Diablo 3 RNG. When and How are Item Stats Determined?

We seem to be approaching peak Diablo 3 DiabloWikiRNG, as fan speculation about how and when and where items are created grows by the day. Lylirra replied to a grand conspiracy about it on the B.net forums with comforting words about randomization that will make absolutely no impression upon anyone who is predisposed against them.

There’s a theory floating around that you should roll several characters of the same class because their drops are seeded when they are created, meaning Wizard A will get different drops than Wizard B or C. This is an interesting theory and one I’m will to test because I started wondering why some people in my clan seem to have all the “luck”. Going to power level a few wizards and just alternate playing them (since paragon are account wide, just have to swap gear – only difference will be their name). If this is correct, it’s absolutely crazy. Why some characters would have a predetermined drop table that they can pull from sounds like something out of a conspiracy theorists journal but I’m willing to give it a shot. My friend plays 1/10 the time I play and has gotten 3 rings of grandeur and I have yet to see my first…
Lylirra: Hey all!

For clarity, items and their stats are determined at the time that they’re generated in the game world. In most situations, this is when the item physically drops (like from a monster, a chest, or a destructible). In cases where there isn’t a physical drop, items and their stats are determined as soon the game creates them. For example, items from Kadala are generated as soon as they’re purchased. Item types, stats, drops rates, etc. are not determined, or pre-determined, upon creation of a character nor is that kind of information stored in some sort of seed on the character itself.

It’s important to keep in mind that random != uniform or equal distribution. It’s completely natural for random systems to included clusters, and for the human brain to interpret those clusters as patterns (even where there are none).

Thanks for the discussion, though! Theories like this are always interesting, and of course can yield great feedback on and insight into potential related issues. 🙂 I’m going to go ahead and lock this thread up, though, since the theory isn’t actually accurate and I’d like to help curb some of the misinformation.

In parting, for your trouble, I’ll leave you with this.

I won’t repeat my comments from the legendary drop survey post earlier this week, but this sort of thing will continue forever while Diablo 3’s item generation codes remain out of reach of players.

There were countless conspiracy theories and rumors about why and how drops worked and Magic Find functioned in Diablo 1 and Diablo 2, but ultimately fans poked through the code and figured out exactly how item generation worked. That was satisfying, but far from reassuring since numerous bugs and possible exploits were revealed, and there were many ways players could gain an advantage or cheat the system, beyond simply “kill monster X repeatedly.” So with Diablo III, 1) we don’t have access to the code to see if it’s actually working as we’re told, and 2) we know that every piece of software has bugs, especially in regards to random number generation.

I don’t think theories like the one offered by the OP in this post are reasonable, since it would just be dumb on the part of the devs. Why would they put some code on a character that forever limited their possible item finds? However that doesn’t mean there aren’t other bugs or issues that do affect randomization.

Item drops in Diablo 1 were determined by the time the game was created. The uniques possible via gambling in Diablo 2 were set upon game creation and players learned to use packet sniffers to find out of they could gamble X unique in that game. Numerous item types and legendary items in Diablo 2 were undroppable, or 1/1000th as common as they should have been due to errors or mistakes in the coding.

It’s not only possible, it’s probable that similar issues exist or have existed in Diablo 3’s complicated item drop game code, which is why fans are forever attempting to ferret them out.

Destined to Conspiracy

This sort of thing is simply our lot. We are a poorly-evolved mammalian species with brains ill-equipped to accept chance and coincidence. It’s why all human societies have always invented superstitions and witchcraft and gods and curses and so on. We see weird things happen and we want to attribute them to a cause or an agency. That’s true for the real world, but especially true for video games, with good reason. Video games actually are intelligently designed pocket universes, where things do happen for a reason. Either a direct “X leads to Y” function, or else because of a system designed to keep the virtual world running by it’s quasi-random rules.

Tagged As: | Categories: Blizzard People, Blue Posts, Items, Legendary Items


You're not logged in. Register or login to post a comment.
  1. In the end, what matters is the base, the random numbers generator is built upon. Most methods are not producing true random numbers, thus have tendencies to produce reliable patterns, thus loopholes.

