No. I’m just going through the game trying to 100% each area. I just finished Pandora.
Agree
The PS4 clears it’s own cache when you turn it off.
Are the drops random?
Yes. Anything that can world drop can appear from any enemy.
But, rather like BL1, some legendaries are more rare than others. The BL1 Volcano rained from the sky whenever the player sneezed. The Reaper was arguably rarer than pearlescents.
Similarly, in this game, guns like Woodblocker and Kill O Wisp appear relatively frequently, but a world-drop Laser-Sploder or Scourge is quite unusual.
My point is that the drops are “random,” but not all of equal probability. Those aren’t the same thing.
3 Likes
So instead of RNG people might say ‘it’s just proportional probability…’
RNG is not purely random. It is a term to describe psuedo random mathematical algorithms… Thus probabilities are weighted by devs… Always will be…
And this explains it better than I can.
3 Likes
If it was truly random we’d have better odds 
2 Likes
Join the club, I have four players two complete games on each, and have had the same problem and the drop are terrible drops.
2 Likes
So, it is true that RNG isn’t truly random, to the extent that the algorithms used to generate numbers are deterministic. That doesn’t mean they are necessarily weighted. In the case in the article you linked, it sounds like the Aion devs used a bad algorithm, or the player in question was cheating.
A proper pseudo RNG algorithm shouldn’t have patterns that are trivial to recognize. Typically, to generate a number between 1-100, you would first generate a random number in a much larger range (depending on the algorithm, it could be between the minimum and maximum number that can be stored in an integer in a particular programming language, or perhaps between 0 and the max). Then you would do a modulus operation - basically, divide by a number and take the remainder. You’d then add 1 to get your result, since in the case in the article we want numbers between 1 and 100.
So, if the pseudo RNG algorithm generated 102:
(102 % 100) + 1 = 3
Or, if it generated 99:
(99 % 100) + 1 = 100
So, if they use a good algorithm it should be very difficult to discern a pattern, even if you know what algorithm they were using. There would be many different ways to generate the same several numbers between 1-100 in a row. However, I suspect there may have been something else going on in the case of the linked article.
Here’s the thing - when saying a machine is deterministic, it means it will come up with the same sequences of numbers literally every time. It is impossible to build a computer that has truly random number generation. The best we can is look at randomness in the natural world and use that as a proxy. So, for example, a “truly random” algorithm might look at the last few digits past the decimal point in the barometric pressure at a local weather station. Hypothetically, one might take advantage of the random properties of quantum mechanics, but that’s a bit outside my scope of knowledge.
Instead, what we typically do in almost every case is “seed” the RNG algorithm. We use a number called a seed - most often derived from the system time - to determine where in the algorithm we start out. Given the same seed, the algorithm will spit out the exact same numbers every time. Hypothetically, if you had access to the seed and the algorithm used in a particular game, you could predict the exact pattern of numbers that would appear.
My guess is that in Aion incident in that article, one of two things happened. Either the RNG implementation was really bad, to the point that there was a short enough sequence that a human could recognize it. This is possible, but NCSoft didn’t use a very good algorithm if that is the case. It is also possible that people were taking advantage of things that were done client side. If the seed was stored client side, you could write a program that would give you the exact sequence that the game was going to produce with a bit of guesswork about what algorithm it used.
Having said all that, I expect jmthsauer’s explanation is probably correct about why some items seem to be weighted more heavily than others. It is however quite possible to write code where item drops are equally likely to occur. This is important - modern cryptography in general wouldn’t work with out good random number generation. However, if the sequence described in jm’s post is correct (gun type, manufacturer, rarity, etc) you would actually have to weight the rolls in order to make all legendaries equally likely.
tldr; RNG is pseudo RNG and likely will be for the foreseeable future because our computers are designed to come up with the same result every time they are given a particular calculation. It doesn’t mean they have to be weighted towards particular outcomes, but if developers aren’t careful - or want rolls to be weighted - they can be.
Agree.
