ZOS, fix your RNG 90 keys gone, no Ilambris medium

MrBrownstone

But why are you trying so hard to get a medium one? You got a heavy already, go get a medium helm and you're set.

gatekeeper13

LordGavus wrote: »

gatekeeper13 wrote: »

e.g CoA. 14 drops, 8 Sunderflame, 6 Embershield, 0 BSW armor. How is this "random"?

But that is random. A low chance of happening, sure. But its random.

Random doesnt mean evenly spread out or fair. And the previous drop has no bearing on the next. Randomness is clumpy.

Nope, thats not random. Not giving a popular set is not random. My screenshot shows that either the RNG got stuck somehow and didnt drop BSW or its designed not to drop it. Nothing random with having a drop rate of 1/10 for Embershield or Sunder staves (that no one needs) and 1/100 for BSW (that everyone farms like crazy).

Algorithm gives less chances to popular items to drop and more to trash sets. And its happening in all aspects of the game.

Infixo

MrBrownstone wrote: »

But why are you trying so hard to get a medium one? You got a heavy already, go get a medium helm and you're set.

I already had heavy divines head, so naturally I wanted medium, any trait, would transmute it. The very first 4 coffers were those nerien’s epaulets I mentioned above. I was like ‘what the heck, 4 in a row ???’ bad luck, ok, but finally ilambris gonna pop, right? No... At some point I decided to check how much ‘bad luck’ I have this time, it was totally not rational, I am fully aware.

But to change the topic a bit, after this experience I am starting to think that maybe in the long run, buying those fully random 1-key coffer is better approach. Especially if one has multiple toons, and experimenting with builds, etc. in this scenario you will end up with 5x more shoulders, basically building your own stockpile ready to be used when needed. You just need lots of storage space to keep them What do you think?

Infixo

LordGavus wrote: »

gatekeeper13 wrote: »

e.g CoA. 14 drops, 8 Sunderflame, 6 Embershield, 0 BSW armor. How is this "random"?

But that is random. A low chance of happening, sure. But its random.
Random doesnt mean evenly spread out or fair. And the previous drop has no bearing on the next. Randomness is clumpy.

Random means that on a sample big enough you should get evenly spread results. So... let's see trials, SunSpire, very popular. 12 people, chests, etc. tons of loot, like 100+ per run. I'd say pretty much statistically relevant with each run, but ok, let's take multiple runs.
Every time I run SS (literally) approx. 40-50% of the loot is Nahvintaas. I decon'ed tons of it. The rest I'd say is evenly spread among other sets (FG, Lokke, Yolna). So... I suppose there is someone who goes to SS and half of his/her loot is False God. Or Lokke. Be honest - how many times did it happen to you?

TheGreatBlackBear

Consider yourself lucky. It's a garbage set. You might have been tempted to run it if you had gotten those shoulders

Olauron

Infixo wrote: »

But to change the topic a bit, after this experience I am starting to think that maybe in the long run, buying those fully random 1-key coffer is better approach. Especially if one has multiple toons, and experimenting with builds, etc. in this scenario you will end up with 5x more shoulders, basically building your own stockpile ready to be used when needed. You just need lots of storage space to keep them What do you think?

Fully random 1-key is better only if you need other sets too. If you already have other sets and need just the one then 5-key is mathematically better.

DeepSeaDiver

Don't mean to brag or rub salt into the wounds but some of you must have terrible rng and I'm usually not that lucky ...

Got 2 Ilambris Med shoulders with divines & impen traits, neither transmutated. Total 5 Ilambri's shoulders inc. the meds. Got 5 Nerien's shoulders to put it into context.

Got 3 of each weight of Grothdarr shoulders.
Got 4 Skoria Med shoulders as well.
I don't burn through keys to get this stuff.

Diff note I got something like 3 BSW Inf staffs in 4 days, gave another away to a nice pair of players who grouped with me.

I've pug farmed vet dungs as a tank for helms/weps so less GF q time (tanks don't even need food for a number of them). You can even DD solo easy dungs such as COA1 even in crafted gear to try for BSW/staff if you want. Optionally do on norm and just improve the gear.

But then I never got purple MS jewelry even though I farmed for days (think how many dolmens u can do in that time). Also never got Levi necklace that I wanted at the time (some kind person gave one in the end) and dislike how many times I tanked/DD norm & vet COH1 & COH2, solo, pairs & full group.