    If Blizz worked in user input on some level, you may bring it to a degree near to real randomness, but leaves the rng vulnerable to manipulation through the player, thus cheating. (And once the input chosen cannot be hidden anymore, your game is fracked.)

    And if they really got it done and produced a rng delievering real randomness, then there’s the question, if it’s near enough to true randomness that the accumulation process*, that’s nowadays proven to be a working concept even at a simple, statistical level, is already kicking in. Although I have no idea, how this could possibly translate into game mechanics, I’m assuming that Lylirra meant exactly that when typing “It’s completely natural for random systems to included clusters, […]”

    *This process is known to most through the phrase “the rich get richer”.

    • Small correction:

      […], but is leaving the rng vulnerable to manipulation through the player, thus allows for cheating.

      Sorry for not rereading before sending off … 🙁

    • The reason why the rich get richer is true in D3 is because rich people get better gear which in turn let’s them farm faster and more efficiently. It has nothing to do wth RNG. While it’s true that computer generated numers are not random in the strict sense, it doesn’t matter for our slow (when compared to computers) brains.

      • Better gear leads to getting even more better gear faster. Sure. Simply by the games design. That’s so Captain Obvious that I’m asking myself how you could even think that anyone in this thread could have forgotten it. And as you’ve said it yourself: It’s offtopic.

        On the other hand: It’s only a matter of good (iterated) programming to provide a program with real randomness. The only limit in reaching true randomness being the issue of only having 1 and 0 at it’s base. But even if this would be reached, it’s only a question of how big the basis on your statistic is that limits the ability of drawing elaborate conclusions. Once you’ve gained insight on the type of pattern emerging and how that tranlates into the game, you can easily find out ways to optimize your drops.

    • Most RNGs produce reliable patterns indeed, but unless Blizzard was wildly incompetent (which may or may not be realistic) the period will be way longer than any players should be able to meet in a play session. For the most used pseudorandon numbers generator, Mersenne twister, the period is 2^19937, which translates to a number around six thousands digits long. Even with a much less robust approach to random number generation, the odds of players being able to notice and use a pattern is close to non-existant, especially when these calculations are server-side

      If Blizzard works in user input, and once again assuming they do it in a meaningful fashion, they do it in a way where players can’t influence it. They could, for example, use the millisecond you open the rift as a seed. With a lot of practice and luck, a player can possibly time it to open a rift in the 1/100th of a second of their choosing, assuming stable latency. 1/1000th on the other hand, is impossible, making the seeding impossible to abuse

      I read Lylirra’s quote about clusters as simply indicating that in the sample sizes we typically see inside the game, data points may appear clustered (that is, non-randomized). You will get in a game where legendaries drop like snow outside my windows right now (it’s May dammit!), while at other time you will go several rifts without seeing anything orange.

      • You’re right. An individual player couldn’t possibly produce a statistical base big enough for patterns to show and depending on the method (or methods) used, may only allow for discerning tendencies, even if all the millions of players would collaborate on building up the statistical base.

        But I think you’re wrong on the influence of player input through measurement of time being stretchable for long. The lower the possibilities of a player hitting a “jackpotrift” (or game, for that matter) are, the more it accumulates frustration instead of providing more fun to the game. And even taking greater margins to account, linking droprate to a certain timeframe only ammounts to being patchwork providing nothing more to the game than hiding the other mechanics (, botched or not, ) behind a curtain.

  2. I remember watching a show about RNG in which is was seemingly proven that the human brain can intervene with RNG iirc. It showed signifact differences when something big was about to happen (9/11, the tidlewave hitting in the pacific/Indian oceans). Im a bit foggy on the details but its interesting at least 😀

  3. Some of the theories really look like “dark ages” material, in a sense that I’m more inclined to believe we’re living on a flat planet and, therefore, cruising ships are dangerous.

    On the other hand, some theories are so disturbing and match with my results so badly that I can actually believe on them – or maybe the theorist know a thing or two about mod making, game coding and made a theory that looks good who knows…

    I don’t know what’s the reason behind the secrecy about the foundation of drop system (they can’t explain everything in great detail to prevent exploits and hacks, but a post with there aren’t little loot tables; if you kill a white mob anytime, anywhere, with any IP address, you can get anything – or not).

