nMA random drops do not appear to be random, here is probably why

furiouslog

So, after multiple runs on different mag and stam toons (mostly mag) in nMA post-Greymoor, here's a tally of my drops:

Cruel Flurry: 11 sets (2A/3D/5M/12S)
Thunderous Volley: 3 sets
Merciless Charge: 5 sets (1BA/2GS/2M)
Crushing Wall: 2 sets (2L)
Precise Regeneration: 1 set
Rampaging Slash: 0 sets

If the RNG is set based (which would make logical sense), you should have an equal chance of getting any one of the sets. That would appear to make the most sense from a programming and player rewards standpoint, since individual weapons don't drop. In this design, there would be a nested probability that you will get a particular weapon within a set given that you received a particular set. After 22 runs, there should be a 98 percent probability that I'd get at least one of every set. The most likely outcome after 22 runs is that I'd have at least 3 of each set. NOTE: NEW DATA ARE AVAILABLE IN A LATER POST SHOWING 47 RUNS - SCROLL DOWN TO SEE.

Looking at drop probabilities, that gives us a probability table that looks like this:



with detailed drops like so:



If, however, the RNG is based on the weapon drop and if the shield drop is tied to the Rampaging Slash set while Cruel Flurry is tied to one-handers, the probability table looks like this:



with detailed drops like so:



which more closely matches the distribution of drops I'm currently experiencing. The probability that I didn't receive a single drop of Rampaging Slash in the first case is less than 2 percent, and in the second case, it's 17%.

As such, I'd bet that the RNG is tied to weapon drops (the second case), and if that is true, and further assuming that it is not biased in any way, two things to note for you mag toons: you are much more likely to get Cruel Flurry when you run the arena than anything else, and the number of times you'd need to run it to have a 99% chance of getting a single Inferno Staff is 58. I'm trying to get one for each of my 4 mag dps toons who do not currently have it, which means to come close to guaranteeing that each of of them get an inferno staff, I'll need to run nMA 232 times, at about 45 minutes per run (including loading times), or roughly 174 hours of gametime. If I have 3-4 hours to play each day, including writs, trials, etc, I'd have to forgo all other activities and just run nMA every day on multiple alts for my entire gaming time for nearly 50 days straight to achieve that goal. Or, I can do it once per day and just wait until next year to finish the outfitting process for my mag dps toons, I suppose.

Bummer.

I guess I'll continue keeping this record until I get so sick of nMA that I give up - if anyone has data they've tallied that they'd like to add, feel free to chime in.

Luke_Flamesword

22 runs is not enough to make a statistic. You can after 222 runs and then maybe it will be show some trends.

Today I went to nMA first time after Greymoor patch. I wanted Bow and I get it in first try That's it for me for legendary Maelstorm Bow. It was just too easy

Seriously, it was too easy. It shouldn't be that way.

furiouslog

Luke_Flamesword wrote: »

22 runs is not enough to make a statistic. You can after 222 runs and then maybe it will be show some trends.

Today I went to nMA first time after Greymoor patch. I wanted Bow and I get it in first try That's it for me for legendary Maelstorm Bow. It was just too easy

Seriously, it was too easy. It shouldn't be that way.

Statistics can be calculated from any representative sample. It's definitely true that a larger sample gives a more accurate picture of trends, but in this case a Kolmogorov-Smirnov test shows that the weapon-based distribution more closely represents my actual results with a confidence of 93%, so I think I'm probably right, absent any additional information.

Chilly-McFreeze

Luke_Flamesword wrote: »

Seriously, it was too easy. It shouldn't be that way.

It was on normal, how hard should that be?

Or what do you suggest? [snip]

[Edited for Bashing]

SidraWillowsky

furiouslog wrote: »

Luke_Flamesword wrote: »

22 runs is not enough to make a statistic. You can after 222 runs and then maybe it will be show some trends.

Today I went to nMA first time after Greymoor patch. I wanted Bow and I get it in first try That's it for me for legendary Maelstorm Bow. It was just too easy

Seriously, it was too easy. It shouldn't be that way.

Statistics can be calculated from any representative sample. It's definitely true that a larger sample gives a more accurate picture of trends, but in this case a Kolmogorov-Smirnov test shows that the weapon-based distribution more closely represents my actual results with a confidence of 93%, so I think I'm probably right, absent any additional information.

You could also run a Chi-square using your observed counts vs. expected counts. I backed into your counts using your N and probabilities and tried it-



Take the observed count vs. expected count, square the difference, divide by the expected value, then sum all to get your Chi-square stat. Degrees of freedom is (# rows-1)x(# columns - 1), so 5 here. Comparing to a Chi-square distribution, your p-value is less than .001. So the counts that you got are very different from those that would be expected if the drop rates were the same.

furiouslog

SidraWillowsky wrote: »

furiouslog wrote: »

Luke_Flamesword wrote: »

22 runs is not enough to make a statistic. You can after 222 runs and then maybe it will be show some trends.

