherwise it wouldn't be "P".
but it should be close enough to being random, unless shitty implementation makes it no where near random.

You're very simply wrong.

Given 3 tails out of 6 coin tosses there's 4 possible permutations of a streak:

XXX000, 0XXX00, 00XXX0 and 000XXX.

There's 2 possible permutations of an even spread:

X0X0X0 and 0X0X0X.

It's twice as likely when getting flipping a coin and getting three tails for it to have been a streak than it to have been evenly spread.

There's other possible outcomes certainly that are neither evenly spread nor a streak. But streaks are nonetheless more likely than even spreads. And their likeliness increases with more possible outcomes than a simple coin toss.

Because a streak requires only a subset of all results to follow the predicted pattern, the rest doesn't matter. An even spread requires all the results to follow the predicted pattern, which is much less likely.

"

Gobla wrote:

Given 3 tails out of 6 coin tosses ....

There's 2 possible permutations of an even spread:

X0X0X0 and 0X0X0X.

sorry, seems I didn't understand what you meant by "even spread".
now I do.

but my definition still holds in the most simple form:
in True-Random, you can't know if a sequence will be a streak or not.
in Pseudo-Random, you can't know for most of the time, but you can provide a good "guess" for the outcome.

in terribly-implemented, buggy Pseudo-Random - the answer is "yes".

"

johnKeys wrote:

sorry, seems I didn't understand what you meant by "even spread".
now I do.

but my definition still holds in the most simple form:
in True-Random, you can't know if a sequence will be a streak or not.
in Pseudo-Random, you can't know for most of the time, but you can provide a good "guess" for the outcome.

in terribly-implemented, buggy Pseudo-Random - the answer is "yes".

The thing is, you don't know how randomization was implemented in PoE, you just assume it must be terrible because you think you have terrible luck (I'm more inclined to believe you're simply bad at the game, no offence). Fun fact: even systems with a virtually true randomization (like most professional poker sites use, or the one this site uses http://roll20.net/) frequently get accused of being buggy pseudo-random systems because people simply suck at understanding the irrational nature of RNG.

"

johnKeys wrote:

"

Gobla wrote:

Given 3 tails out of 6 coin tosses ....

There's 2 possible permutations of an even spread:

X0X0X0 and 0X0X0X.

sorry, seems I didn't understand what you meant by "even spread".
now I do.

but my definition still holds in the most simple form:
in True-Random, you can't know if a sequence will be a streak or not.
in Pseudo-Random, you can't know for most of the time, but you can provide a good "guess" for the outcome.

in terribly-implemented, buggy Pseudo-Random - the answer is "yes".

I can provide a pretty good guess when needing to roll a perfect-20 on a d20 to succeed. I'm going to fail.

I agree that PoE's RNG isn't very well implemented, but that's not due to bugs. I'm almost certain there's no bugs there because I can't think of a single possible reason why PoE wouldn't use an RNG from a standard library, depending on the OS the servers run. Alternatively there's the option of various popular open-source libraries. There's absolutely no incentive to write their own algorithm and the possible bugs that come with that.

PoE's RNG simply isn't well implemented due to a lack of interesting outcomes. The vast majority of outcomes are universally bad, a minority are universally good and basically none at all make you think how you could use this.

"

Gobla wrote:

PoE's RNG simply isn't well implemented due to a lack of interesting outcomes. The vast majority of outcomes are universally bad, a minority are universally good and basically none at all make you think how you could use this.

Very true.

But you know, I'm not sure GGG took an existing library for their RNG, being the stubborn people they are.
After all, they insisted on creating their own graphics engine - and all this time and FPS drops/freezes/resource-leaks later, it's safe to say they suck at that right now.

Though even if they did "black box" a good, stable PRNG - that doesn't mean they aren't giving it the kind of input, or manipulating its output, to make the game what it is.

Even after demonstrating you know nothing about randomness and distribution, you still keep claiming that GGG's RNG is poorly implemented, presenting all of zero evidence. It's astonishing.

If I was Dictator of the World, the *FIRST* thing I would do would be to mandate probability and statistics classes *every year* so people can eventually get trained out of their inability to understand random events.

"

BGSacho wrote:

If I was Dictator of the World, the *FIRST* thing I would do would be to mandate probability and statistics classes *every year* so people can eventually get trained out of their inability to understand random events.

BUT MATH IS HARD!

"

Mark_GGG wrote:

"

BGSacho wrote:

If I was Dictator of the World, the *FIRST* thing I would do would be to mandate probability and statistics classes *every year* so people can eventually get trained out of their inability to understand random events.

You'd have my vote (if anyone was allowed to vote under this hypothetical dictatorial regime).

A dictatorship usually is run under the "one man, one vote" system. The dictator is the man and he gets the vote.