Playing for approx 1.5 yrs on EU (shouldn't make any diff). Go figure.

tmbrinks

Olauron wrote: »

tmbrinks wrote: »

18 tries. <snip> Sorry for your bad luck, but nowhere near a statistically relevant sample size.

The thing is player is not expected to make a statistically relevant number of draws in this case. That is not a case when one player will make thousands or tens of thousands draws. Using RNG outside of cases with statistically relevant number of draws for one player is bad game design.

Yes, some of it is to get people to re-play content in a game. The RNG factor is done by design. Many of us do some tasks to the point of making them statistically relevant so that we can determine if the RNG is weighted or not.

Yes, there are also things in this game that are weighted RNG. The Uncracked Geodes, Master Writ Voucher values, etc. There is nothing inherently wrong with either of those.

For every person who comes in an complains that it took them 40 runs to get a vMA inferno, there's another person who got it in the first 4 runs. But who is going to come to the forums to complain? The person it took 40. The one who only took 4 is unlikely to say anything about it. So, the perception is that the RNG is skewed, but everybody is simply looking at incomplete, biased data and drawing conclusions that can't be made.

nafensoriel

Eliahnus wrote: »

nafensoriel wrote: »

You want them to fix RANDOM by making it less random?

It's not random by any means, it's pseudo-random, meaning the implementation tries to simulate randomness, but obviously it has it's flaws, as in most other games.

It baffles me when people actually do think that luck/unluck in a software program is really random.

It baffles me when people think pseudo-random is more random than actually random.
Yes, we do actually random all the time. Random is not what people think it is.

There is no statistical difference between the algorithms used for random chance and flipping a coin mathematically.
It's a fallacy that computers cannot provide an effectively random number due to some weird occult like thought that they are forced into "patterns".

/edit
Oh and "weighted random" is STILL RANDOM.
If the roll is 1/100 it will still only ever land on ONE number out of ONE HUNDRED.

Neoealth

Sorry OP, I got a few of those shoulders yesterday and destroyed them. ( ͡° ͜ʖ ͡°)

Kartalin

Part of the issue is sample size as someone said above. In order for data to be at least approximately normal in distribution, we need:

n*p*(1-p) >= 10

In this case n, the sample size, is 18. The probability of success, p, could be calculated as 1/6 or about 0.167. (2 monster sets and 3 armor weights gives 6 total options of which 1 is desired) Thus,

n*p*(1-p) = 18*0.167*0.833 = 2.5

which is way less than 10. In fact, we'd need the sample size to be about 4 times as large (72 statistical, not ESO, trials) in order to assume some degree of normality, which is to say a reasonable expectation of a somewhat even distribution of results.

So that's unfortunate, but we can still look at the binomial distribution just to see what that tells us.

Ex: What is the probability of receiving 0 medium ilambris shoulders out of 18 coffers? Without going into too much of the formula or calculation detail, the result comes out to be 0.0373 or just under 4% chance of this occurring. So relatively unlikely but not outside the realm of possibility unfortunately.

So now, just for fun, let's try a hypothesis test. This is really spurious because we'd be making the claim that the proportion of drops that is theoretically 0.167 was observed to be 0 in a sample of 18. But what the heck, I'm just sitting in my office right now before class.

Null hypothesis: p = 0.167
Alternative hypothesis: p < 0.167

I'm going to use the P-value approach here, this gives us a test statistic of z0 = -1.89 Calculating the P-value, the probability that z < -1.89, we get P=0.0294. Basically what we can conclude here is that if we repeat this process of buying 18 coffers we anticipate getting less than 3 ilambris medium shoulders in about 3 of 100 repetitions. And unless we were using an extremely low level of significance (0.01 or 0.02 for example) then we can be reasonably confident that we should reject the null hypothesis and conclude that the data tells us the drop rate is in fact lower than 1 out of 6.

TL;DR

1. Results are not assumed to be normally distributed with less than 72 coffers.
2. The outcome of 0 medium Ilambris shoulders out of 18 is expected about 4% of the time.
3. Hypothesis testing tells us the drop rate is actually less than 1 out of 6 for medium Ilambris, but it's hard to get behind this idea at this sample size.

