The RNG needs to be updated urgently!

[sporealanw]

Hey guys,

I am really getting tired of that RNG (for those who don't know: RNG = Random Number Generator) that is used in SWTOR. After it annoying me several times in the past months I am now really sick of it, sorry to say so.

 

There is a TD;DR version at the bottom.

 

I will give three examples, all of them are in 4.0 content, however I don't know any other examples atm.

 

First and most up-to-date example: I was trying to get this armor's head for my commando. You get it from the Alliance Supply Crates dropped in heroics.

I quickly completed all rep's heroics where you get Aygo's supply crates. I skipped a few disproportionate long missions and I got about 14 crates for Aygo. Of course I knew from my earlier experiences that I won't have any chance to get this helmet in only 14 crates, but this is really ridiculous: http://i.imgur.com/xNR4OSO.jpg

The red area is the total loot I received, the yellow is the one I am slightly disturbed of. In only 14 crates these 4 items (out of 70 possible armor items that can be dropped from Aygo's crates) came twice! And that is definitely not the first time I saw something like this!

 

Second example (Feb 2016): Star Fortresses. I was trying to get the achievement "Fallen Knights of the Empire" (Defeated each of the Star Fortress Paladins aboard all of the Star Fortresses). Of course I had to do several rounds to get it, but there's also a part where it get's ridiculous. The following happened: I finished all paladin achievements except the Voss Star Fortress. There are 4 paladins that can spawn there. For easier explanation I will give them numbers:

Bellic Hondi = 1

Zai Branen = 2

Orus Jind = 3

Avia MUr = 4

I already got 3 and 4. In each Star Fortress run exactly 2 paladins out of these 4 will appear. That means there are 6 possible combinations (1 2 - 1 3 - 1 4 - 2 3 - 2 4 - 3 4). Now here's a little bit maths: the probability that

1) the 2 paladins that I need will appear, is 1/6 - to be specific: 1 2

2) one of the paladins that I need will appear, is 4/6 - to be specific: 1 3 - 1 4 - 2 3 - 2 4

3) the 2 paladins that I do not need will appear, is 1/6 again - to be specific: 3 4

Now I start one Star Fortress instance, and of course the third case occured. That might still be normal, it had a chance of 1/6 against 5/6 that I get at least one of the missing paladins. But here it comes: I start another Star Fortress and again case 3 occurs! That is twice in a row 1/6: 1/6 * 1/6 = 1/36 which is about 2%!

In words, it was most unlikely that this could happen to me - the 2% against 98% that out of 2 Star Fortresses I get at least one paladin I needed.

 

Third example (spring 2016): another go with Alliance Supply Crates: Unfortunately I have no proof for this because I never expected something like this. I wanted to collect this armor (all items) for my Jedi Sage. So I needed 7 of 71 possible items (71 because there are two possible drops for Underworld Knight's chest). Of course I knew I had to collect quite a lot crates even with a well-working RNG. But I seriously collected much more than 170 supply crates within 3 weeks and I was still waiting for 2 more items to get the full armor set! You can imagine how angry I was after that.

 

 

TL;DR area

Example 1: In 14 Alliance Supply Crates, 4 armor items have been dropped twice, check the image links above.

Example 2: Star Fortresses: 2% chance occured against 98% chance while working on an achievement.

Example 3: Tried to get one full armor out of Supply Crates (link above), needed way more than 170 crates to get the pieces together.

[JWagner]

Summary:

I already got this item, so it shouldn't be in the RNG?

Definitely disagree with that idea. As awesome as that is, it would lose the RNG aspect of it to have any control of it.

[sporealanw]

I expected that answer. I don't mean to make the RNG user-friendly, I don't want it to be controllable, but more like a random number generator should be!

As you can see from the examples above (and that is the exact reason I mentioned them!) this is nothing about RNG but always the same boring stuff coming out of the crates, the RNG created the same numbers over and over. You wouldn't sit down to a poker table if you knew that the cards are the same each round, would'ya?

Edited October 13, 2016 by sporealanw

[Stoofa]

But you do understand how random works right?

If I say there is 1 of 10 possible items in a box, you could open any number of boxes and get exactly the same item each time - and it would still be random.

Your "tests" are nowhere near big enough to prove/disprove that random is or isn't random.

[dr_mike]

But you do understand how random works right?

 

http://dilbert.com/strip/2001-10-25

 

If you search the forums for hypercrates, you'll see that some players think that BW has modified the randomness in an attempt to get players to do more grinding. (Bad choice of words but I hope you get what I mean.)

 

I have a feeling this is just another example of that.

[sporealanw]

But you do understand how random works right?

If I say there is 1 of 10 possible items in a box, you could open any number of boxes and get exactly the same item each time - and it would still be random.

But that is the problem, there are way more than 10 items you can get. That means the chance you get the same item is a lot lower and the fact that you still get the same items every time makes you easily think that the RNG prefers several numbers and skips the others. As I said an RNG that will give you the same number over and over again is not random, it is predictable.

 

Your "tests" are nowhere near big enough to prove/disprove that random is or isn't random.

That were no tests, that was just examples that happened to me in daily playing. And I said examples, not a full scientifical documentation. This is just a part of what problems I already had with that RNG. What I didn't mention is that in example 3 I usually went to Sana Rae with ~20 boxes per load. And in every load there were several items just like in example 1 that I double, tripple or even quadruple looted, while other items have never shown up. An RNG is supposed to go through all items with the same chance of looting which means that if you only collected enough you would have all lootable items in the same amount. But from what I have seen in opening over 400 crates in total is that I would have massive piles of these items while only a few single ones oth those items.