Today I went to nMA first time after Greymoor patch. I wanted Bow and I get it in first try That's it for me for legendary Maelstorm Bow. It was just too easy

Seriously, it was too easy. It shouldn't be that way.

Statistics can be calculated from any representative sample. It's definitely true that a larger sample gives a more accurate picture of trends, but in this case a Kolmogorov-Smirnov test shows that the weapon-based distribution more closely represents my actual results with a confidence of 93%, so I think I'm probably right, absent any additional information.

You could also run a Chi-square using your observed counts vs. expected counts. I backed into your counts using your N and probabilities and tried it-



Take the observed count vs. expected count, square the difference, divide by the expected value, then sum all to get your Chi-square stat. Degrees of freedom is (# rows-1)x(# columns - 1), so 5 here. Comparing to a Chi-square distribution, your p-value is less than .001. So the counts that you got are very different from those that would be expected if the drop rates were the same.

That's a better way to do it, but I use Stata and KS is a lazy way to compare distributions. Thanks for the add.

Luke_Flamesword

It was on normal, how hard should that be?

Or what do you suggest?

I suggest that before patch MA and DSA were to hard to obtain and now, after patch - arena stuff is to easy to obtain. There is to much difference between normal and veteran difficulty - especially on boring, super slow and too easy DSA. I think that there should be some middle ground.

gatekeeper13

Nothing random with ESO RNG system.

Proof?

Burning Spellweave Inferno Staff. Probably the lowest drop rate I ever seen in an item. And what a coincidence? It's one of the most sought after staves.

Bradyfjord

From what I remember about statistics, you need a sample size in the thirties to have the minimum amount of information to start making inferences from.

Tannus15

gatekeeper13 wrote: »

Nothing random with ESO RNG system.

Proof?

Burning Spellweave Inferno Staff. Probably the lowest drop rate I ever seen in an item. And what a coincidence? It's one of the most sought after staves.

I've never farmed this item and I have 3.

A guild mate has organised 22 runs of normal cloudrest and not a single siroria inferno staff has dropped. Your previous drop has no bearing on the likely hood of your next drop.
@furiouslog You've rolled a d20 22 times and then you're shocked that you didn't roll a 20 and that you got 10 three times.
It happens.
Move on.

furiouslog

Bradyfjord wrote: »

From what I remember about statistics, you need a sample size in the thirties to have the minimum amount of information to start making inferences from.

That's not entirely true, that sounds like a rule of thumb, and than rule highly depends on what you are trying to test. However, I got additional data from my guildmate, and here's an updated table, now with Chi-square tests (low numbers means that the outcomes don't fit the expected outcomes). There are 47 observations now:

https://ibb.co/QFr97qb

It's pretty clear that the drop rate is most likely being anchored to a single weapon, which is attached to a set.

Tannus15 wrote: »

@furiouslog You've rolled a d20 22 times and then you're shocked that you didn't roll a 20 and that you got 10 three times.
It happens.
Move on.

This unnecessarily dismissive comment runs counter to the point of the post, which is to determine why I'm not getting expected results. I do statistics for a living. We're not talking about a 20 sided die, we're talking about a six sided one, for which a sample of 22 rolls is pretty good to tell you if it might not be a fair die if you are getting extremely weird results. 47 rolls is even better. To a scientific certainty? No, you'd have to run it thousands of times, but ignore the data if you like - one of these RNG methods is clearly more likely than the others.

I am fully aware that one drop does not influence another drop, but the combination of possible drops when you roll the die over and over again is still a measurable outcome that has an expected distribution. That is what I am analyzing, and my confidence that I have the set odds correctly calculated for nMA is pretty high.

As for your Siroria staff example, think of the number of drops you can get on Z'Maja. There are 4 sets, with 13 weapons (including shield) plus 2 jewels. That's 60 possible combinations you have to draw from (assuming each is equivalent), which is a pretty low probability. Even if you ran it 60 times, there is a 36% chance that you personally never would have gotten the Siroria staff any of those times. That's why people have a tough time getting individual weapons in certain sets, like BSW. Comparing that to this is comparing apples and oranges.

virtus753

The drop rate is tied to a single weapon, as suggested above, per the last official statement (Feb. 2017):

https://forums.elderscrollsonline.com/en/discussion/318838/pc-mac-patch-notes-v2-7-5#latest

Sergykid

wtb currency system on rng arenas, or at least tie them to coffers like undaunted.
anyway, arenas should be done for fun not for loot. They should work at the fun part, like less randomness in stages

red_emu

Any mathematician will tell you: You can prove any point with statistics.

furiouslog

virtus753 wrote: »

The drop rate is tied to a single weapon, as suggested above, per the last official statement (Feb. 2017):

https://forums.elderscrollsonline.com/en/discussion/318838/pc-mac-patch-notes-v2-7-5#latest

I saw that too, but it was unclear what the second part of the statement meant: "Now, it is based on the individual item type (Dagger, One Handed Sword, etc), and all damage weapons within the item type have an equal probability to drop." Does "damage weapons" refer to traits? Sets? No idea.

