Guild Wars 2

Since the “discuss rng conceptually” thread appeared, I have put more time into learning about current RNG models.

I felt, anecdotally, that there were a number of instances in my own experience, and in forum posts, regarding “runs.”

For example, 3 of 5 members pf a partty I was in for Tequatl recently got exactly the same aquabreather, and none of us had seen one in quite a while before that.

I got two Truths in two consecutive bags in EoTM, two weeks previous to that.

Back to the question..

In my research, I learned that there is a statistically disproportionate chance of runs in some pseudorandom number generation models (apparently, Java uses one of them, for example.)

Is the model used by Guild Wars 2 using the generator of its language, custom designed, chipset seeded, or…?

John, I imagine you have the tools to do an analysis of the quality of the rng model for the game. Has this been done, or are you willing to? I do understand you may not be able to publish your findings…

EDIT: This is a real, twchnical question with a finite answer. There -are- issues with some random number generators. Please read this to educate yourself on the topic:

http://en.m.wikipedia.org/wiki/Pseudorandom_number_generators

Generally speaking, the level of non-randomness seen in pseudo-RNG is so small as to be unnoticeable by even the posters in this thread pooling all our data. Coincidences happen, so it’s not proof of anything that the OP has seen several of the same items popping up among their acquaintances and friends.

That said, OP is correct in suggesting that the choice of pRNG tool could matter, especially if seed selection isn’t properly managed. And OP is also correct in thinking that ANet chose one tool or another and someone could tell us. OP is also probably correct in thinking John Smith would be a good person to ask.

However, I can’t imagine why ANet would give us that sort of information. Those who don’t understand pRNG or probability can’t make use of the knowledge; those who are intimately familiar with the details might be able to take advantage of such data. But more importantly, the vast majority of us, who only think we know something about the topic, would draw incorrect conclusions based on insufficient information about how GW2 actually works. As Trahearne used to say, “this won’t end well.”

tl;dr it’s most likely that the question is moot and even if it weren’t, it’s not in ANet’s best interests to tell us.

Try this sticky thread for questions about the economy:

https://forum-en.gw2archive.eu/forum/game/bltc/I-have-a-question-about-the-economy/page/16#post4538345

Assume they’re using a quick and dirty random number generator – an LCG, or even a raw XORshift – because passing every single test of randomness just isn’t critical for their application while minimal computation time is.

Imagine they’re using a vanilla XORshift with a generic clock seed on server reset. What would you expect to be the difference between that and a cryptographic quality RNG with an analog seed?

Ensign.2189:

What would you expect to be the difference between that and a cryptographic quality RNG with an analog seed?

Predictability.

If there is a correlation of successive values in the results of the random number generation, there is an increased likelihood of repeated success… and potentially why we see some players who pop multiple precursors out of the forge in a relatively short time period.

Please note that I’m not a disgruntled RNG’er looking for rationalization of personal bad luck, this has come out of playing this game for over 5600 hours and 5000-ish on GW1, where I was a pre searing trader who noticed patterns in when to expect players to visit me with their black dye drops to sell.

If the RNG system for the GW1/GW2 engine is the same, these patterns over time from the old game may be relevant to the new one.

Either way, it’s more of an academic question — I don’t expect, demand or hope it to be changed.

Wanze.8410:

Try this sticky thread for questions about the economy

This isn’t about the economy.

Allisa Wonderland.8192:

Wanze.8410:

Try this sticky thread for questions about the economy

This isn’t about the economy.

Why do you direct your question at the game economist then?

I have no idea.

That being said, it might be your mind playing with yourself.

It’s like the birthday question. Say in a class of 30 people, the probability of 2 people having the same birthday is quite high.

http://en.wikipedia.org/wiki/Birthday_problem

Wanze.8410:

Why do you direct your question at the game economist then?

Read paragraph 1 of the original post.

laokoko.7403:

I have no idea.

Yep… And no, that’s not where I am coming from.

To clarify: I am a logically minded thinker who knows what gambler’s fallacy is.

Allisa Wonderland.8192:

Wanze.8410:

Why do you direct your question at the game economist then?

