Question about so-called "Luck Accounts"

[Last Warrior Lord.7248]

Been playing for 9 years and only1  Precursor (underwater weapon) I cant remember nothing else, this was years ago. For as luck I have the worst luck with a 150% account wide luck. I have gotten ascended drops many times but that's bout it. It is what it is.

[HnRkLnXqZ.1870]

Sadly these topics are mostly a question of faith. You have people who believe the system is 100 % fair and you have those who question it. Each side accuses the others of defying the truth and ignoring the facts. 

The facts of those who claim the system is fair are basically quotes from staff-members. Would you buy a car, if the sales-man tells you the lights do not always work? Probably not. Would the sales-man tell you this information in the first place? Probably not, he wants to sell cars.

The facts of those who claim the system to be unfair mostly rely on zooming into the program itself. That is not what a player is supposed to do. The deeper you go, the more problems you find. Computers are far from being perfect, even today. The issue about creating a perfect RNG is just one of the problems. It is not the only one, but it is probably the only one we can discuss without derailing into tech gobbledygook too much.

The big question: Does it matter?

The sad answer: We do not know!

From the player-perspective, we lack the depth knowledge about how the system truly works. Without that understanding, we cannot see how much impact the inaccuracy of our RNG truly has for our luck. 

Best we can do is guessing according to our observations.

[Random Wax Orc.7695]

Quote

There are lies, damned lies, and statistics.

 

Pseudo Random Number Generators are not random.  They provide complex patterns that are statistically random and statistically fair when viewed as an aggregate.  Beyond that, anything might be true.  Examining the statistics on a different scale (e.g. individual players or new players, or whatever) will not necessarily show the same random distribution of the aggregate.  The implementation of the code, method of distribution of the numbers, and other factors can easily result in emergent patterns of number distribution from the complex patterns that are provided by PRNG.

What can you do about it?  Nothing... except perhaps add randomness in your behavior, and hope for some confirmation bias to give you the impression of luck.

Criticism about the implementation of RNG in games?  Short of generating their RNG from white noise generators, not much can be done.  However, the tiered loot tables are a sloppy implementation of RNG code.  Using numbers from a complex non-random pattern repeatedly in roll for table one, then a roll for table two, then a roll for table three, etc... can cause emergent patterns that can stop even remotely appearing random, especially when the size of the loot tables are often close to the size of the random seed.  What does that mean?  If the probability of a result is small (on the order of maximum number of results of the RNG) then the RNG pattern is going to diverge from random, except on a gross statistical level.  Nesting calls for small results in tiered loot tables will amplify the problem.  Yet statistically, it will still be globally fair, but only globally.  However, we can't audit Anet's code nor collect proper data, so it is a moot point.

 

Anyway, there is nothing you can do about it.  The horse is dead, and there are no horses in Tyria.  It is simply an academic issue... that essentially everyone strawmans one way or the other, based on their biases.

[Astralporing.1957]

2 hours ago, Random Wax Orc.7695 said:

However, the tiered loot tables are a sloppy implementation of RNG code.  Using numbers from a complex non-random pattern repeatedly in roll for table one, then a roll for table two, then a roll for table three, etc... can cause emergent patterns that can stop even remotely appearing random, especially when the size of the loot tables are often close to the size of the random seed.  What does that mean?  If the probability of a result is small (on the order of maximum number of results of the RNG) then the RNG pattern is going to diverge from random, except on a gross statistical level.  Nesting calls for small results in tiered loot tables will amplify the problem.  Yet statistically, it will still be globally fair, but only globally. 

And why exactly is that? from my understanding of probability, having tiered loot tables by itself does not change absolutely anything. It's just an easier way to split one bigger table into smaller ones to make it easier to manage, but without having any impact whatsoever on end results.

[rune.9572]

If they truly wanted to debunk this they'd release the source code of their RNG for everyone to inspect ¯\_(ツ)_/¯

But they don't have to. We already know all RNGs are imperfect because deterministic machines can't produce true randomness.

Even if the RNG used a truly random source, the value is still stored as a binary structure in a set amount of memory, which all have fixed entropy.