Honestly, all I'm trying to do is figure out how exactly how long I'll have to grind this boring content until I get what I want so I can determine if it's worth it or not. Right now, it's looking like it's not worth it.

I didn't mind running nCR hundreds of times because I can chat with my pals while I am doing it, and it's interactive, which is actually fun for me. It's a social activity. nMA is a non-immersive and un-challenging solo grind fest that offers no value apart from gaining (what appears to be) a 1 in 13 shot at a desired piece of gear. It's like playing a single level of Doom over and over and over again. It's literally a waste of my time. At this rate, I'll also be deconning about 140 Cruel Flurry weapons along the way, so I guess I need to take the benefit of the extra blue mats into account.

mairwen85

furiouslog wrote: »

virtus753 wrote: »

The drop rate is tied to a single weapon, as suggested above, per the last official statement (Feb. 2017):

https://forums.elderscrollsonline.com/en/discussion/318838/pc-mac-patch-notes-v2-7-5#latest

I saw that too, but it was unclear what the second part of the statement meant: "Now, it is based on the individual item type (Dagger, One Handed Sword, etc), and all damage weapons within the item type have an equal probability to drop." Does "damage weapons" refer to traits? Sets? No idea.

Honestly, all I'm trying to do is figure out how exactly how long I'll have to grind this boring content until I get what I want so I can determine if it's worth it or not. Right now, it's looking like it's not worth it.

I didn't mind running nCR hundreds of times because I can chat with my pals while I am doing it, and it's interactive, which is actually fun for me. It's a social activity. nMA is a non-immersive and un-challenging solo grind fest that offers no value apart from gaining (what appears to be) a 1 in 13 shot at a desired piece of gear. It's like playing a single level of Doom over and over and over again. It's literally a waste of my time. At this rate, I'll also be deconning about 140 Cruel Flurry weapons along the way, so I guess I need to take the benefit of the extra blue mats into account.

But at least you can do the arena with your eyes closed now. The biggest danger in nMA is falling asleep. So, yes, it sucks that along with the old vet weapon nMA got the vet RNG too, but still winning because it's so freaking easy. No major effort required is a good trade off. Its leisurely, more time waiting for spawns than anything else, you can get a cup of tea between rounds, have a smoke between stages. That's your compensation for the quirky, obviously biased rng.

The whole pre-greymoor attitude toward players wrt perfected weapons and that they should suck it up and go refarm--and here we are talking about nMA and how getting the weapon of choice could be a protracted period of time covering anything from 20 to 200 runs, or more. Its almost as if the complaints about undone effort and wasted time investment were warranted.

RNG has always been, is now, and likely ever will be a major complaint for maelstrom arena. It does feel skewed and weighed unevenly, and potentially for all the reasons you've mentioned and then some. Unfortunately, it's just a case of the odds being stacked heavily against you.

Drdeath20

Yeah seems tedious. Im bout 20 runs on vma at ~1 hour a pop and similar returns

gatekeeper13

Tannus15 wrote: »

I've never farmed this item and I have 3.

Just because you were lucky enough to get 3 staves, doesn't mean it drops for everyone else (I am one of the lucky ones too, has dropped 4 times for me but kept only 2, gave the rest to guild mates). The absolute majority of people I know had insane difficulty getting an inferno BSW. And there were people in this forum who said they ran CoA1 around 200 times without inferno BSW dropping even once. The former leader of a guild I am a member of, never got one. And he has been playing ESO 24/7 for years.

Gundug

22 runs isn’t much of a sample size. This reminds me of my own personal experiences collecting Psijic Ambrosia and New Moon Acolyte pages. And in my case, this is resulting from years in the case of Psijic Ambrosia, and months of New Moon collecting. Just because I have ten of page 5 and none of page 3 doesn’t mean that’s in any way representative of the theoretical chance of each page to drop. I have spoken to other players that had similar results, yet with completely different combinations of pages.

When I used to be into programming as a kid, I sometimes messed with random number generators. It would always look weird to have a setup designed to give a random number from 1 to 10, and get five 3s in a row. In our minds, we want to see a different number every time, but the algorithm doesn’t work that way.

TheGreatBlackBear

Bruh. They made maelstrom weapons to appease [snip] you and you're still complaining? What would you like instead. A free weapon for logging in?