Read paragraph 1 of the original post.

So i guess it is about the economy.

Allisa Wonderland.8192:

laokoko.7403:

I have no idea.

Yep… And no, that’s not where I am coming from.

The same topic comes up on the forum with every ARPG I played. People always wonder why a few of the same item drop at the same time and ask on the forum.

The reason I bring up the birthday parodox is because of this.

The probability of 2 people in a class of 30 with both birthday on january first may be really small. But the probability of 2 people actaully having the same birthday is really high.

It’s right, I don’t know if what you said is actually correct. But most likely there isn’t any problem with the RNG system. I think what you said is perfectly normal. Getting 2 truth in a row or have 3/5 people getting the same aqua breather.

Because think about it, getting 2 of a specific item is rare, but getting 2 of the same item in a row isn’t that rare.

edit: I was just typing out this message, and someone in my guild link the truth in guild chat! What a coincident!

Allisa Wonderland.8192:

If there is a correlation of successive values in the results of the random number generation, there is an increased likelihood of repeated success…

Or, a reduced likelihood of repeated success; it’s much easier, and much more likely, for a generator to traverse its domain too uniformly than the alternative (at least, while passing even more trivial tests of randomness).

Sometimes I’ll even use such generators on purpose, since for some applications a lower entropy generator that covers the while space more quickly can be quite valuable.

Ensign.2189:

Allisa Wonderland.8192:

If there is a correlation of successive values in the results of the random number generation, there is an increased likelihood of repeated success…

Or, a reduced likelihood of repeated success; it’s much easier, and much more likely, for a generator to traverse its domain too uniformly than the alternative (at least, while passing even more trivial tests of randomness).

Sometimes I’ll even use such generators on purpose, since for some applications a lower entropy generator that covers the while space more quickly can be quite valuable.

Probably. But I say it is more a math paradox than actually tech problem. Like the monty hall problem.

If I could have 1$ every time I heard this type of question the OP have on ARPG/MMORPG forum, I have enough money to buy a PS4.

It’s more about how our brains work than the way the program works. We don’t remember the thousands of boring loot drops, where everyone gets something different, we remember the one moment of synchronicity when two people got the same item. When there are 10,000 loot drops per second, it’s going to happen once in a while. When it happens fifty times in ten minutes to the same group of people, there’s a problem.

Noooo… I am asking a technical question!!! Please do not hijack this thread talking about how the brain works and math paradox problems. There is a real and finite answer to my question.

Please read the wikipedia link I added to the original post.

Allisa Wonderland.8192:

Please read the wikipedia link I added to the original post.

Right. If you did any programming before you know when you want to randomize something, the first you give is a seed.

For example if you want to pick a number between 1-10. You first need to give it a seed number say 234. And the random number generated will be base on that seed number.

But the output of the number is actually random base on the seed number. It’ll give you a random number with the same probability.

So if you rerun the program and give the same seed number the output will always be the same. But it’ll actually be randomized that it wont’ be noticeable by the human eye.

PRNG is not random, because the outcome is determined by the predefined number (seed).

But you make it sound like your coincidence of getting 2 truth is due to the flaw of PRNG, it is definitely not true. You wont’ be able to notice it.

Here is some link you can read.
http://en.wikipedia.org/wiki/Mersenne_twister
http://stackoverflow.com/questions/15500621/c-c-algorithm-to-produce-same-pseudo-random-number-sequences-from-same-seed-on

That being said, could the algorithm used to build those random number be really simple so it is less labor intensive for the CPU? Maybe. You need to ask the person making the GW2 game engine. I’m not sure if he is bored enough to answer you that.

Allisa Wonderland.8192:

Noooo… I am asking a technical question!!! Please do not hijack this thread talking about how the brain works and math paradox problems. There is a real and finite answer to my question.

The question will not be answered, because it involves revealing proprietary information about the program that Anet will not discuss in public.

What you are experiencing is an illusion based on the way we think about things. If there were a problem with the code, Anet has methods to test and correct this. For example, a while back players discovered a bug in the way ectos were produced when items were salvaged. I remember JS mentioning that he spawned and salvaged millions of items in order to test for the bug that was reported, and found that the process was indeed producing unintended results.