Edit: Lol, just realized I should use a T-distribution not Z for the hypothesis test. Ah well, not reliable results anyway
.

idk

Browiseth wrote: »

it aint broke if it's doing its job innit

(that job being psuedo random and unpredictable)

Pretty much this. Even OP acknowledges it is RNG. They are just complaining that RNG is not in their favor.

Remember when we not only needed to get the correct set, weight and trait? So nice that we can transmute the trait not as it makes RNG less of a pain.

Lucious90

Not a huge sample size, but I did 100 bags of Potentates looking for the shield.

Ended up with 90 rings, 8 necklaces and 2 Staffs.

Did Over 50 of the Warriors fury box. Ended up with the 5 pieces I needed because of transmute.

I disgree with the RNG gate. I like the idea of the keys being used as a token to get what you want. Especially if you have to grind dungeons, AND need them to progress harder content and Trials. While I do agree that the Current trial for progression needs to stay RNG. The basics shouldn't be.

PvP gear should never be hidden behind RNG. Keep the regional gear bags as RNG, but introduce more profession bags

ATomiX69

now imagine how it was before trait rerolling kek.

Alinhbo_Tyaka

I've often wondered if there is a bug somewhere with the seed number being used. Most pseudo random functions will return the same sequence of random numbers if provided the same seed number each time they are initialized. To get closer to a true random number sequence you need to also provide a different seed number to change the pseudo random number sequence. There a number of ways to generate a pseudo unique seed number but even then there is a chance for duplicates., Of course without seeing the actual code it is impossible to know what they are using to create the pseudo random number or how everything is initialized. So until then I will sacrifice a goat to RNGesus any time I really need the numbers to roll in my favor.

Waynerx8

Kartalin wrote: »

Part of the issue is sample size as someone said above. In order for data to be at least approximately normal in distribution, we need:

n*p*(1-p) >= 10

In this case n, the sample size, is 18. The probability of success, p, could be calculated as 1/6 or about 0.167. (2 monster sets and 3 armor weights gives 6 total options of which 1 is desired) Thus,

n*p*(1-p) = 18*0.167*0.833 = 2.5

which is way less than 10. In fact, we'd need the sample size to be about 4 times as large (72 statistical, not ESO, trials) in order to assume some degree of normality, which is to say a reasonable expectation of a somewhat even distribution of results.

So that's unfortunate, but we can still look at the binomial distribution just to see what that tells us.

Ex: What is the probability of receiving 0 medium ilambris shoulders out of 18 coffers? Without going into too much of the formula or calculation detail, the result comes out to be 0.0373 or just under 4% chance of this occurring. So relatively unlikely but not outside the realm of possibility unfortunately.

So now, just for fun, let's try a hypothesis test. This is really spurious because we'd be making the claim that the proportion of drops that is theoretically 0.167 was observed to be 0 in a sample of 18. But what the heck, I'm just sitting in my office right now before class.

Null hypothesis: p = 0.167
Alternative hypothesis: p < 0.167

I'm going to use the P-value approach here, this gives us a test statistic of z0 = -1.89 Calculating the P-value, the probability that z < -1.89, we get P=0.0294. Basically what we can conclude here is that if we repeat this process of buying 18 coffers we anticipate getting less than 3 ilambris medium shoulders in about 3 of 100 repetitions. And unless we were using an extremely low level of significance (0.01 or 0.02 for example) then we can be reasonably confident that we should reject the null hypothesis and conclude that the data tells us the drop rate is in fact lower than 1 out of 6.

TL;DR

1. Results are not assumed to be normally distributed with less than 72 coffers.
2. The outcome of 0 medium Ilambris shoulders out of 18 is expected about 4% of the time.
3. Hypothesis testing tells us the drop rate is actually less than 1 out of 6 for medium Ilambris, but it's hard to get behind this idea at this sample size.

Edit: Lol, just realized I should use a T-distribution not Z for the hypothesis test. Ah well, not reliable results anyway
.

You lost me at "part of the issue"

AlnilamE

I'm amused at the premise that ZOS knew the OP was looking for Medium Ilambris and skewed the RNG against them.

Kartalin

It might be easier to use a heavy shoulder and farm for a medium helmet. At least your odds in getting the helmet are only 2 to 1 instead of 5 to 1 from the coffers.

Plus you could potentially trade with someone else in group which increases your chances.
.