Some statistical gymnastics so the zero-drop-gang can also move on

"

Archer6621 wrote:

Joking aside, your claims are not credible if you cannot point out what is "incorrect" about the statistics,

Let's start with the fact that you were unaware that drop chance during day two was 3%, not 1.5%. Then, the average loss % being 1.8% after watching both streams. Then, the fact that 12k participants was only a Day 1 metric. Then, the fact that many participants weren't counted as eligible to begin with.

"

Zykor wrote:

"

Archer6621 wrote:

Joking aside, your claims are not credible if you cannot point out what is "incorrect" about the statistics,

Let's start with the fact that you were unaware that drop chance during day two was 3%, not 1.5%. Then, the average loss % being 1.8% after watching both streams. Then, the fact that 12k participants was only a Day 1 metric. Then, the fact that many participants weren't counted as eligible to begin with.

137,000 were. So it's safe to assume anybody who didn't get something that should have didn't update their Twitch link.

"

EstocSlayer wrote:

You didn't even know the drop rates were doubled on the 2nd day. Chances are you made more mistakes.
Also you have no credibility.

So take your fake trash biased incorrect statistics out of here.

But that part got corrected already! You're not willing to explain what is wrong but make an assumption instead. That still means you have exactly zero credibility.

Assuming you are capable of thinking along, consider this:
- Every 5 minutes, a user can get a drop with some probability p. One such event can be simulated by a Bernoulli trial.
- 12000 users were linked
- Since day 2 had double drop-chance, we take an average of the first and second day as overall drop-chance. Likely though, there were more viewers on day 1 because of the announcements, but it is not clear how many of these were linked.
- All we now need, is to know how many hours each user watched, which we need to get a from a distribution. The fitted distribution is nice in the sense that most of its mass is in the 8+ hour range, since it is likely for many of the people that linked their accounts to leave the stream on for drops.
- Having all this, we can just iterate over each user, do as many trials as their amount of hours allow, sum the result, and then we have the drops per user for that run.

What is fake and trash about this? Could you be more specific instead of discarding something you're not willing to understand?

"

EstocSlayer wrote:

I would do my own calculations but I should be doing my assignment as I have a deadline tomorrow and I also work full time.

I would love to see your approach after your deadline if you have time (I'm not joking). I'm currently a computer science master student, so I'm all too familiar with pesky deadlines, but managed to deal with mine for now.

"

EstocSlayer wrote:

The fact remains thay the problem wasn't RNG. It was the fact that GGG messed up and saw those accounts as ineligible when they actually should have been.

But this is the thing: it's an unknown. So it is hard to state that the problem was not RNG when it is not clear whether someone not receiving drops either was part of the "bug", or part of the zero-drop slice from the RNG. Be reasonable, it cannot be said with certainty without additional data.

"

Bluppins wrote:

"

Zykor wrote:

"

Archer6621 wrote:

Joking aside, your claims are not credible if you cannot point out what is "incorrect" about the statistics,

Let's start with the fact that you were unaware that drop chance during day two was 3%, not 1.5%. Then, the average loss % being 1.8% after watching both streams. Then, the fact that 12k participants was only a Day 1 metric. Then, the fact that many participants weren't counted as eligible to begin with.

137,000 were. So it's safe to assume anybody who didn't get something that should have didn't update their Twitch link.

Unsure why so many people are unable to read.

It's 137k rewards, not winners.

"

Zykor wrote:

"

Archer6621 wrote:

Joking aside, your claims are not credible if you cannot point out what is "incorrect" about the statistics,

Let's start with the fact that you were unaware that drop chance during day two was 3%, not 1.5%. Then, the average loss % being 1.8% after watching both streams. Then, the fact that 12k participants was only a Day 1 metric. Then, the fact that many participants weren't counted as eligible to begin with.

I corrected the first part.

For the second point, I found during messing around with the simulation that if you assume everyone participating has watched the full 16 hours, you will indeed find such low percentages. But this is quite unrealistic to assume for the full audience. So I assume that not everyone watched the full 16 hours, according to that distribution in the screenshot (although that distribution does have significant amount of mass in the 8+ hour range, which is reasonable, since people that linked their account probably left the stream on to get more drops). Essentially you can try this by replacing the "user_hours_s[j]" by "16" on line 82. You will indeed find a very low "loss" portion, but this number is based on unrealistic assumptions.

For the third point, it is true that we don't have the number for day 2. But if you think about it, it's not even that relevant! Just try it in the notebook, increase the number of participants, you will see that the distribution among the participants themselves will barely change (deviation will decrease as a result of the law of large numbers).

"

innervation wrote:

This experiment, noble as it is, fails to capture the most important part of all of this: the complainers don't want to understand probability, they want to be outraged.

Its good that someone did the numbers on them though, I didn't get dick and i'm not outraged :p

Tbh i can't see how you can play PoE and not understand probability at all, every league I play i got some crazy statistical anomalies.

"

Archer6621 wrote:

"

Zykor wrote:

"

Archer6621 wrote:

Joking aside, your claims are not credible if you cannot point out what is "incorrect" about the statistics,

Let's start with the fact that you were unaware that drop chance during day two was 3%, not 1.5%. Then, the average loss % being 1.8% after watching both streams. Then, the fact that 12k participants was only a Day 1 metric. Then, the fact that many participants weren't counted as eligible to begin with.

I corrected the first part.

For the second point, I found during messing around with the simulation that if you assume everyone participating has watched the full 16 hours, you will indeed find such low percentages. But this is quite unrealistic to assume for the full audience. So I assume that not everyone watched the full 16 hours, according to that distribution in the screenshot. Essentially you can try this by replacing the "user_hours_s[j]" by "16" on line 82. You will indeed find a very low "loss" chance, but this number is based on unrealistic assumptions.

"One example didn't align with the narrative I wanted to tell, therefore I made up speculative data which falls more in line with the narrative i want to tell"

Hell yeah bro gj.

This whole thread is absolutely useless, you were right about one thing though: Your title saying this "data" is mental gymnastics.

Thread fits perfectly the derogatory meaning for mental gymnastics.

The fact that op even mentioned stretching out as a sure way to get ripped muscles just goes to prove he doesn't go to the gym. Even though he claims he has a gym membership card.

"

Zykor wrote:

"One example didn't align with the narrative I wanted to tell, therefore I made up speculative data which falls more in line with the narrative i want to tell"

Hell yeah bro gj.

This whole thread is absolutely useless, good thing you've admitted it's mental gymnastics in the title.

Are you claiming that it is realistic to assume that every linked user has watched the full 16 hour stream? Because it is not. What kind of distribution of hours would you suggest otherwise?

If that's not your issue, then I'm not sure what you mean.

"

Archer6621 wrote:

"

Zykor wrote:

"One example didn't align with the narrative I wanted to tell, therefore I made up speculative data which falls more in line with the narrative i want to tell"

Hell yeah bro gj.

This whole thread is absolutely useless, good thing you've admitted it's mental gymnastics in the title.

Are you claiming that it is realistic to assume that every linked user has watched the full 16 hour stream? Because it is not. What kind of distribution of hours would you suggest otherwise?

If that's not your issue, then I'm not sure what you mean.

I'm claiming your speculative data has absolutely no credible support whatsover when you aren't even aware of the data you're working with in the first place.

This is a waste of time.