So if anyone at Anet thinks that loot spawns are bugged, they can use diagnostic tools to generate millions of items and look for patterns in the data that would indicate non-random results. Therefore, any anecdotal evidence can easily be confirmed to be a product of players’ imaginations and not a bug in the code.

Maybe it’s me, but I can’t imagine Anet telling you what type of RNG they use. It seems that it could be considered a trade secret and covered by an NDA.

Allisa Wonderland.8192:

Since the “discuss rng conceptually” thread appeared, I have put more time into learning about current RNG models.

I felt, anecdotally, that there were a number of instances in my own experience, and in forum posts, regarding “runs.”

For example, 3 of 5 members pf a partty I was in for Tequatl recently got exactly the same aquabreather, and none of us had seen one in quite a while before that.

I got two Truths in two consecutive bags in EoTM, two weeks previous to that.

Back to the question..

In my research, I learned that there is a statistically disproportionate chance of runs in some pseudorandom number generation models (apparently, Java uses one of them, for example.)

Is the model used by Guild Wars 2 using the generator of its language, custom designed, chipset seeded, or…?

John, I imagine you have the tools to do an analysis of the quality of the rng model for the game. Has this been done, or are you willing to? I do understand you may not be able to publish your findings…

EDIT: This is a real, twchnical question with a finite answer. There -are- issues with some random number generators. Please read this to educate yourself on the topic:

http://en.m.wikipedia.org/wiki/Pseudorandom_number_generators

Could it be that your examples are actually examples of the Baader-Meinhof phenomenon and not Anet’s RNG system?

http://en.wikipedia.org/wiki/List_of_cognitive_biases#Frequency_illusion

It’s interesting to note that even in GW1, it would quite often happen that the game would spawn two identical drops almost simultaneously, but for two different players in the same instance. It’s quite possible that they are using pretty much the same system as in GW1, with the same flaws…. or it could just be a whole lot of coincidences.

And when I say identical drops, I mean that the game would not only drop the same rare skin twice in a row, but with the same statistics.

Allisa Wonderland.8192:

Noooo… I am asking a technical question!!! Please do not hijack this thread talking about how the brain works and math paradox problems. There is a real and finite answer to my question.

Maybe posting your question in the thread i linked looks plausible to you now because thats where John answers questions.

Mad Queen Malafide.7512:

It’s interesting to note that even in GW1, it would quite often happen that the game would spawn two identical drops almost simultaneously, but for two different players in the same instance. It’s quite possible that they are using pretty much the same system as in GW1, with the same flaws…. or it could just be a whole lot of coincidences.

And when I say identical drops, I mean that the game would not only drop the same rare skin twice in a row, but with the same statistics.

I didn’t play GW1, so not sure what is going on there. Depend on how many stats combo there are, it does felt quite weird a rare item drop with the same stats if that happens frequently.

Allisa Wonderland.8192:

For example, 3 of 5 members pf a partty I was in for Tequatl recently got exactly the same aquabreather, and none of us had seen one in quite a while before that.

I got two Truths in two consecutive bags in EoTM, two weeks previous to that.

Neither of these are completely unexpected.

You seem to consider that you and only you, or your group and only your group, are the only ones accessing the RNG at any particular time. When you fought Tequatl, there were probably 50+ players in the area fighting him. Did you pool THEIR drops too? Did you check their drops in all past runs? And were all those aquabreathers the same weight? They do all have the same prefix and rune, but the different weight aquabreathers are fundamentally different items.

When a monster enters the map, the map tracks it in a list based upon its time of entry, and the same seems to hold true for players. What then makes sense is that treasure rolls are doled out in the same order of entry. For a massive event like Tequatl, these rolls may or may not be performed as a single batch. If not, then the rng function is being accessed by players/objects/events throughout the game. This really messes with your model.

As for the second instance, you are probably not the only person opening a bag at any particular time. Whatever the reason, you got lucky twice and got an exotic. Why the same exotic? That’s not random, is it?