There are definate RNG patterns in this game that can be noticed, but nothing quite like that… Like farming for a Carbuncle or Masher in Traunt’s loot boxes… Knowing that within every hour a Victory Rush will drop from Sloth/Thunk. Knowing that after about 4 hours you may see an otto idol from Tyreen.
And yeah the Algorithms are more sophisticated. It was merely an example to counter the idea that devs have nothing to do with RNG - like too many people argue…
Agree.
Two places I have noticed that something is going wrong with BL3’s rng are in Maliwan Takedown, and for some reason the Agonizer 9000 fight. In those areas, on PC at least, while in cooperative, I often find identical drops to other players on LAN. Typically, one (or two) drops will be different, and the rest will be the same. I suspect what is happening here is that the seed is getting set for each client from the system time of the host, so only a few clock cycles have passed, resulting into two sequences of numbers that are mostly the same being generated.
I also noticed that happening with Agonizer. Split-screen farming generates exactly the same items on both screens in cooperation mode.
Point here is that drops are “random” in the sense of probabilistic (and subject to caveats discussed above Re: algorithmic pseudo-RNG) , but as our experiences demonstrate, if the coding of that RNG is screwed up by developers, players may get screwed by bad programming, be able to exploit an obvious fault in their favor, or notice genuinely non-random patterns in the behavior of the pseudo-RNG.
I would only add that for that last one, reasonably concluding non-randomness or deviation from “intended” behavior requires large data samples. Usually much larger than our intuition suggests (anyone who has read Kahneman or has taken an advanced statistics course knows what I mean). One afternoon or even full day of farming by one person usually isn’t nearly enough.
2 Likes
Technical nitpicking. The actual random number generator used in the Unreal Engine is a pretty decent one - that was demonstrated very thoroughly on the old forums with BL2 by someone who knew their stats and programming.
True. But the weights (in the loot tables) simply define the chance of one thing relative to another whenever the RNG function is polled. If you’ve read the “Inside the Box” loot articles, you’ll know exactly how that works.
And the PS4 clearing it’s cache is irrelevant - the game doesn’t use a random number table, it uses a recursive algorithm (that is properly seeded rather than being initialised to zero whenever the game fires up).
Apologies for nitpicking the nitpick, but regardless of whether there is a solid RNG algorithm in place, it is almost certainly pseudo random, namely because algorithms for deterministic machines (aka, any physical computer in the last century) that produce random numbers don’t actually exist. They may as well be random when they are seeded properly, but given the same input, they produce the same output and the same sequence of numbers every single time.
In Borderlands 3, there are a few areas where it appears someone may have seeded the algorithm again somewhere they shouldn’t have, or seeded multiple players from the hosts clock. (Maliwan Takedown and Agonizer 9000, at least…) The chances of identical drops are otherwise astronomically low. But any player actions intended to influence the drops? Far more likely to be based on confirmation bias than anything else.
From a bit of poking around, it looks like the Unreal engine just calls C rand (at least in the days of BL2). Nothing cutting edge, but if used properly, it should provide a more than sufficiently random distribution of numbers. Two exceptions exist (and the first is true of pretty much any algorithm) - if it is improperly seeded, you could end up with the situation I described above. The other possible issue is with something called race conditions, which basically has to do with multiple CPU threads attempting to access the same object in memory, without being protected by something called a mutex (mutual exclusion). However, I believe Unreal uses C++ rather than C, so this probably shouldn’t be an issue here, as C++ rand() implementations can guarantee no data races.
In any case, I agree with the gist of your post, but was compelled to point out that we don’t actually use true random number generation in modern computing.
ETA: Also not necessarily saying you were refuting the idea that RNG in BL3 is pseudo random, but just clarifying for anyone else following the conversation or who cares about these sorts of things…
1 Like
I mean, the bottom line is that when folks say “It’s RNG” it’s basically short form for “drops have an element of random behaviour”, meaning: don’t expect to consistently get $item once every $tries; sometimes it’ll be more, sometimes less,
Interesting point up-thread about why Woodblockers etc. might be so common - that makes sense as far as world drops are concerned. So we need at least one more legendary Hyperion sniper. Hmm… anyone have any candidates from previous games that aren’t in BL3 yet?
