How does the so called "random generator" for minipets work?

Off topic, but I remember in Hich School we had one of those stupid "lock-ins" after graduation where they had a mock Casino and this guy tried to convince us that keeping track of the numbers that came before in Roulette made sense.

We told him, no, it really doesn't matter, but there's no convincing some people.

Ok. First of all that wasn't really a complaint/whine. I haven't sold any of it except the Rurik 'cause I don't like him - therefore I can't complain about getting profit (which is why people want gold/green mini pets most of the time). I know people who've gotten like 3 turtles, 2 wallows and like 4 whiptails throughout their account, of which they haven't played for 2 years yet - now THAT is crappy. Sure I got it balanced pretty much so far, but yeah...

Perhaps I should've talked about more of the ODDS of getting stuff too... whether you get 3 Gwens or 3 whiptails, something's wrong IMHO. Base it on luck (or lack of...) or mathematical stuff, you still get three of the same kind. I'm guessing it's a normal average (roll a dice a few times, see what you get - seems 1 through 4 come rather often, chances that you get 5-6 still happens, but seems less often... yet you still have 1/6 chances of getting each number) but there's something wrong when someone keeps getting the same number all the time.

I guess a system that could be implemented is that, if the mini pet works via number generator, depending of the way they work too (account? character? other way?), to remove that number from your character/account so you stop getting duplicates/triplicates, etc. Removing a name from a hat really; can't be gotten twice. Maybe I'm seeing too big for GW really...

I'm not made for maths really >_>.

/endrant

Flip a quarter for 50 times, and tell me how many times it comes up tails or head in a row. (if it is truly random, it should have series up coming up all heads or tails at some point).

Patterns in randomness is random. It's counterintuitive, but if I would trust less a system where I never got the same pet, than a system where I get 2 or 3 of the same pet. (ie, if I never got the same pet, then it's not random).

http://www.cs.sunysb.edu/~skiena/jai...pts/node7.html

5 hydras in a row, random my A$$.

You're just really lucky (or unlucky, depending on your point of view).

Again, people who get 4 or 5 of the same pet are more likely to report it here than people who haven't, so these "testimonials" are pretty useless from a scientific/mathmatical point of view.

The fact we all aren't getting 5 of the same pet is equally telling.

Math: streaks are a part of any random series. Read up on statistics if you don't understand that. Eventually someone in GW should get 10 in a row if the generator is truly random, although the probability of any one of us being that person is astronomically low.

On a computer note, I remember reading that a computer can NEVER generate a truly random series of numbers, but it can come credibly close to doing so. Comps are limited to preset mathematical formulae, whereas analog devices such as dice are limited only by physics. Or something like that.

Luck: My wife's last two mini's were a Rurik and a Lich. Mine was a(nother) fungal wallow, and the rarest I've ever gotten was an elf (purple).

Elfs are gross.

Wheres the SMITE CRAWLERS?! I want that! QQ

First it checks to see if the mini-pet is going to my account. If it is it generates a Fungal Wallow. If not, it does some other random thing.

LOL... poor arcady.

Best post in this thread

True, but the incident happened to some people (like me) when we opened the presents within a short timeframe.
This arose the suspect that the result of the package could be related to the opening time.

The thread was not started like "one month ago i received a troll, after 2 months another troll, after 8 months another, OMG it's a bug".
You and other are answering as if this was the subject, but sorry we're not talking about that.

The thread reported a situation of a streak happened in an interval of one hour.


And we both are discussing about suppositions, since no clue comes from the developers of the software.
A group supposes that the GW software doesn't take time/date parameters into account and behaves like a random generator, another group is not fully convinced.

Someone from the dev team could have simply clarified that there's no link between the opening time and the result, that the GW software doesn't work this way.
Unfortunately, this didn't happen, and it's very unlikely that will happen.

How about we buy lots and lots of 1st year presents and make some experiments?

Whos got cash for 10,000 or so?

Well said. People who always got a different mini won't complain....

I don't think that test is conclusive that loot is determined upon entering an explorable. For one thing from what Anet has said in the past I sorta assume that you get the equivalent of your own Virtual Machine running your instance that is forked off from the main server. That is, when you entered the explorable you got your own copy of GW1.1.334(holiday) running basically in a sandbox. Then when the main program is reverted back to GW1.1.334 your instance hangs around until you left it and the garbage collector returned your resources. In this case you and the other person are *not* running the same instance so you can not compare each other's drops, loot could still be generated upon a kill. This is why when you zone back into an a town you do not see the people that logged out and re-logged into the different build.