Well, the loot table isn’t nearly as random as people like to think, particularly when it comes to bag contents, so getting doubles of a particular item isn’t really noteworthy. It happens with Masterwork Dire Krytan Spears of Ogre Slaying, but you don’t see anyone posting about getting two of those in their inventory, because who cares about salvage fodder?

It might also help to note that dropped Exotics are typically matched to a particular source of loot bag. One I like to mention is Lord Taeres Shadow, which you can get from champ spiders. The same is (probably) also true for any other class of loot.

The thing about random numbers is that you are guaranteed to eventually see ‘runs’. It happens. It’s the difference between ‘random’ and an even distribution.
The thing about humans is that they are designed to detect patterns.
So you will go months with getting random results with no discernible rhyme or reason, then one day you and your buddies all get aqua breathers, and boom, there’s a pattern and your brain raise all sorts of flags and alerts you to the weirdness that you should pay attention to.

In terms of how they decide the randomness, this is my guess:
each mob has a loot table of things it is able to drop, and each thing has a number range which decides how likely it is to drop
for example, let’s say cog is numbers 1-4, Bag of Experimental Materials is numbers 5-6, generic white item is numbers 7-8, generic blue item is numbers 9-10, generic green item is numbers 11-12, generic yellow item is number 13, generic orange item is number 14
then they use the generic random number generator function of whatever programming language they use to get a random number from 1 to 14. probably something like: $result = rand(1,14);
whatever number comes up is what item you receive.
obviously, the tables they are dealing with use much higher numbers. I would guess in the 9-digit range.

I appreciate that there are a number of very intrlligent and knowledgable people here, offering their insight into some of the reasons that i posted to provide a background to my question. Inhave read all of it and have’t learned anything new, in spite of it.

Perhaps i should have taken out ny anecdotes, altogether.

The closest i saw to an answer is that we won’t find out, because the information is proprietary. I also expected this, but was hoping to at least get a “no, i can’t say” or “sure, it’s this!” answer to the core question.

Please?

Mystic.5934:

The thing about random numbers is that you are guaranteed to eventually see ‘runs’. It happens. It’s the difference between ‘random’ and an even distribution.
The thing about humans is that they are designed to detect patterns.
So you will go months with getting random results with no discernible rhyme or reason, then one day you and your buddies all get aqua breathers, and boom, there’s a pattern and your brain raise all sorts of flags and alerts you to the weirdness that you should pay attention to.

In terms of how they decide the randomness, this is my guess:
each mob has a loot table of things it is able to drop, and each thing has a number range which decides how likely it is to drop
for example, let’s say cog is numbers 1-4, Bag of Experimental Materials is numbers 5-6, generic white item is numbers 7-8, generic blue item is numbers 9-10, generic green item is numbers 11-12, generic yellow item is number 13, generic orange item is number 14
then they use the generic random number generator function of whatever programming language they use to get a random number from 1 to 14. probably something like: $result = rand(1,14);
whatever number comes up is what item you receive.
obviously, the tables they are dealing with use much higher numbers. I would guess in the 9-digit range.

Your algorithm matches pretty well with my observations. I’d like to posit additionally that the items you get from a particular source are specific. We just had a massive amount of items come in via the farm in SW. I tried to post a random masterwork from my gains and saw the market was glutted for that particular weapon (some 800+ items at minimum value). Bonetti’s Rapier is rather famous for being farmable in SW too, enough to drive the price from 66G down to 6.6G

I’m not sure what you’re referring to as far as a 9 digit number, but there’s only about 38000 items listed in the API, and a good portion of those do not drop.

Allisa: look for red posts in the other thread, because I’m pretty sure JS has already said he isn’t going to divulge details on the RNG system GW2 uses.

Sariel V.7024:

Allisa: look for red posts in the other thread, because I’m pretty sure JS has already said he isn’t going to divulge details on the RNG system GW2 uses.

I try to read everything he posts and I don’t recall seeing it (insert complaint about forum search here).

java uses http://en.wikipedia.org/wiki/Linear_congruential_generator. It’ll repeat itself after 2^48 numbers (281 ,474 976 ,710 656).

Without knowing whether the sequence of numbers is used for an account, for an instance, for a map, for a server, of for some other group of players, this discussion is meaningless.