[edited fro baiting]

furiouslog

TheGreatBlackBear wrote: »

Bruh. They made maelstrom weapons to appease [snip] you and you're still complaining? What would you like instead. A free weapon for logging in?

[edited fro baiting]

I'm not a casual. I have been playing since beta (with a break), and I have 12 alts, cleared all vet trials, all vet dungeons (most of them on HM), and gotten about 29K of achievement points on my main. Not really casual. My complaint is that the number of times I need to run nMA to achieve my goal requires an unbearably boring time sink that is way too long given other stuff I could be doing with my limited game time. Is that not a legitimate complaint?

@Gundug - read down a bit, I updated the sample with new data, and it is now 47. But either way, statistical tests take your sample into account when you are trying to prove or disprove a hypothesis, and 22 rolls of a 6 sided die where half of them come up on "1" is enough to conclude that the die is not balanced. You can google that, if you like, there is a lot of content on that particular problem. Either way, my hypothesis is confirmed by the wording in the patch notes. That second table shows the likely drop rates for different sets and weapon combos, but it might even be worse for bows, healing staves, and the sword and board.

Danksta

22 runs means nothing when you think of the thousands of runs that have already taken place since the expansion dropped.

furiouslog

Danksta wrote: »

22 runs means nothing when you think of the thousands of runs that have already taken place since the expansion dropped.

I'm not sure what point you are trying to make here. Can you clarify your statement?

Danksta

furiouslog wrote: »

Danksta wrote: »

22 runs means nothing when you think of the thousands of runs that have already taken place since the expansion dropped.

I'm not sure what point you are trying to make here. Can you clarify your statement?

I'm saying your samples size is much too small to come to any reliable conclusion.

kathandira

furiouslog wrote: »

snip

I have amazing RNG.

Literally got all weapons in a matter of 2 days. Only 2 duplicates.

Stealthk80

So I too have kept a log of my runs post patch 26 on xbox in vMA

I got a slashing rampage set last night on my 28th run. That was the first I've had, still no inferno.
I run a mag sorc in there - takes roughly 45 to 55 mins, normally about 47 mins if flawless.

here's my drops from 28 runs

8 resto staff
3 ice staff
3 lightning staff
4 t. Volley
4 cruel flurry (axe/dagger, sword/axe, dagger/dagger)
3 2h axe
1 maul
1 2h sword
1 rampaging slashes
0 inferno

mairwen85

kathandira wrote: »

furiouslog wrote: »

snip

I have amazing RNG.

Literally got all weapons in a matter of 2 days. Only 2 duplicates.

Which god do you sacrifice your goats to?

kathandira

mairwen85 wrote: »

kathandira wrote: »

furiouslog wrote: »

snip

I have amazing RNG.

Literally got all weapons in a matter of 2 days. Only 2 duplicates.

Which god do you sacrifice your goats to?

Pelor

mairwen85

furiouslog wrote: »

TheGreatBlackBear wrote: »

Bruh. They made maelstrom weapons to appease casuals like you and you're still complaining? What would you like instead. A free weapon for logging in?

I'm not a casual. I have been playing since beta (with a break), and I have 12 alts, cleared all vet trials, all vet dungeons (most of them on HM), and gotten about 29K of achievement points on my main. Not really casual. My complaint is that the number of times I need to run nMA to achieve my goal requires an unbearably boring time sink that is way too long given other stuff I could be doing with my limited game time. Is that not a legitimate complaint?

Aparently it wasn't a legitimate complaint for vma as per my previous post.

furiouslog

Stealthk80 wrote: »

So I too have kept a log of my runs post patch 26 on xbox in vMA

I got a slashing rampage set last night on my 28th run. That was the first I've had, still no inferno.
I run a mag sorc in there - takes roughly 45 to 55 mins, normally about 47 mins if flawless.

here's my drops from 28 runs

7 resto staff
4 ice staff
3 lightning staff
4 t. Volley
3 cruel flurry (axe/dagger, sword/axe, dagger/dagger)
3 2h axe
1 maul
1 2h sword
1 rampaging slashes
0 inferno

We'd need only your post-Greymoor runs to get apples to apples, because they changed the drop rates. You ran vMA 28 times since the update?

furiouslog

Danksta wrote: »

furiouslog wrote: »

Danksta wrote: »

22 runs means nothing when you think of the thousands of runs that have already taken place since the expansion dropped.

I'm not sure what point you are trying to make here. Can you clarify your statement?

I'm saying your samples size is much too small to come to any reliable conclusion.

The updated table has 47, not 22. But as previously explained, sample size is a function of the test specifications, and the statistics let you know if the test passes or fails. I'm not trying to calculate the odds to a scientific certainty, I'm trying to compare two templates for RNG against actual data to see which of the templates is more likely, and for that, I have a sufficient sample.