Edited November 19, 2021 by rune.9572

[Ashen.2907]

I have gotten so many Invisible boot drops that I have taken to selling them to the merchant to free up bank tabs. 

[Randulf.7614]

11 minutes ago, Ashen.2907 said:

I have gotten so many Invisible boot drops that I have taken to selling them to the merchant to free up bank tabs. 

I reckon I’ve had loads drop too. Trouble is, I just can’t seem to see them

[Random Wax Orc.7695]

3 minutes ago, Astralporing.1957 said:

And why exactly is that? from my understanding of probability, having tiered loot tables by itself does not change absolutely anything. It's just an easier way to split one bigger table into smaller ones to make it easier to manage, but without having any impact whatsoever on end results.

Because the pseudo random numbers pulled for each tier are not actually random.  The set of not-really-random numbers are not independent of one another.  As single numbers they can statistically appear as random numbers, but as sets of numbers derived from seeds (also using code on server/client that processes at a fairly consistent rate) there will increasingly likely be non-random patterns.

In the simple situation of one seed per player, instead of (number of random seeds)^(number of tiers) possibilities you get something closer to (number of random seeds)^1 results.  In the case of an MMO, you don't get your own seed, but the processing of the server, client, connection speed, and any number of code related factors can shift/limit the patterns experienced by each player.  But there are still patterns.  That rare RNG result for a tier that triggers the next tier might have generally unfortunate patterns for some and extremely fortunate patterns for others.

The point being that since RNG is based on patterns not true randomness, any complex use of RNG will result in patterns related to the complexity, and less randomness.

[Khisanth.2948]

58 minutes ago, rune.9572 said:

If they truly wanted to debunk this they'd release the source code of their RNG for everyone to inspect ¯\_(ツ)_/¯

But they don't have to. We already know all RNGs are imperfect because deterministic machines can't produce true randomness.

Even if the RNG used a truly random source, the value is still stored as a binary structure in a set amount of memory, which all have fixed entropy.

According to THIRDPARTYSOFTWAREREADME.txt -> http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/emt19937ar.html

[Astralporing.1957]

49 minutes ago, Random Wax Orc.7695 said:

Because the pseudo random numbers pulled for each tier are not actually random.  The set of not-really-random numbers are not independent of one another.  As single numbers they can statistically appear as random numbers, but as sets of numbers derived from seeds (also using code on server/client that processes at a fairly consistent rate) there will increasingly likely be non-random patterns.

In the simple situation of one seed per player, instead of (number of random seeds)^(number of tiers) possibilities you get something closer to (number of random seeds)^1 results.  In the case of an MMO, you don't get your own seed, but the processing of the server, client, connection speed, and any number of code related factors can shift/limit the patterns experienced by each player.  But there are still patterns.  That rare RNG result for a tier that triggers the next tier might have generally unfortunate patterns for some and extremely fortunate patterns for others.

The point being that since RNG is based on patterns not true randomness, any complex use of RNG will result in patterns related to the complexity, and less randomness.

There will be patterns working exactly the same in a true RNG-generated chain of results as well. The patterns you speak of are not a result of "pseudonandomness". They are a result of your mind having a tendency of assigning patterns to pretty much anything (look up rorschach test, for example).

[Random Wax Orc.7695]

2 hours ago, Astralporing.1957 said:

There will be patterns working exactly the same in a true RNG-generated chain of results as well. The patterns you speak of are not a result of "pseudonandomness". They are a result of your mind having a tendency of assigning patterns to pretty much anything (look up rorschach test, for example).