[Cassianus]

http://dilbert.com/strip/2001-10-25

If you search the forums for hypercrates, you'll see that some players think that BW has modified the randomness in an attempt to get players to do more grinding. (Bad choice of words but I hope you get what I mean.)

 

I have a feeling this is just another example of that.

In other words, loot generator has the non-normal distribution (either because of poorly implemented RNG, or intentionally). I also strongly suspect this, because of very few pieces which I should get to complete the full sets (I gather all the sets) and still no luck; at the same time I quite often get 2 same pieces from 6-8 crates opened in line.

Yes, it can be counted as an anecdotal evidence because I don't carry out these experiments with the same initial conditions and don't write the results. Sure, it's possible to do such experiments, even though it will be very time-consuming and tedious. And if we will open 100 crates, someone can always argue that it's statistically incorrect, need to open 10 times more. And if we open 1000 crates, one person may say that it's not yet enough...

 

Well, let's suppose that we got enough data showing that RNG distribution is non-normal. What next?

 

Me: BW, it's unfair! Please fix the wrong RNG.

BW: We never promised to give you the strictly equal loot generator. It is random, it's enough.

Me: But with current distribution I will be gathering all the pieces for centuries!

BW: Players like the unpredictability and fun. We think that randomness is OK. If you dislike it, you are in minority, sorry for that.

[SteveTheCynic]

But that is the problem, there are way more than 10 items you can get. That means the chance you get the same item is a lot lower and the fact that you still get the same items every time makes you easily think that the RNG prefers several numbers and skips the others. As I said an RNG that will give you the same number over and over again is not random, it is predictable.

Have you heard of the "birthday paradox"?

 

Basically, if you have an item chosen randomly (e.g. birthdays, although they aren't random, nor are they uniformly distributed during the year, but close enough), and you pick repeatedly (and it's something where results are "independent" like rolling dice rather than "dependent" like picking cards from a deck without shuffling the ones already picked back in, so you can't the Queen of Diamonds again), then you have a strong chance of having a duplicate much sooner than you'd think. (For birthdays, the chance of at least one duplicate passes 50% after just 23 selections.)

 

So for 70 different armour pieces available, and given that it is of uniform distribution and independent selection (neither of which is guaranteed, but doing it otherwise doesn't make sense), you can normally expect a 50% chance of duplicates after around 10 picks. (The crude formula is 1.2 times the square root of the number of different possibilities.)

 

Also, on the distribution of values picked by a random function: if it cannot produce a run of arbitrary length of the same value, it isn't random. If it cannot skip a particular value during hundreds and thousands and millions of tries, it isn't random.

 

No, as far as I can tell, the RNG is just fine. Leave it alone.

[Adric_the_Red]

I've seen these kinds of threads in the forums for pretty much every MMO I've ever played, going back more than 15 years. In pretty much every case, it's people who don't understand that getting duplicate results in just a few tries, or conversely, not "hitting" on a 20% chance after five tries, doesn't mean there's a problem with the RNG.

 

As an example, I've had folks tell me that with the new 60% chance to RE since patch 4.0, you can "guarantee" a success by making 2 of whatever it is you want to RE, and you will "always" succeed on the second try. Except I know from my personal experience that this is not true. I've had success with only a single try, and I've had failures with 2 tries, or 3, or even more. Because it's, well, random. And your results on one try don't influence your results on the next.

 

I suspect they think that a 50% chance of getting x result means that you will succeed on every other try, so 60% chance means you are absolutely certain to succeed within two tries. Like thinking if you get heads on your first coin toss, the second will be tails, because the chance of getting heads is 50% and you already got it once. And they don't understand that the 50% chance applies to every toss, and the results of the second toss aren't influenced by the first. Getting heads on the first toss doesn't mean you have a 100% chance to get tails on the second.

 

That Dilbert cartoon is right on point. Ironically, it is entirely possible that the demon spitting out all nines is in fact producing random results, in the same way that it is possible - not extremely likely, but possible - to do a truly random coin toss and get heads 10 times, or even 100 times, in a row.

 

And on the rare occasions when the game developers address the player concerns, they typically say they have run tests on the RNG that cover hundreds of thousands of "rolls" (or more), and their analysis shows the results are indeed random.

 

And typically, neither explanation satisfies the folks complaining. They hold fast to that, "But my results don't look random to me, so there must be a flaw in the RNG".

Edited October 14, 2016 by Adric_the_Red

[SerraShar]

No, as far as I can tell, the RNG is just fine. Leave it alone.

 

Not it isn't..Thanx to IT...I have hundred of gloves and Boots and belt and no pants or armor lol !

 

*Bombardement*

well..gotta use them for something

[Rolodome]

As far as I know, the only discernible way to better your odds is to raise your immediate sample size. Meaning, say you have 10 boxes to open. Instead of opening 2, waiting a couple of hours, and then opening 2 again, you open all 10 at once, in succession.

 

However, I'm not sure that even does anything, as there is a time gap between the generation of each number. It's not like running a computer program on a million attempts where you can generate every result in miliseconds.

 

In the main, there are two things at work that make RNG odd in games:

 

1) The sample size is basically never big enough. If you consider the sample size as the compiled roll of every player in the game, then the sample size becomes big enough enough and you start to see what might look like reasonable distribution. But personal sample size? Not even remotely close. Usually the highest players get with items is in the hundreds and that's when they're trying to open up a bunch at once for science or show.

 

2) Computers don't use true randomness. They use a kind of randomness that is close and tries to simulate real randomness, but it's not truly random. From what I understand, the closest thing we have to true randomness in a simulation is random.org, which uses atmospheric noise.

 

That said, people tend to latch onto the knowledge that computers are not truly random and think that means it's affecting them moment to moment. But usually, it's not going to be because the sample size is so poor anyway, that the randomness being slightly off is not going to make a noticeable difference.