Even if there are runs, you don’t know who those runs affect.

Wanderer.3248:

Without knowing whether the sequence of numbers is used for an account, for an instance, for a map, for a server, of for some other group of players, this discussion is meaningless.

Even if there are runs, you don’t know who those runs affect.

Yes. I’m not looking for replication, I am curious about the quality of the numbers being generated and how that affects rewards potential.

It may not be a realistic goal to predict whether one success may be an implicit precursor to the next, but to me it is academically interesting to know is there is some statistically possible predictability to such an outcome because of limitations in the random generation algorithms.

Enough googling later and I have, at least, the answer to a) language and more importantly, b) most likely compiler. (Unless this has changed, but I would be shockingly surprised.)

Still more to learn…. This is more fun that speculating on the living story.

Allisa Wonderland.8192:

but to me it is academically interesting to know is there is some statistically possible predictability to such an outcome because of limitations in the random generation algorithms.

Do you have any computer program background?

laokoko.7403:

Allisa Wonderland.8192:

but to me it is academically interesting to know is there is some statistically possible predictability to such an outcome because of limitations in the random generation algorithms.

Do you have any computer program background?

Yes.

Allisa Wonderland.8192:

I appreciate that there are a number of very intrlligent and knowledgable people here, offering their insight into some of the reasons that i posted to provide a background to my question. Inhave read all of it and have’t learned anything new, in spite of it.

Perhaps i should have taken out ny anecdotes, altogether.

The closest i saw to an answer is that we won’t find out, because the information is proprietary. I also expected this, but was hoping to at least get a “no, i can’t say” or “sure, it’s this!” answer to the core question.

Please?

Also, those who say “method X is flawed because of Y, and if they use this method that could be the problem,” doesn’t make any sense. If it’s so well known that some random guy on the internet can find info about it through google, programmers wouldn’t use the method.

Anyway, there’s been a lot of talk lately on various forums about getting info from the devs. The thing a lot of people don’t seem to realize is that the Anet staff who communicate through the forums are not responsible for the policies they post under. If those who do determine policy say “don’t talk about these subjects, not even to confirm or deny rumors,” then they won’t say anything about it at all.

JS is one of the most open and active staff members on the forums. If he doesn’t answer a question, there’s probably a reason for it.

Allisa Wonderland.8192:

laokoko.7403:

Allisa Wonderland.8192:

but to me it is academically interesting to know is there is some statistically possible predictability to such an outcome because of limitations in the random generation algorithms.

Do you have any computer program background?

Yes.

So you understand there is basically no random number. The randon number is just the starting number going through a complicated math formula which generated the randon number. And that random number goes through the math formula again to generate the next number.

Weather there is predictability depend on the complexity of the math formula. For example Java use a linear algorithm. And the formula start repeating itself after 2^48 numbers. Which is astronomical big.

Most other programming language use Mersenne Twister. Which is better than Java’s linear algorithm. I presumeit is more complex so it take longer for the computer to process. Java also have a secureRandom function, which is slower but provides better complexity.

Regardless, most programming language “math formula” is complex enough you won’t be able to see runs after an astronomical list of random numbers. For example java’s number only repeat after 2^48 random number is generated.

If there is a pattern, it most likely occurs because for some reason the coder somehow use the same seed number in their coding. For example 2 process use timeoftheday as seed numbers at the exact same time. Or the pool of seed number the program use is very small.

GW2’s game engine is home made. So how GW2 design their loot table, only they know. I dont’ know much about game engine so nothing I can add to that.

Allisa Wonderland.8192:

Sariel V.7024:

Allisa: look for red posts in the other thread, because I’m pretty sure JS has already said he isn’t going to divulge details on the RNG system GW2 uses.

I try to read everything he posts and I don’t recall seeing it (insert complaint about forum search here).

Like other forum posters, you can click on JS’s name to the left of his posts to bring up a forum profile. Then you can select “see all of this forum user’s posts” to look through everything he has posted.

laokoko.7403:

So you understand there is basically no random number. The randon number is just the starting number going through a complicated math formula which generated the randon number. And that random number goes through the math formula again to generate the next number.

