64 Key Opening (AKA Not as Many As Deltia OH WELL)

chimneyswift_ESO · October 2016

Enjoy

https://www.youtube.com/watch?v=kbO5eOyA_Cg

KingMagaw · October 2016

Nice video - ~5 Divines out of all that.

With more videos like this coming to light, it truly shows the drop rates are fixed, to me anyway.

@ZOS_RichLambert - Corruption is getting more transparent by the day.

Foxic · October 2016

It took me over 100 keys to get my first divines ilambris

Khairiah · October 2016

Wouldnt that be calculated to about the same % of chance as Rich Lambert showed in a different thread? Im not good at math so im not gonna throw myself into putting up some big example to prove something.

Ahzek · October 2016

KingMagaw wrote: »

Nice video - ~5 Divines out of all that.

With more videos like this coming to light, it truly shows the drop rates are fixed, to me anyway.

@ZOS_RichLambert - Corruption is getting more transparent by the day.

Isnt 5 out of 64 12.8% ? Isnt the drop rate supposed to be 12.5% ? Arent you just wildy throwing accusations in some kind of tantrum that you didnt get exactly what you wanted ?

Edit: I indeed did my math in entirely the wrong way. Its interesting how fast you forget things, especially when the result is obviously false like in this case. I apologize formally.

chaserstorm16909 · October 2016

Ahzek wrote: »

KingMagaw wrote: »

Nice video - ~5 Divines out of all that.

With more videos like this coming to light, it truly shows the drop rates are fixed, to me anyway.

@ZOS_RichLambert - Corruption is getting more transparent by the day.

Isnt 5 out of 64 12.8% ? Isnt the drop rate supposed to be 12.5% ? Arent you just wildy throwing accusations in some kind of tantrum that you didnt get exactly what you wanted ?

5 out of 64 is 7.8%.

Khairiah · October 2016

5/64x100 = 7.8%
Still not a bad rate.

DRXHarbinger · October 2016

Open the chest with the old sets only in. I bet it's full of divines mephala and malubeth.

Bad_Company · October 2016

So, this is basically what happened.
You got:

8x Well-Fitted
15x Prosperous
14x Training
7x Reinforced
5x Divines
5x Impenetrable
5x Infused
5x Sturdy

clocksstoppe · October 2016

It's fairly obvious that the dev claiming they are equally likely was flat out lying.

DRXHarbinger · October 2016

clocksstoppe wrote: »

It's fairly obvious that the dev claiming they are equally likely was flat out lying.

Yep. For sure. I'd love to see them post the same for vma weapons saying destros are just as likely...you know call the entire player base wrong.

GeneralPardon · October 2016

The issue is that you only look at you own results instead of the bigger picture. I have been lucky enough to get various sets in divine even more than prosperous or training (thank the divine).
You just compensate for my stuff, but you don't get the stuff YOU want so it's broken. Just deal with it!

code65536 · October 2016

clocksstoppe wrote: »

It's fairly obvious that the dev claiming they are equally likely was flat out lying.

A sample size of 64 is just far too small when measuring the distribution of something split 8-ways. A perfectly even distribution would've been 8--that's way too small and too easily subject to swings of luck. 5 vs. 8 extrapolates too much compared to something like 55 vs. 58.

Every PTS, I've spent a large (but still relatively small) amount of keys:

For example, in the Update 10 data, my silver keys yielded 38 shoulders. Perfect distribution would've been 5.4. The lowest from that batch was 2, the highest was 8. That's within the fairly narrow bounds of 5±3, but one is 4x higher than the other.

With the larger sample set of the Update 10 gold keys, perfect would've been 14.3, and the counts were 11 on the low end and 19 on the high end. It's actually a wider numeric spread, but the percentage spread is much less.

In the Update 11 data set, we're looking at 291 shoulders total. Here, the most interesting spread is that of Malubeth coming in with just 15 drops and Kena coming in with 31. Do you really think ZOS would go through the time and trouble to weight the tables (and if you knew how itemization works in ESO, you'll come to realize just how much extra work it would take for them to do that) to favor Kena 2-to-1 over Malubeth?

