There is no way that the containers in Bruma are “random”

cozmic72 · May 11

I’m only missing 5 Cyrodiil set pieces and they’re all one handed weapons. I farm town daily quests and get way too many bows, shields and resto staffs that don’t sell on PSNA.

Stafford197 · May 11

One thing I will add to this is that I went to Infinite Archive and spent 140K currency on companion Heavy Armor and 40K currency on companion Jewelry. This is 28 pieces of Heavy Armor and 8 pieces of Jewelry. Nearly all drops were in useless traits like Augmented (buff duration), Shattering (penetration), or Focused (crit chance). To add on to this I only got a single drop of Quickened (cooldown reduction) and Aggressive (damage done) which are the most desired traits.

On the bright side, if you are making a tank, the game considers Vigorous (health buff) a bad trait so you have a good chance to get those pieces.

The whole point here is that sure it might be RNG… but I am convinced that certain drops are weighted more than others to get you to spend even more currency for the more valuable items.

sayswhoto · May 11

I'm not a statistician. Just curious on a few items.

@furiouslog
Would the first Chi-square test (with 25 items) not be valid due to having less than 5 expected values in each category?

@RicAlmighty
You could also use a binomial distribution as sort of an initial check. It would be less work than a Chi-square test. With a "success rate" of around 0.143 the probability of 9 shields or more is about 0.0058. Roughly 6 in 1000 attempts (with 25 containers) would have a similar experience, which may not seem too unreasonable given the size of this game.

However, if it was 9 shields in a row, that could be something else.

To the forum:
I know it was identified as Excel mistake, but 16.7 x 7 = 116.9%. This is over 100% chance. I felt like that should've been caught way earlier.

furiouslog · May 11

sayswhoto wrote: »

I'm not a statistician. Just curious on a few items.

@furiouslog
Would the first Chi-square test (with 25 items) not be valid due to having less than 5 expected values in each category?

@RicAlmighty
You could also use a binomial distribution as sort of an initial check. It would be less work than a Chi-square test. With a "success rate" of around 0.143 the probability of 9 shields or more is about 0.0058. Roughly 6 in 1000 attempts (with 25 containers) would have a similar experience, which may not seem too unreasonable given the size of this game.

However, if it was 9 shields in a row, that could be something else.

To the forum:
I know it was identified as Excel mistake, but 16.7 x 7 = 116.9%. This is over 100% chance. I felt like that should've been caught way earlier.

Well, next time I won't do statistics for free in public I guess, if I've had a few gummies. Yeah, I know that the sample size was an issue for the first one, but it was still an indicator of what might be happening. It's a gaming forum, not a paper for the AAAS. Feel free to do more detailed work on your own if you like.

sayswhoto · May 11

No problem. I thought the 378 item one was very thorough.

The only reason I happened to look through this thread is because it look related to a suspected RNG issue in Tales of Tribute. I just have a feeling something isn't right here, but I think I should let it go.

furiouslog · May 11

sayswhoto wrote: »

No problem. I thought the 378 item one was very thorough.

The only reason I happened to look through this thread is because it look related to a suspected RNG issue in Tales of Tribute. I just have a feeling something isn't right here, but I think I should let it go.

No! Let's figure it out! It's a thing I love to do.

Amottica · May 12

EF321 wrote: »

Opened 25 on PTS:

1xAxe
1xMace
2xSword
2xGreatsword
6xBow
1xLightning staff
6xResto
3xShield
1xNeck
2xRing

I am almost sure (no proof, just personal observation from collecting everything in Cyro with random coffers) that it first rolls type, and only then exact piece of that type:
4x1h
2x2h
6xbow
1xDestro
6xResto
3xShield
3xJewlery

If you look at it like this, bows, shields and restos should be far more common, than subtype like dagger or axe, but not more common that all 1h combined. Just speculation, I have no proof or real evidence.

Such a small sampling does open up to imbalanced results because the results of one container have no bearing on the results of the next container. Try 100 or even 200 containers on the PTS and see what those results yield.

Varana · May 12

If someone wants to spend a lot of things on stuff, it might also be an idea to rule out streaks of the random number generator. Or test if the RNG has streaks, which is my personal superstition based on anecdotal evidence.

I.e. if you're opening a lot of containers in rapid succession, then the RNG tends to give similar results. But again, this is not based on any factual evidence other than the impression of always standing in the wrong line at checkout.

EF321 · May 12

Varana wrote: »

if you're opening a lot of containers in rapid succession, then the RNG tends to give similar results.

This is very much possible, random number generators often use time as seed.

frogthroat · May 13

EF321 wrote: »

Here's 344:

That is much better. When talking about random, people expect uniform and maximum entropy. Reality isn't like that. When writing down a random number sequence, you can easily distinguish between a human trying to write random and a random number generator. The human writes uniform randomness. No repeating numbers more than maybe just a couple in a row. But in real randomness you can ... and most likely will have repeating sequences. The bigger the pool, the more likely it is.

When a human writes down "random", you could pick any short sequence and it would have close to the expected values. But with an actual random number generator you need a lot of samples. It will only approach the expected values when there are enough samples. The bigger the sample size, the close to the actual percentages you get. So 344 is an excellent start.