something fishy - rng is not rng

"

Gemdraco wrote:

"

Mark_GGG wrote:

For the record, the system being described by Arrowneous does not exist in the game. Unique drops are not recorded, and have no effect on future unique drops*.

Unique (and other) item drops are independent events.

*except for if a unique drops which grants IIQ or IIR stats to the user, which would, technically, affect future drops if you equip it. Which isn't what we're discussing at all, but I'm worried if I don't explicitly mention it, someone's going to jump on that specific case as a way to prove me wrong.

What about mirror drops? I wouldn't dismiss what Arrowneous stated so quickly.

Mark_GGG, i too would like to have that assurance as well for high end currency. for the record.

For the record, has anybody within GGG ever been deceptive about internal mechanics such as the loot system?

Considering they never discuss droprates in absolute terms, you can't really trust anything GGG staff says about the subject, since the culture seems to be to hide information rather than share it, unlike Diablo 2 with the Arreat Summit.

I loved that website.

"

geradon wrote:

Mark_GGG, i too would like to have that assurance as well for high end currency. for the record.

Tin foil hats much?

I saw a video of a guy doing a vaal side area. He found 2 exalted orbs within 1 minute of each other. If it were gated like that, that would mean that every minute someone finds an exalted orb...

"

Gobla wrote:

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.

What you talk about the fact that streaks are more likely to happen that even distribution is right, but the problem here is that our event has a 0.0001% chance to happen. So streaks of FAILURE are more likely to happen, and evenly distribution here is far more likely to happen than a streak of SUCCESSES. And this is true to almost all cases where the sucess event has less chance to happen than the failure event. If you witness a certain amount of events, for example, you kill 50000 mobs, and you drop 5 shavroness, but in a streak, then it means that something is PROBABLY wrong. In this case, the most sucesses in the streak, the less likely it is to happen(kind of obviously, but your supperior math vision seems to highily blind you). In perfect random distributions(what the drop system in poe supposedly is) you will see a uniform distribution of events if you repeat the process enough times. It doesn't means that it will happen exaclty like success/fail/success/fail/etc..., just that a pattern representing it's actual average chance to happen will appear. So you will really find that, in general, a shavroness drop every 10000 kills. Because that is the chance of it to happen and it's what the probability really means, or else it would just not make any sense at all. For example, saying that something has 50% chance to happen(throwing a coin and get one of the sides) and then saying: I already got head, so we can expect it to happen again (in a streak) than getting tail this time? THAT is not how probability works AT ALL. It will be 50% chance to get head or tail again, since the first event already happened and doesn't matter anymore, we suppose at least. So aways dropping the same item in a streak rather than in a sort of uniform distribution indicates that there is a pattern in the distribution. And it is also very obvious.

Try this, create a random image made of white and black dots with enough resolution(1000x1000 might be enough). Then make each pixel have 50% chance to be black or white. And then check the image and, what a suprise, it's gray! Now make another image with each pixel having 99% chance to be white. I will let you guess it. Now make yet another image where you have still have 99% chance of each pixel being white, but then make that when on point happens to be black, each point after that one has 100% chance to be black - 1% chance for each new black point in a streak, until it eventually happen to be white again. WOW IS THIS BLACK MAGIC?

"

Veruski wrote:

Considering they never discuss droprates in absolute terms, you can't really trust anything GGG staff says about the subject, since the culture seems to be to hide information rather than share it, unlike Diablo 2 with the Arreat Summit.

I personally only trust second account forum trolls for factually accurate statements about POE and its workings.

"

Jiero wrote:

TRNG does and to an extent some prng systems will but these are to slow/heavy for use to use as a arpg prng system, Enterprise level PRNG that can work for environments like this do not... the difference being that these kinds of prng are made to be light and fast with minimal resource use and near zero wait time (so just a seed and a cpu clock check). That means it has numerous issues, such as a more limited subset of results that will repeat themselves a lot quicker then more fully featured prng systems plus heavier tendencies to cluster results more tightly. Once that pattern has exhausted itself and repeats then it doesn't matter how many more times you run it because the distribution and results will be the same (just the seed can change this table, which is often based on date from the system clock). These are also predictable enough that people have written programs that can tell you what its generating and is why these prng systems cannot be used in either computer/server/internet security or in gambling games (with real money flowing). Hence they are deemed GOOD ENOUGH to fill the needs of a server based arpg, they are not however actually good systems.

In fact there was a gif once posted here that displayed how it reacted on a windows system compared to a linux based one as well. The linux one formed nice little clusters that while predictable and limited simulated some form of psuedo randomness (still had a lot of white space in clean patterns, with almost all results clustered) while the exact same code in windows formed bars lining across the image with massive white space. The only thing I think that can alter this requires a macro outlook that extends the results table so that multiple clusters of results occur within each desired possibility (using a lot more resources but will hide the clusters) instead of just the usual dividing the result by the number of possible entries to get a particular item.


edit - Heck, look at all the games in which you can indirectly MANIPULATE the results (fairly close to all games). It lets me pick what 'random' enemy I fight and what is rewarded in any rpg game with 100% accuracy (if done right), lets people pick what pokemon they would encounter, allowed people to pick the 'random' charm table they would get in monster hunter 3u and etc... and all it needed was to set the system clock or to reset the game in order to gain control of these 'random' results.