No.  The patterns of numbers from pseudorandomness are not an observer phenomena.  There is a finite set of possible random seeds and each seed provides a fixed series of numbers from the equations.  Any patterns to selecting a seed or selecting elements of a series will create patterns in the output.  It is an emergent statistical phenomena based on complexity, nothing to do with psychology.  This would not happen with true randomness (like from a white noise generator).  However, with PRNG there is essentially a finite set of strings/series of numbers, one for each possible seed.  Using a large amount of numbers from each string, will create patterns because any patterns of selection are superimposed on the base pattern of the finite set.  (look up emergence and complexity)

A proper (overly simple) example would be take the numbers from 1 to 10, and rearrange them in a variety of ways.  There are 10! ways to arrange them (10*9*8*7*6*5*4*3*2*1 = 3,628,800 ways) but you can only keep maybe 1% of the arrangements.  So 99% of the arrangements will be missing.  The numbers in the each spot of the sequence can be evenly distributed over the whole set of arrangements, so any selected number can be evenly populated both within a seed and within a position in all seeds.  However, if you select two consecutive numbers within a seed, that set of two numbers will unlikely be evenly populated within the set.  Three numbers even less likely.  Something like 1,5,7,2 vs 2,1,5,7 vs 7,2,1,5 vs 5,7,2,1 will almost never be equally likely.  Such a simple pattern can break the randomness of the number generator.  As the patterns of selection increase in complexity, patterns in the finite arrangements will emerge.

But as I said, there isn't really a fix for this phenomena.  The science/math however indicates that an individual should not expect "fair" results from RNG.  The statistical distribution of results may seem level globally, but emergent patterns can easily result in an improbable distribution of individual results.

[Zok.4956]

On 11/19/2021 at 6:39 PM, HnRkLnXqZ.1870 said:

The facts of those who claim the system to be unfair mostly rely on zooming into the program itself.

 

Would you be so kind to write down those facts? But please, only facts, no opinions or speculations. Thanks.

 

[Leo.3428]

20 hours ago, Random Wax Orc.7695 said:

No.  The patterns of numbers from pseudorandomness are not an observer phenomena.  There is a finite set of possible random seeds and each seed provides a fixed series of numbers from the equations.  Any patterns to selecting a seed or selecting elements of a series will create patterns in the output.  It is an emergent statistical phenomena based on complexity, nothing to do with psychology.  This would not happen with true randomness (like from a white noise generator).  However, with PRNG there is essentially a finite set of strings/series of numbers, one for each possible seed.  Using a large amount of numbers from each string, will create patterns because any patterns of selection are superimposed on the base pattern of the finite set.  (look up emergence and complexity)

A proper (overly simple) example would be take the numbers from 1 to 10, and rearrange them in a variety of ways.  There are 10! ways to arrange them (10*9*8*7*6*5*4*3*2*1 = 3,628,800 ways) but you can only keep maybe 1% of the arrangements.  So 99% of the arrangements will be missing.  The numbers in the each spot of the sequence can be evenly distributed over the whole set of arrangements, so any selected number can be evenly populated both within a seed and within a position in all seeds.  However, if you select two consecutive numbers within a seed, that set of two numbers will unlikely be evenly populated within the set.  Three numbers even less likely.  Something like 1,5,7,2 vs 2,1,5,7 vs 7,2,1,5 vs 5,7,2,1 will almost never be equally likely.  Such a simple pattern can break the randomness of the number generator.  As the patterns of selection increase in complexity, patterns in the finite arrangements will emerge.

But as I said, there isn't really a fix for this phenomena.  The science/math however indicates that an individual should not expect "fair" results from RNG.  The statistical distribution of results may seem level globally, but emergent patterns can easily result in an improbable distribution of individual results.

Would a single tier mechanism with a much larger set of numbers counter the problem with skewed sequences, or would it just shift the problem, as the full rotation would miss even more sequences?

[Abyssisis.3971]

9 years, 0 precursors here so definitely not on the lucky list. Had a guildie that would vomit out precursors at least once a week, so I reckon there is some truth to lucky accounts.

[Random Wax Orc.7695]

12 hours ago, Leo.3428 said:

Would a single tier mechanism with a much larger set of numbers counter the problem with skewed sequences, or would it just shift the problem, as the full rotation would miss even more sequences?

That is a good question.  I don't have a good answer.  Too many unknowns might be involved for any certainty.  The only solution I am aware of is to make an end run around PRNG and use a white noise generator for random number generation instead.  Search for "white noise random number generator".  I don't know if there is a practical one for the purpose, based on the necessary volume of random calls needed for a MMO.  If I was going to start up an MMO or anything else with "gambling" involved in its infrastructure, I would have someone research that topic.  If a practical one doesn't already exist, creating a fast efficient plug and play WNRNG might be a good startup.