For the Update 12 data set, even with a large data set of 500 shoulders, there was still a pretty big numeric spread. Prosperous came in last with 56 drops, and Sturdy had the most, with 82. However, Update 12 was also the first update where I automated the logging of my drop-testing, and as a result, I still have the original raw log file of my testing on the Update 12 PTS.

Going through my log and looking at only the first 100 logged items (86 shoulders, 14 motifs), I see:

Divines: 11 (vs. 78 in the final count)
Infused: 17 (vs. 63 in the final count)
Prosperous: 10 (vs. 56 in the final count)
Reinforced: 13 (vs. 77 in the final conunt)
Sturdy: 10 (vs. 82 in the final count)
Training: 16 (vs. 70 in the final count)
Well-Fitted: 9 (vs. 74 in the final count)

So, if we look at the distribution of the first 86 shoulders vs. the final distribution of 500 shoulders, we see that the front-runner after 86--Infused with 17--ends up 2nd-to-last with 63. Prosperous and Sturdy, which were tied for 2nd-to-last after 86 shoulders come out in first place (Sturdy) and last place (Prosperous) after 500. Folks, that's what randomness is.

If you get two groups of people together, and tell one group to flip a coin 100 times and jot it down, and tell the other group to pretend to flip a coin and jot down what they think a random distribution would be, you could tell which group did the actual coin flips because it's probably the one with long streaks of tails or heads. And the one without streaks and with a nice-and-even distribution is likely the one that was fabricated. People naturally want to avoid patterns when they are told to make up "random" data. That's not randomness.

The moral of the story, folks, is that randomness is not the same as a perfectly-even distribution. There will be streaks and droughts. There will be shapes in the clouds in the skies. And that's true randomness.

Addendum 1: And on the forums, the people most likely to post about their experiences are people who have been burned by the randomness. We hear a lot of "I spent 50 keys and didn't get what I want!", the people who get what they want after the first dozen keys usually don't bother to come here to vent.

Addendum 2: To clarify, RNG can be harsh--if you look at my log data from the Update 12 key openings, even after spending 200 keys on Maj's chest on the PTS, the best Kragh pieces that I got was heavy divines and medium infused--I didn't get a single BiS medium divines item. I'm not saying that RNG can't screw people over--it can and it often does, and that is a legitimate concern that ZOS should address (tokens!). What I am saying is that there is absolutely zero credible evidence of ZOS tipping the scales in favor of any particular set or trait.

raidentenshu_ESO · October 2016

clocksstoppe wrote: »

It's fairly obvious that the dev claiming they are equally likely was flat out lying.

chimneyswift_ESO · October 2016

Bad_Company wrote: »

So, this is basically what happened.
You got:

8x Well-Fitted
15x Prosperous
14x Training
7x Reinforced
5x Divines
5x Impenetrable
5x Infused
5x Sturdy

These graphs are beautiful and should be posted on every video like this. Training and Prosperous need to be EXPOSED!

Ankael07 · October 2016

@code65536 PTS's RNG codes can be different right? I mean they could be coding the drop rates equally since everyone there can open tens of boxes often. Now I dont want to sound all gloomy but Im also skeptical about the graphics that Rich posted on the other forums. I mean its not like he released the codes for the drop rates so we can observe them with a third party program.

Seri · October 2016

code65536 wrote: »

Going through my log and looking at only the first 100 logged items (86 shoulders, 14 motifs), I see:
Divines: 11 (vs. 78 in the final count)

Infused: 17 (vs. 63 in the final count)

Prosperous: 10 (vs. 56 in the final count)

Reinforced: 13 (vs. 77 in the final conunt)

Sturdy: 10 (vs. 82 in the final count)

Training: 16 (vs. 70 in the final count)

Well-Fitted: 9 (vs. 74 in the final count)

Or in graph form since people like pictures.