    The scariest thought is that maybe not even they know everything (as weird as it sounds, as in “not an exploit, but not working as intended in a serious bad way”). Did they stress tested/simulated X million drops to see if their random is really random (in the way random use to behave on random… randomness)?

    I think this simulation was possible on D2 after a bunch of hours (not millions of thousands, but enough to get meaningful results. If they didn’t simulate at all or enough, maybe there’s a “cause”…

    Yes, I’m talking about something I don’t know, but I think that true randomness actually tends to a certain pattern or distribution – 10N Blackthrones, 2N Tal’s 1N Vyr’s chests dropped after a certain amount of time – not just Y BT Z Tals and X Vyr because random, since random isn’t complete chaos beyond chaos theory. If the sample size is large enough (millions of players doing thousands of rolls), some stability would show up, making builds possible. “Stochastic convergence” is what I mean – I think (I’m really bad at anything that looks like math). Maybe there’s something on the RNG to delay this, to avoid the “everyone having the same gear” phenomenon, maybe not.

    Some people are calling it bo-ring anyway… And I think my WD is cursed and my wizard isn’t.

  4. I always told people in my party to completly leave the game and restart when they weren’t getting legendaries. It worked so many many times that I personally don’t believe it’s just coincedence.

    I believe the drops are determined the moment you “create” a game.

    • Yeah i dont believe in the conspiracies, but i have had similar anecdotal results. I believe in ‘lucky games’, where, if i’m getting a couple legs already, i’m better off running other areas in the game because they keep coming. Conversely, if it’s a dry spell, I’m better off exiting and creating a new game.

      I have also noticed the pub game theory where one player gets most of the leg drops, having been on both sides of that. It’s happened too often to ignore.

      But, like Flux mentioned, we’re prone to find patterns where none exists so… I’ll just end with RNG is RNG. Where ‘R’ is not truly ‘random’ but a random selection from a predetermined array of possibilities. 😉

      • After reading I’ve tested this for myself and think that this is not far from the truth behind rng. Whenever I got two of more legs during one acts bounties, running Rifts afterwards provided about 3 to 5 legs each three Rifts. And getting one or no legs during the bountyrun, Rifts afterwards never provided more than two legs for 5 Rifts. (Torment 1; 64% mf; WD)

        BUT: The latter provided me with better rolls allover. Thus, if D3s programming really results into creating games, where RNG likes you and games, where it doesn’t, really testing it out should include testing for the possible situation of there being different ways, RNG could like you.

        For Light or any other, selfproclaimed detective: This is purely speculation, as in weighing possibilities. It says nothing about how the Deathnote works…

  5. I’d best try get into the sweet spot of those clusters then :D.

  6. As one of my clan members said, ‘Do you define soon the same way Blizzard defines soon ?’ How about legendary? There are many other words I could use but you get the point. So is Random random or is it random?

    • “So is Random random or is it random?”

      It’s not perfect. Same as everything else in the universe.

      • We are not discussing perfect or imperfect. We are discussing defining terms. A word means something (only if you are not a relativist). If it doesn’t mean something then language is of no use. I am saying that Blizzard is defining something as random when it and the whole system is NOT RANDOM. It is ok if they don’t want a random system. Thats fine. But don’t call it random when its not.

        • I’ll rephrase and elaborate then…:

          Does it really matter?

          If there was lazyness involved in trying to reach randomness, resulting in a rng that ain’t random, then it would be just all too human, resulting in bonus sympathy points for the programmers. Still saying that it is random only further substracts from the sympathy point pool of the marketing dudes, which is still far away from climbing back into the positive values to begin with. Thus, from my point of view, it would simply even out Blizz-wide.

        • Just to be sure: Are you really talking about randomness? Or are you talking about equality of drops? Because randomness naturally has the innert quality of producing inequality. And to reach an even item distribution amongst the players actually means taking out randomization by steering/controlling the results of RNG back to a acceptabel degree of equality.

Comments are closed.