That’s certainly true - just because it is pseudo-random doesn’t mean the chance of getting an item after n tries with chance to drop P won’t follow basic rules of probability. So, we should have:
1 - (1 - P)^n
I think in general a lot of complaints about RNG in games hinge on what the drop rates are to begin with. So, say you want an item with a 6% drop rate that drops anointed 17% of the time, and you are only interested in 1/10 of the possible anointments. You’d have a 0.0102% chance of finding it, or a little over 1/10000. Intuitively, most people think that in a situation like this, after 10000 runs, they should have found their item. However, there would still be a ~36% of not having found it by then. Even after 20,000 runs, there would still be ~13% of not having found it. On the other hand, with thousands of people playing, some people will find what they are looking for on the first try. The gap between people’s intuitive understanding of probability and how it actually works can definitely lead to the feeling that things are rigged.
On the topic of Hyperion snipers, I believe the Invader was in all the previous games. It functions a bit like the Woodblocker, but I think the recoil was more manageable. The Longbow is similar to the Masterwork Crossbow, but that one can’t world drop. The Omnicannon is also a candidate, but its rather similar to the Headsplosion. Of the three, I’d vote for either the Invader or the Omnicannon, but they might need to be changed up a bit to differentiate from current offerings.
That’s crap. I’ve only seen maybe 6 or 7 of those weapons and I’ve beaten it over 100 times. I keep getting the Woodblocker, Gunerang and Brainstormer over and over again.
There are exceptions to this, and perhaps GBX programmers use one.
For example, when I write code that needs to be “really random” I utilize (and pay for, it’s cheap) Random.org’s API set to retrieve truly random numbers as I need them, I usually batch request 10,000 at a time and re-request when I’m down to 2,000.
Random.org is random.
“RANDOM.ORG offers true random numbers to anyone on the Internet. The randomness comes from atmospheric noise, which for many purposes is better than the pseudo-random number algorithms typically used in computer programs. People use RANDOM.ORG for holding drawings, lotteries and sweepstakes, to drive online games, for scientific applications and for art and music.”
And again, any programmer can use this, either the free sets or pay for batching.
–RoA
1 Like
There’s the Invader but that’s basically the same effect as the Woodblocker. Borderlands 2 had the Longbow but that would overlap with the Masterwork Crossbow (although the Longbow was much better than the Crossbow is now). Other than that there’s the Fremingtons Edge but I’m not sure if it would make for a good gun since it’s basically just the Monocle but Hyperion (high zoom to the point where it’s near unusable, increased crit when scoped in), the Morning Star (might be a candidate because it has an interesting effect where hitting crits grants you a stacking crit bonus but on the other hand it’s also one of the most annoying guns in Borderlands History as it keeps on complaining and berating you for being a horrible person) and the Omnicanon (which basically just adds some explosive damage).
Truth be told, if they are gonna add more Hyperion snipers to world drop, I think they will have to come up with new ones.
2 Likes
Good point - in an earlier comment I mentioned the possibility of using barometric pressure, but I didn’t know about Random.org. Thanks for sharing.
I don’t think it’s likely that this is used in Borderlands, however, for several reasons. One, from what I could dig up, it looks like the Unreal Engine’s rand just calls C++'s rand. (Of course, they wouldn’t have to use that, and could have included some other library or even rolled their own.) Two, it is possible to play the game without an internet connection. This indicates that most likely - barring downloading large batches of numbers and storing them on the hard drive somewhere - that they are using some internal mechanism. Finally, there are some areas in the game (MT, Agonizer 9000) where item drops are identical for players in LAN on cooperation mode, but in most other areas they are not. I suspect this is a result of both players getting seeded from the same system clock.
1 Like