Since the numbers output would be actually random, you would still have lucky and unlucky people, but they would be distributed more like a bell curve around the average... without enduring spike patterns we might anecdotally see with PRNG.

[Tanek.5983]

I can't say what is going on behind the scenes, but my main account is on NA and was created back at launch.  Earlier this year I made an EU account and in just that amount of time it has been "luckier" with drops than the NA account has been in 9 years.

I figured it was just the game deities messing around with me.

[Sobx.1758]

On 11/21/2021 at 1:24 AM, Abyssisis.3971 said:

9 years, 0 precursors here so definitely not on the lucky list. Had a guildie that would vomit out precursors at least once a week, so I reckon there is some truth to lucky accounts.

If I flip a coin and I get 4 "tails" in a row, it means the coin is rigged.

Random doesn't mean everyone gets even number of rare drops. That's not how it works at all.

Edited November 22, 2021 by Sobx.1758

[MarkoGold.7126]

the only thing i heard related to the op question is accounts having a hidden value that affect drops, supposedly created when you create the account and rng. but i have no idea if its true or not. 

Edited November 22, 2021 by MarkoGold.7126

[sewenewes.4176]

I don't know if there are lucky and unlucky accounts or if luck ever changes but I have never experienced good fortune in this game and it's one of the primary reasons I keep taking breaks from it. The highest value item that ever dropped for me was Genesis, one of the weapons needed for the Exotic Hunter achievement. Never got a precursor and never got a rare drop from black lion chests. No Tequatl weapons. No invisible shoes. No infusions that weren't bought.

 

While I enjoy doing content for its own sake in short bursts, I'm not motivated to stick around and farm for long when I'm never rewarded for it. Similarly I have also stopped spending on keys because the fun of the anticipation is just not there anymore as I know I will absolutely never get anything interesting.

[Tytos.3076]

If anything, the dramatic unevenness in 'luck' between different accounts is an argument for dynamically adjusting drop rates and reeling in the extreme outliers a bit. Of course, this would lead to shrieks of "WELFARE!" from the usual suspects, so it could never be implemented openly.

[mythical.6315]

1 hour ago, Tytos.3076 said:

If anything, the dramatic unevenness in 'luck' between different accounts is an argument for dynamically adjusting drop rates and reeling in the extreme outliers a bit. Of course, this would lead to shrieks of "WELFARE!" from the usual suspects, so it could never be implemented openly.

How are you measuring this “unevenness”?  Do you have access to some data which we do not have?  Threads like these are just another case of people not understanding randomness. 

[Abyssisis.3971]

11 hours ago, Sobx.1758 said:

If I flip a coin and I get 4 "tails" in a row, it means the coin is rigged.

Random doesn't mean everyone gets even number of rare drops. That's not how it works at all.

 I gave my opinion based on my experiences in game. If you don’t like it, I really don’t care so whatever floats your boat. 😂

[Danikat.8537]

On 11/19/2021 at 7:00 PM, Random Wax Orc.7695 said:

 

Pseudo Random Number Generators are not random.  They provide complex patterns that are statistically random and statistically fair when viewed as an aggregate.  Beyond that, anything might be true. 

There's a huge difference between 'computers cannot generate truly random numbers' and 'anything might be true' in this context. The question was whether specific accounts are more likely to get rare and valuable drops than others. Ensuring that doesn't happen does not require a truly random number generator, it just requires that everyone is using the same one and it's not tied to their account in any way.

The only way even an imperfect RNG system could create lucky and unlucky accounts is if the account is part of the number generation process in some way. As long as the RNG is using the same process for everyone it doesn't matter if that process is genuinely random or if it simply produces results that appear statistically random - each person has the same chance of getting a good or bad result.