That said, I still wouldn't complain if prosperous/training was removed from the drop table :P

code65536 · October 2016

Ankael07 wrote: »

@code65536 PTS's RNG codes can be different right? I mean they could be coding the drop rates equally since everyone there can open tens of boxes often. Now I dont want to sound all gloomy but Im also skeptical about the graphics that Rich posted on the other forums. I mean its not like he released the codes for the drop rates so we can observe them with a third party program.

I don't doubt what he says for the following reasons:

Although I don't know for sure how the game's loot tables work, I do know enough about how the game handles itemization and about programming in general to be fairly certain in saying that weighting the drops in favor of certain traits would be complicated and involve extra work. It's not like they can go in and tweak a single line of code to change a trait's share from 10% to 15%.*
Lost in all the noise about drop rates, there have been people who have reported in the forums (example) and in-game that their drop rates are relatively even. Again, keep in mind that people are more likely to speak up if they got lousy rolls. I got my Sharpened vMA Inferno staff after about a couple dozen runs. But until just now, I have never said this on the forums because I have nothing to complain about.
None of the cases of bad drops, once the numbers are shown, give me reason to think that they are anything other than luck. Deltia's 250-key video, if you actually count all the drops in the video, is relatively even. Yes, he did get a lot more training than divines, but that's consistent with the kinds of spreads that I've seen on the PTS. Again, my Kena and Malubeth drops were in a 2-to-1 ratio in the Update 11 PTS, and that was with a 291-shoulder sample set. It's just how randomness works.
Rich has been straight with us before on things like internal data, and I see no reason to start doubting him now.

I am very adverse to conspiracy theories in general. Such things require a burden of proof and evidence, and there simply is none. There are data points suggesting an even drop, and the data points suggesting uneven drops are not convincing enough to make me think that they are outside of the norms of randomness.

In any case, if I had the inventory space, I'd go on a key spree on Live and settle this matter (I have 1230+ keys and do plan on spending some of them); it's easy to do on the PTS where I can indiscriminately destroy things (I was auto-destroying the shoulders as I received them on the PTS), but I won't be able to do any big sprees on Live until I figure out what the hell to do with the shoulders that I receive. My biggest itemization complaint is that they added all these new BoP items without giving us more inventory upgrades. What the hell.

And as I've said before, I dislike Training and Prosperous (esp. Prosperous, a trait that I've been quite vocal about in the forums), and I am not defending either trait. I am, however, saying that the drop rates are even (though that doesn't mean luck can't screw you) and compared to what things used to be (when bad RNG meant you got nothing), getting a set item of a useless trait isn't that bad in comparison.

* To clarify, on the question of whether an item drops at all, I expect that to be easy to adjust--but when it comes to the question of which item drops out of a pool of possibilities, from what I have gleaned about the game's inner workings, I find it very unlikely that they would go through the kind of trouble and work that I expect would be needed to make such weightings.

Tabbycat · October 2016

It looks like the greater the sample size the closer you get to ZOS' percentages.

I'd say RNG is being RNG.

SkylarkX · October 2016

The argument "why would ZOS weight the drop table and create more work for themselves" is bogus.

a.) they've already stated that they weight the tables for other reasons
b.) this is an mmorpg

RobTheAxe · October 2016

Nobody wants training or prosperous

Why not entirely remove them?

SkylarkX · October 2016

Seri wrote: »

code65536 wrote: »

Going through my log and looking at only the first 100 logged items (86 shoulders, 14 motifs), I see:
Divines: 11 (vs. 78 in the final count)

Infused: 17 (vs. 63 in the final count)

Prosperous: 10 (vs. 56 in the final count)

Reinforced: 13 (vs. 77 in the final conunt)

Sturdy: 10 (vs. 82 in the final count)

Training: 16 (vs. 70 in the final count)

Well-Fitted: 9 (vs. 74 in the final count)

Or in graph form since people like pictures.

That said, I still wouldn't complain if prosperous/training was removed from the drop table :P

If i'm reading thing right I think the first 86 were on PTS? And the 500 total were from updates 10, 11 and 12 (all on the PTS)?