Further, even if not totally true (and I suspect that it isn't 100%, probably something between what you describe and what I do) they could still have drops randomly selected upon kill but have a loot table (or tables) statically created when creating an instance. In fact, I would guess this to be the way it works as it is about the easiest way with the most flexibility (and probably the best in terms of scalability too). It would be trivial to have the loot determination method walk down a list of loot tables until it hit a NULL and have them either tacked on to the last list or have a separate roll . We know that bosses sometimes drop multiple items so there pretty much must be someway that they do this. This would also allow some interesting tweaking of drops based on location, species of critter, individual critter (read boss), critters around the one you killed, or any combination of that. IIRC we do see that in places.

*shrug* then there is always the possibility someone came up with a *really* nifty idea and we are both grossly incorrect

I would think that such an occurance would be sufficently enough to say there's something wrong with the random generator. There's such a thing as being too improbable. Logically speaking, yes there is a minute chance of something like that happening once, just like there's a chance that a bunch of chimps with computers will type out one of shakespear's plays, but I think if something like that happened, you'd really have to check it out.

At some point the odds of the random generator being broken are more likely than it to keep spitting out the same number. At that point you double check it.


BTW, since we are dealing with computers, it's only a psuedo-random number.


PS. The above quote is why I hate trying to talk to people on this subject, there are some that will blindly accept that what is happening is random more likely than the code being broken.

PPS. Lyra if you were playing dice and betting 1 dollar per toss, how many snake eyes in a row would it take you to wonder if you were playing with loaded dice?

I don't think that test is conclusive that loot is determined upon entering an explorable. For one thing from what Anet has said in the past I sorta assume that you get the equivalent of your own Virtual Machine running your instance that is forked off from the main server. That is, when you entered the explorable you got your own copy of GW1.1.334(holiday) running basically in a sandbox. Then when the main program is reverted back to GW1.1.334 your instance hangs around until you left it and the garbage collector returned your resources. In this case you and the other person are *not* running the same instance so you can not compare each other's drops, loot could still be generated upon a kill server side as nothing in you universe has changed. This is why when you zone back into an a town you do not see the people that logged out and re-logged into the different build.

Further, even if not totally true (and I suspect that it isn't 100%, probably something between what you describe and what I do) they could still have drops randomly selected upon kill but have a loot table (or tables) statically created when creating an instance. In fact, I would guess this to be the way it works as it is about the easiest way with the most flexibility (and probably the best in terms of scalability too). It would be trivial to have the loot determination method walk down a list of loot tables until it hit a NULL and have them either tacked on to the last list or have a separate roll . We know that bosses sometimes drop multiple items so there pretty much must be someway that they do this. This would also allow some interesting tweaking of drops based on location, species of critter, individual critter (read boss), critters around the one you killed, or any combination of that. IIRC we do see that in places.

*shrug* then there is always the possibility someone came up with a *really* nifty idea and we are both grossly incorrect

Hmm ... that's an interesting proposition for sure, a 'Santa Claus' mechanism where the overall contents of the sled are determined by the time when SC starts his journey and individual presents are pulled from the stack of possible alternatives on each stop. This might be an ideal mechanism for another type of loot dropping 'monster' called a chest.

However, some monsters possess a variety of discernible equipment and have a tendency to drop the exact equipment that they are carrying. I've done enough Fahranur HM rounds to see right away what the next mob has in offering. So clearly at least some monsters have 'carrying capacity', and it wouldn't thus be too far-fetched to assume that all monsters have enough internal state for personal belongings (the fact that otherwise identical monsters can have different equipment also gives a reason for the somewhat resource intensive determine-during-instance-creation mechanism). And since that state is set for some monsters when the instance is created, it would be the most logical choice to seed all potential drops at the same time. Of course one can have a kill-time algorithm that first checks whether the monster should drop its carried equipment, then if not proceed to generate random other loot. I'd just find such a solution less elegant.