IMO the bigger problem being demonstrated here is that the majority of players do not want or expect a truly random outcome. What they want is a fair system where rare and valuable drops are distributed equally among players. That would actually require the opposite - scrapping the RNG and deliberately tracking and assigning drops per account - something in the game saying 'you got an invisible boot box last week so you don't get one today, but this person hasn't had one yet so now it's their turn to get it'. Or something like the 'mercy of the mad king' achievement where after you've opened X containers you're guaranteed a rare drop.

Of course it couldn't be completely equal because player behaviour will affect it. For example if the game wanted to give me an equal share of precursors from the Mystic Forge I'd probably have to get one every time I put exotic weapons in, because I almost never do that. My lack of precursors from the Forge isn't due to bad luck but due to not trying, and there's no sensible way to accommodate that, in spite of the fact that it's another important factor which is almost always forgotten here.

I've gotten more ascended drops in the past year than I have in the last 8 or so years. Is that because my luck has changed? No, it's because I've been playing WvW a lot more and your chance of getting ascended drops from rank up chests is much higher than anything else I normally do. (I spend a lot of time playing story instances and open-world maps where I'm not fighting world bosses so a lot of the time it's impossible for me to get an ascended drop no matter how lucky I am.)

 

10 hours ago, MarkoGold.7126 said:

the only thing i heard related to the op question is accounts having a hidden value that affect drops, supposedly created when you create the account and rng. but i have no idea if its true or not. 

Yes, that's the rumour being discussed here. The problem is no one has ever offered any proof for it beyond "I know a guy who says he's always getting precursors".

What we'd need is someone to prove it. One of those people who claim they've done some deep analysis of the games behaviour to actually share their evidence and for it to be statistically sound. For example take a group of people who claim to have lucky accounts and ones who claim to have unlucky accounts and have them all put the same items into the Mystic Forge multiple times then record all the results. Not how many precursors they got or what their total profit/loss was or how they feel about it but every single item that went in and came out for each account and then compare them to show that the lucky accounts actually do get better drops and aren't simply remembering and sharing their good results more frequently, making more attempts to get rare items or using items more likely to get good results. (It gives items equivalent to what you put in so if you throw in only level 80 exotic greatswords you'll only get level 80 exotic greatswords back and have a much higher chance of a precursor than if you throw in rare and exotic items of all types and most of your results will be armour and trinkets.)

[Random Wax Orc.7695]

1 hour ago, Danikat.8537 said:

The only way even an imperfect RNG system could create lucky and unlucky accounts is if the account is part of the number generation process in some way. As long as the RNG is using the same process for everyone it doesn't matter if that process is genuinely random or if it simply produces results that appear statistically random - each person has the same chance of getting a good or bad result.

This is not true, unless you are broadly defining an "account" to include speed of the computer used on the account, the speed of the internet connection used on the account, and any latency/ping issues.  It is a bit like increasing favorable roulette results based on the the speed of the ball, where it was released, and then selecting a block of numbers from the wheel where the ball is likely to stop.  In GW2, how accounts get queued up for a massive stack of PRNG results can put some accounts habitually in lucky zones in that stack and some accounts in unlucky zones in that stack.  The difference might be generally imperceptible but the chance of getting an extremely rare drop is generally imperceptible as well... which can make that difference important.

My point in previous replies was exactly that.  Not only is the PRNG generator not random but the distribution of PRNG numbers will not be completely random either.  The distribution will be mindbogglingly complex, but not random.  The problem is that complexity leads to emergence and emergence leads to patterns.

 

If Harry and Ron both play GW2, but Ron's slow computer almost always gives him late results in the queue and Harry's fast Nimbus 2000 computer often gets him early results in the queue, then they can have completely different experiences from that queue.

 

BTW, I used the phrase "anything might be true" in the sense of "there is no way of knowing".

[Astralporing.1957]

14 hours ago, ewenness.6482 said:

While I enjoy doing content for its own sake in short bursts, I'm not motivated to stick around and farm for long when I'm never rewarded for it.

And that's the main reason for "lack of luck" right there. If you were to check for the "most lucky" accounts, you'd definitely find that most of them belong to people that do a lot of farming.

The chance of rolling getting 1 in a 1-100 roll is only 1%, but make enough rolls, and your chance of getting that result at least once will skyrocket.