Not really valid for the topic at hand!

It seems like the only verified data that is relevant to Update 12 live is OP's video..

code65536 · October 2016

SkylarkX wrote: »

Seri wrote: »

code65536 wrote: »

Going through my log and looking at only the first 100 logged items (86 shoulders, 14 motifs), I see:
Divines: 11 (vs. 78 in the final count)

Infused: 17 (vs. 63 in the final count)

Prosperous: 10 (vs. 56 in the final count)

Reinforced: 13 (vs. 77 in the final conunt)

Sturdy: 10 (vs. 82 in the final count)

Training: 16 (vs. 70 in the final count)

Well-Fitted: 9 (vs. 74 in the final count)

Or in graph form since people like pictures.

That said, I still wouldn't complain if prosperous/training was removed from the drop table :P

If i'm reading thing right I think the first 86 were on PTS? And the 500 total were from updates 10, 11 and 12 (all on the PTS)?

Not really valid for the topic at hand!

It seems like the only verified data that is relevant to Update 12 live is OP's video..

No, the 86 and 500 are both from the Update 12 PTS. It's the same key opening spree. It was mostly to show that there is no such thing as trait favoritism. After the first 86 keys, the most "favored" trait is Infused. Well-Fitted was the least favored (almost half the drops compared to Infused) and that Sturdy and Prosperous were tied as the 2nd-least favored. The key opening continued until 500 were spent, and now what used to be the most-favored trait is now the 2nd-least favored (with a total drop rate below that of Well-Fitted), and Sturdy and Prosperous, which were tied after 86 keys, have diverged to opposite ends, with Sturdy being the most-favored trait after 500 keys and Prosperous being the least-favored.

If someone is going to posit that there's a favored trait, then tell me, what is the favored trait? If people looked at that spree after 86 keys, they'd see Infused leading Well-Fitted 17-to-9--almost double--and would naively call shenanigans. But if they keep going, then they'll end up with Sturdy leading Prosperous 82-to-56. Wait, wasn't Infused the unfairly favored trait? So why is it now at the bottom and something else at the top?

The whole notion that there is favoritism is ridiculous and unsupported by evidence. Though not quite as ridiculous as these conspiracy theories that ZOS secretly changes drop rates to be more fair on the PTS to fool people who test. Seriously? It's just luck. Some people have good luck, some people have bad luck, and you can have streaks and droughts and turns of luck. That's what luck is! If I saw a perfectly-even distribution, then that's when I am more inclined to call shenanigans because perfectly even distributions is not something that you expect to see in random data unless you start having lots and lots of data points. Like more than 500 and certainly more than 64. The notion that random necessitates a perfectly even distribution over small data sets is a common fallacy held by those who don't understand probability and statistics.

People like to ascribe meaning to things that have no meaning. It's like trying to find shapes in the clouds and stars. It's natural, but also irrational. Random is not the same as patternless, and patterns that naturally arise from randomness are just that--random patterns--and do not warrant people trying to fit a shape to them. Doing so is how we get superstitions and these silly conspiracy theories about favoritism or PTS tampering.

SkylarkX · October 2016

code65536 wrote: »

SkylarkX wrote: »

Seri wrote: »

code65536 wrote: »

Going through my log and looking at only the first 100 logged items (86 shoulders, 14 motifs), I see:
Divines: 11 (vs. 78 in the final count)

Infused: 17 (vs. 63 in the final count)

Prosperous: 10 (vs. 56 in the final count)

Reinforced: 13 (vs. 77 in the final conunt)

Sturdy: 10 (vs. 82 in the final count)

Training: 16 (vs. 70 in the final count)

Well-Fitted: 9 (vs. 74 in the final count)

Or in graph form since people like pictures.

That said, I still wouldn't complain if prosperous/training was removed from the drop table :P

If i'm reading thing right I think the first 86 were on PTS? And the 500 total were from updates 10, 11 and 12 (all on the PTS)?