The term I’ve heard used is “sufficiently random.” Which is to say that situations like the anecdotal evidence “my buddy and I got the same loot” is not because the random number generator is flawed, but is a result of our looking for patterns where there are none, or at least where the pattern is so complex it would take several decades of constant effort to observe it.

At any rate, the simplest computer algorithm is far more complex than random numbers chosen by a human, since humans would be reluctant to choose the same number twice, choose numbers that are near each other, etc. So if a person chooses a string of five numbers from 1-100, it is unlikely he will choose something like 1,2,3,3,99, while a random number generator is as likely to choose this string of numbers as any other.

Alisa, I’m glad you have found some answers to your questions elsewhere, I hope it satisfies your curiosity.

The times I’ve seen clustering has been when I sell some item whose trading history indicates a relatively stable supply that fluctuates in a narrow range and an hour or two after I post mine for sale, there’s an atypical spike in supply with multiple prices significantly undercutting mine.

As if my slight undercutting of the previous stable price opened the floodgates causing hoarders to dump their supply OR it was just a slightly more common drop that evening.

laokoko.7403:

If there is a pattern, it most likely occurs because for some reason the coder somehow use the same seed number in their coding.

That would actually be my biggest concern if I was in their shoes.

GW2 generates a lot of random numbers – every attack has crit and damage rolls, plus loot drops on kill, all the procs, rolls for the AI; you get the picture. Generating high quality random numbers is not anywhere near as important as generating lots of random numbers very, very quickly. Forget using a twister, you need speed.

It’s probably multi-threaded, because I cannot imagine throttling all the rolls they need through a single locked thread. Which means they need a fast generator with a whole bunch of different seeds, which sets up a risk of parallel threads randomly overlapping and kicking out the same sequence of numbers.

How you set those seeds is really important, and not at all easy. I wonder how you’d do that to avoid overlapping runs; it’s not like you can just generate random seeds for all of them without some risk of one landing in a sequence of another. Maybe there’s a trick to that, or maybe you just throw your hands up and not care and occasionally get parallel runs kicking out the same numbers.

Ensign.2189:

Which means they need a fast generator with a whole bunch of different seeds, which sets up a risk of parallel threads randomly overlapping and kicking out the same sequence of numbers.

I don’t think that matters. Even if two different players received the exact same sequence of numbers, those numbers will be being used for different things, depending on the players actions, and the results are still going to look fairly unpredictable anyway.

You don’t need a cryptographically strong random number generator when the numbers are being applied to unpredictable player actions. All you need is one that generates an approximately uniform distribution in the long term.

Illconceived Was Na.9781:

Generally speaking, the level of non-randomness seen in pseudo-RNG is so small as to be unnoticeable by even the posters in this thread pooling all our data. Coincidences happen, so it’s not proof of anything that the OP has seen several of the same items popping up among their acquaintances and friends.

That said, OP is correct in suggesting that the choice of pRNG tool could matter, especially if seed selection isn’t properly managed. And OP is also correct in thinking that ANet chose one tool or another and someone could tell us. OP is also probably correct in thinking John Smith would be a good person to ask.

However, I can’t imagine why ANet would give us that sort of information. Those who don’t understand pRNG or probability can’t make use of the knowledge; those who are intimately familiar with the details might be able to take advantage of such data. But more importantly, the vast majority of us, who only think we know something about the topic, would draw incorrect conclusions based on insufficient information about how GW2 actually works. As Trahearne used to say, “this won’t end well.”

tl;dr it’s most likely that the question is moot and even if it weren’t, it’s not in ANet’s best interests to tell us.

Thank you for being the one person to clearly demonstrate their ability to clearly comprehend and respond to my question. So many voices that appear to just love to hear themselves speak..

It was refreshing to see that there are a few people educated enough to discuss the performance benefits of non-cryptographic quality pRNG’s..

Guild Wars 2 having a bespoke engine in no way proves that it has a home built random number generator.

Anyhow, in closing, there are only two likely answers to the question, and drilling down to the time period of the original game’s launch, probably only one that makes any sense.