I wonder if one could device further experiments to probe the alternative mechanisms. It's a nifty but ultimately pointless issue in any case, just the way I like it

I'm really tired of hearing this. The number being pseudorandom has got NOTHING to do with why people are getting duplicates. Nothing. This is why: There are only 14 minipets. That's it. That's all there is to it. There are only 14 minipets and only 8 of them have any real chance of spawning. When anywhere between 500,000 and 1,000,000 people open 2 or 3 presents each and there are only 14 to randomly pick from, you are going to get a lot of duplicates. We're not 'blindingly accepting' anything, thanks. If someone was to put forward actual evidence that something was broken, I would be right there with you wanting an answer. Unfortunately no-one has done that yet. They've put forward a lot of anecdotes that are entirely expected. Not just likely, but expected. Mathematically, thousands should be getting 3 the same and hundreds should be getting 4 the same. It's going to take more than a few people on guru to convince me that there's something broken.

the white mini pets - instances or the random ness of them occuring is higher, so if you say there are only 14 minipets to choose from I understand the dart boards chart by lyra_songs I also understand your theory of there's only 14 minipets and some or most are bound to repeats,

but does this not also means since the instances of the rare ones occuring are much lower, so explain the incident where one account has 4 mini gwen? "it just should not be if the randomness is calibrated correctly" right?

*edit with more of what i am thinking about the randomness.

in my opinion, if the randomness is distributed correctly, it should ideally be most people get white, and sometimes repeats (which sucks) and then once in a while one lucky account gets 1 green/rare one, not one account continually getting 10 hydras and another continually getting the greens mini.

Ahh, I had forgotten about that too - you used to be able to tell if some bosses carried their greens by visually inspecting them also. Alas that is no longer the case - it was neat being able to do that from a RP perspective. In fact I wish *all* critters did this.

But then, still easily doable under my system. I had more or less envisioned a base class that had it's destructor call a method that reads a pointer (or a reference) to a loot table (or linked list of tables) randomly for it's drops. No reason why a specific leaf class can't have it's constructor randomly overwrite whatever they use for a loot table with a a table with one (or a very few) entries along with making sure the correct skin is displayed instead of pointing to the instance global one. That is just as elegant a situation as what you describe and is VERY minor to implement. That would be the same way a "Corsair" would add something like a "Corsair's Pauldron" to it's drop table (or other enemy specific drop) except add it to the loot table linked list.

The problem is we are back to Schroedinger's cat - from the viewpoint of the observer the state of the cat has always been one thing and one thing only and we are arguing if the cat was always in the state we observed or in a superposition. We can only do the initial observation once and there is no real way to know what might have been. If we ever can prove that the other universes exist we can query them and see if they get different drops based on when they observed the outcome

To some extent if they used an object oriented language and used a fairly decent OOP design then I suspect I am more correct than not - it's been to long since I needed it to remember the name but there is a standard pattern close to what I describe. If not, then it depends on what they used, that level of abstraction is only easily done if you have the compiler (or the run time as some of it is dynamic typing) manage the typing information and make the appropriate calls through a jump table. I would then do closer to what you describe.

I'll have to think on it if I can think of any other way to test (it's my bed time and I have consumed a moderate quantity of alcohol). Obviously *some* part is done when the instance is initiated, yet I just feel that having all the drops calculated at the start opens it up to some abuse. I can't think of anyway to do so, yet that just sets red flags off to me and I rather suspect it would for others too. Even back in the early 90's when I found game programming interesting the books I had on it HIGHLY suggested you not do such a thing.

Let's calculate some probabilities based on the facts that for each yearly series there are 8 whites with an assumed probability of 9% each (72% total), 3 grapes of 7% each (21% total), 2 rares of 3% each (6% total) and a single green of 1%. For three consecutive presents opened the likelihood of some outcomes are:

3 whites: 1 in 3 players
3 grapes: 1 in 108 players
3 same whites: 1 in 171 players
3 fungal wallows: 1 in 1372 players
3 rares: 1 in 4630 players
3 greens: 1 in 1000000 players

Yes, considering the size of the player base, there is a large number of people who get all rares out of 3 presents, and probably at least one lucky duck who gets 3 Gwens in a row. 3 same whites in a row is so probable that there may be someone in your guild who got that - to find one in your alliance is almost a certainty (unless you are in a very small alliance).