The point I was making is that GGG don't need to set extra constraints to limit the # of mirrors flowing in -- setting a reasonable probability should result in a steady flow of mirrors.

Though, yes, PRNGs can be attacked (and that can be interesting) :).

Old games used garbage RNG, if any at all -- it was a lot cheaper for them to just compute some hash of the game state and treat that as "random."

I'd still love to get more details about how GGG's RNG/drop calculation works :D.

"

dessloc wrote:

"

Veruski wrote:

Considering they never discuss droprates in absolute terms, you can't really trust anything GGG staff says about the subject, since the culture seems to be to hide information rather than share it, unlike Diablo 2 with the Arreat Summit.

I personally only trust second account forum trolls for factually accurate statements about POE and its workings.

Yes. I'm lying about the arreat summit.

Oh wait. Enjoy:

http://classic.battle.net/diablo2exp/

"

syrioforel wrote:

The point I was making is that GGG don't need to set extra constraints to limit the # of mirrors flowing in -- setting a reasonable probability should result in a steady flow of mirrors.

Though, yes, PRNGs can be attacked (and that can be interesting) :).

Old games used garbage RNG, if any at all -- it was a lot cheaper for them to just compute some hash of the game state and treat that as "random."

I'd still love to get more details about how GGG's RNG/drop calculation works :D.

That is the thing with a clustering PRNG though, the smaller the chance of the result your looking for the less likely it will be close enough to a result cluster that it will occur and that makes its chance far less likely then intended (or actually impossible at times based on the seed). If their PRNG system clusters as much as I have a feeling it does then these extremely small chances are pushed into a rarity level they never intended them to be because of limitations in how it can generate the numbers. And to be fair, it can also occur in the reverse if a result cluster is centered around that near impossible chance and suddenly the things are raining down on you.

You have to do more then just set the chance to a rate you feel is fair, it has to be set up in a way that extends the results table so that results in it are in increments that mirror the lowest drop rate (or a lowest common denominator of all drop rates) with more common items being a multiple of those drop chances. Each drop chance needs to extend across an equal number of clustered results so as to hide them and each cluster needs a roughly equal chance to occur. So if the rarest item is .001% then everything in the table needs to be a multiplier of that with items taking multiple slots as needed, plus the actual table needs to check in increments well below 0.001% to make sure that each table entry contains more then one result cluster.

edit - Monster hunter 3U was just released last year iirc on the 3ds and wii u and still could be manipulated like that. But ya, I would find it interesting to learn more then just inferring what the systems is based on watching drops and anecdotes but they likely can't come out and say what it is due to the chances are that it will invite people finding exploits in it. They also should never reveal their source code of the RNG they use, as a black box system where we really don't know what is contained inside is better for the game. However, they should still continually work behind the scenes to keep the RNG accurately returning results at the rate they desire them to be.

Thanks for the GGG official answer. There are tons of posts stating that high level items (rares and uniques) are not not dropping because you are using "gated" RNG. Since we now have authoritative proof from GGG that this is not true, then we can all go back to our normal griping that the drop rates for higher level quality items are just set too low across the board. This goes hand in hand with PoE's horrible item progression past level 50. For example, I got this to drop:

Itemlevel = 45

So I got this drop while I was at level 50 character level. Now I'm up to level 63 and haven't gotten anything better to drop, not even a better rare. So it's not "gated" RNG but just progression (item scaling) failure that makes PoE "suck donkey" in the pursuit of better gear in the higher levels. Had to be one or the other.

Note: Every time I post on item progression failure in PoE I always get post's against any changes and that I want "loot galore" of the likes of RoS or TL2. Hell no! I've always stated that loot overload would ruin PoE play balance and longevity just as much as loot starvation. Right now GGG has PoE tuned to the low end and loot starvation (on gear that is above itemlevel 50) is the norm. While I never (and neither does GGG) want a flood of quality uniques it is time (actually way overdue since Chris stated over a year ago in an on-line interview that he wasn't happy with item progression) for GGG to tweak the drop rates and give us (the long time PoE players who put up with a lot of game flaws because we love the overall game play and mechanics of this arpg) a buff in above 50 level gear to compensate us for technical problems that subtract from maximum fun. Way too much 1 and 2 socket yellow vendor trash to wade through.

Thank you Mark_GGG for coming in and putting the question of PoE having "gated RNG" or not to rest. This is the latest example of the stellar forum support that GGG delivers to us day in and day out. There are many threads and 100s of posts here daily and I am very pleased that the forum mods and GGG employees oversee and reply to our posts as much as they do. Keep up the great work.

"

syrioforel wrote:

"

Mark_GGG wrote:

For the record, the system being described by Arrowneous does not exist in the game. Unique drops are not recorded, and have no effect on future unique drops*.

Unique (and other) item drops are independent events.

*except for if a unique drops which grants IIQ or IIR stats to the user, which would, technically, affect future drops if you equip it. Which isn't what we're discussing at all, but I'm worried if I don't explicitly mention it, someone's going to jump on that specific case as a way to prove me wrong.

This should be obvious.

It would be rather nice to have such things stated explicitly somewhere to stop all the rabble. :D

Having better technical documentation on PoE would be nice. Having better item progression past level 50 would be better for all (except maybe the PoE masochists).