Not really valid for the topic at hand!

It seems like the only verified data that is relevant to Update 12 live is OP's video..

No, the 86 and 500 are both from the Update 12 PTS. It's the same key opening spree. It was mostly to show that there is no such thing as trait favoritism. After the first 86 keys, the most "favored" trait is Infused. Well-Fitted was the least favored (almost half the drops compared to Infused) and that Sturdy and Prosperous were tied as the 2nd-least favored. The key opening continued until 500 were spent, and now what used to be the most-favored trait is now the 2nd-least favored (with a total drop rate below that of Well-Fitted), and Sturdy and Prosperous, which were tied after 86 keys, have diverged to opposite ends, with Sturdy being the most-favored trait after 500 keys and Prosperous being the least-favored.

If someone is going to posit that there's a favored trait, then tell me, what is the favored trait? If people looked at that spree after 86 keys, they'd see Infused leading Well-Fitted 17-to-9--almost double--and would naively call shenanigans. But if they keep going, then they'll end up with Sturdy leading Prosperous 82-to-56. Wait, wasn't Infused the unfairly favored trait? So why is it now at the bottom and something else at the top?

The whole notion that there is favoritism is ridiculous and unsupported by evidence. Though not quite as ridiculous as these conspiracy theories that ZOS secretly changes drop rates to be more fair on the PTS to fool people who test. Seriously? It's just luck. Some people have good luck, some people have bad luck, and you can have streaks and droughts and turns of luck. That's what luck is! If I saw a perfectly-even distribution, then that's when I am more inclined to call shenanigans because perfectly even distributions is not something that you expect to see in random data unless you start having lots and lots of data points. Like more than 500 and certainly more than 64. The notion that random necessitates a perfectly even distribution over small data sets is a common fallacy held by those who don't understand probability and statistics.

People like to ascribe meaning to things that have no meaning. It's like trying to find shapes in the clouds and stars. It's natural, but also irrational. Random is not the same as patternless, and patterns that naturally arise from randomness are just that--random patterns--and do not warrant people trying to fit a shape to them. Doing so is how we get superstitions and these silly conspiracy theories about favoritism or PTS tampering.

Of course there are people who can't see the forest for the trees, but there are also people who only see the forest. A proportion of the people who are complaining about drop rates are in all probability experiencing disproportionate drops individually, there are plausible explanations for this to be true - including the possibility that the rng system is flawed (rather than weighted).

Such a scenario would explain why some people get more consistently favourable, unfavourable or middle of the road drops while on the whole the global average remains intact.

All systems are fallible until proven otherwise. Until someone who possess absolute truth on the matter can demonstrate that the rng system is infallible, prostrating on the matter is essentially meaningless.

code65536 · October 2016

SkylarkX wrote: »

Of course there are people who can't see the forest for the trees, but there are also people who only see the forest. A proportion of the people who are complaining about drop rates are in all probability experiencing disproportionate drops individually, there are plausible explanations for this to be true - including the possibility that the rng system is flawed (rather than weighted).

Such a scenario would explain why some people get more consistently favourable, unfavourable or middle of the road drops while on the whole the global average remains intact.

All systems are fallible until proven otherwise. Until someone who possess absolute truth on the matter can demonstrate that the rng system is infallible, prostrating on the matter is essentially meaningless.

Nonsense. ZOS's RNG is not "flawed" in the way you think it is. Everything is perfectly consistent with how randomness works, and that's the point that I've been trying to drive in this thread.

That having been said, one could argue that using a random system to distribute rewards is itself flawed. The math suggests that in a perfectly fair, purely random system, there will likely be a non-trivial number of people who will never get the ideal piece that they seek. But this is not a flaw in ZOS's RNG system--this is just the inherent nature of randomness. The flaw is in their design choice to use RNG with no other recourse. This is why people--myself included--support tokens or other ways in which they can get relief from bad luck. If people have a beef with the system, it's not whether the RNG is working as intended--it's whether that intent is the right intent.