I call bs on this, unless some of those have been acquired from outside the account (other accounts of the same player or other players). At the moment any account can have at max 4 characters who have had their second birthday. Furthermore, it is quite uncommon that all those 4 slots would actually contain a character that is now over 2 years old. I'd say that as an order of magnitude estimate there are currently no more than 10000 accounts with four 2+ years old characters (and most of those are probably mules). 4 Gwens out of 4 presents is highly unlikely given the sample size. But if the player has several accounts it's a different issue altogether.

Imagine a six sided dice.
Paint three sides of the dice white.
Paint two sides purple.
Paint one side green.

Now roll the dice four times. It's highly unlikely that you'll get a sequence of four green results.

Now get a million people to roll the dice four times each. With such a large sample size its reasonable to conclude that (order of results being irrelevant) every combination of results will be represented. Or simply; its highly unlikely someone won't get four green results.

---

I know I've oversimplified by the way, a fourteen sided dice would be more accurate and accounts can have more or less than four 'rolls of the dice' but the model still holds up. Given that the sample size in reality is considerably more than one million, I'd be surprised if it was just one account.

Also, it's worth noting that while a software RNG can only be psuedorandom, the results of psuedorandom generator are indistingushable from true random numbers (within a reasonable degree) unless you know the original state of the algorithm and while this method will create recognisable patterns and does not consititute the irreducable indeterminacy of a true random sequence, the degree to which this happens in a capable software PRNG should not significantly impact the viability of the output in any reasonable timeframe; i.e. the next million years or so. The only viable criticism of PRNGs is that if restarted with the same original state, the same sequence will be generated.

I like the six sided dice explanation better... lol thanks for explaning, I am lazy to look up words at the moment, so i'll leave the psuedorandom and PRNG for later, looks like a file extention to me... lol, but I understand the explanation.

it means out of the million people rolling the dice, one of them is bound to have the luck of repeatedly rolled and get the dice to land with the green side on top. right?

As an example, you can visually inspect skeletons in Fahranur for what they are wielding, and what they may also drop. A specific case is the Dead Sword that is always rare and thus exempt from loot scaling. If you see a skeleton with a Dead Sword, killing that skeleton will usually produce a Dead Sword. Furthermore, monsters are wielding actual weapons and not just skins with generic stats, as demonstrated by the occasional vampiric damage (monsters don't get the -1 degen, though).

Until you can show me REPEATABLE, & TESTABLE proof to the contrary, i dont see the point in discussing it as if it was broken.

I'm not blindly accepting it as random, especially when its a computer based system. I'm also not blindly accepting that it is broken based on someone's personal supposition.

Whats the chance of getting a critical hit? (Non-Asssassins) You can swing 10,000 and find out how the actual % ends up being.

Make a chart, open up 10,000 gifts and lets talk statistics.

I dont bet on games of chance. I only bet on things with skill, or at least some semblance of skill.

Dice tossing, loaded or not, is not skill based.

Not that I dissagree with you on your other points, but, Dice tossing is a skill. Depending on the position of the dice in your hand you can predict the number of rolls upon hitting the table therby giving you a forknowlage of the outcome. This requires an incredible amount of skill to control the number of rolls the dice will make, but it is doable.

As far as the random pet problem, its working as intended.

My guess on the working of the system is more likely a pass fail roll vs each pet starting with the most comom ones.

Example: When you click on your present there is a 70/30 chance it will be a whiptail. If you pass that you get a 65/35 chance on a fungal wallow. This continues untill you reach the rarest of pets at which time you either get the pet or are sent back to the start once more. Just a guess but it does make sense.

Dunno, sound's a little over complex to me. The simplest way to determine would be for a software random number generator to generate a two digit number and compare the result to a table where, for example:

00 - 09 Aataxe
10 - 19 Imp
20 - 29 Harpy
30 - 39 Heket
40 - 49 Jugger
50 - 59 Thorn Wolf
60 - 69 Mandragor
70 - 79 Wind Rider
80 - 84 Elf
85 - 89 Koss
90 - 94 Palawa Joko
95 - 96 Lich
97 - 98 Djinn
99 Gwen

I'd imagine the odds aren't exactly as produced above, generating a 3 digit number is probably more accurate given the amount of Gwens you see compared to Wind Riders. I'm also assuming that all minipets of a given rarity have an equal chance to appear while this may not be the case.

Regardless, the model holds up

EDIT:
Right