Issues With Rng Allocation

[DBugII]

Edit: There are many things that this thread does not answer. This is just to give a general idea as to how the system is currently and how it can be made better.

 

This is a General RNG Thread, not just in regards to 2 specific parts hardly anyone has been able to get. The goal is to help explain that the "consistency" or "allocation" issues people have been having with the RNG.

 

The problem with all RNGs is one thing: They aren't random. In fact, they are very predictable. The term is Pseudorandomness. Now, a lot of this gets technical quickly, so I'll explain each step as I proceed. The goal is try and explain the reason why RNGs are more friendly to certain items over others with the same drop-rate.

 

Allocation, or "true drop-rate" as I will go into soon enough, is the placement of these items in a magical "list" known as "loot".

For example,

[Guns, Beer, KFC, Walmart, Money].

If the list was just as I ordered it, we'd have enough chicken to end world hunger - Followed by Walmarts, and tons of booze for our hulky Rhinos. Guns and Money on the other hand, they'd be more of rarity. This is because, from experience, RNGs often avoid the first and last entries in allocation - If the list were to double in size, the allocation of loot would be horrible for items at the end or front of the list because the volume in the middle will grow. This is explained further down, as well as a way to vastly improve it.

 

THIS IS IMPORTANT:

1. RNG is used to determine the loot obtained. It does not determine what loot is available

2. RNG does not like things to "end".

3. RNG is just one part of a whole system!

 

To fix this bad allocation, we manipulate it as follows:

[Guns, Beer, KFC, Beer, Money, Walmart, Guns, Walmart, KFC, Money]

In theory, the allocation, and drop-rate, would be the same because all items show up twice. However, this won't be the case, KFC is now the rarest drop.

Why is KFC the rarest drop? Because BOTH entries are further toward the ends of the list than the rest of the loot. However, it still has a better allocation than Money and Guns did with the previous list(because it exists twice in the list).

 

If we wanted to repeat the results of before, the list would look like this:

[Guns, Guns, Beer, Beer, KFC, KFC, Walmart, Walmart, Money, Money]

This fixes very little. Though indeed, like with the previous list all items appear twice. However, their allocation in the list has not changed: Guns and Money is still toward the end of the list. The ease of getting both does improve slightly, because they don't exist just on the very ends of the list.

 

So, how can we manipulate allocation and make certain items harder while making other easier?:

[Guns, Beer, Beer, KFC, Walmart, KFC, Walmart, Guns]

Guns are now the hardest resource to get, while KFC and Walmart are the easiest. Beer also now has an uncommon, but somewhat likely chance of dropping. Hopefully by now, this pattern of allocation has been established.

-----------

Now, lists of 0-9 (Computers count from 0, not 1) are cute and easy to work with, but most games use values which range from 0-255(Ie. Pokemon) while others may shoot up way higher.

 

Hopefully, this'll help answer some frequent questions and concerns. I agree with DE in saying that the RNG is working as it is intended to, and I do not blame the RNG for the consistency issues as it is a problem with all RNGs with bad allocations.

I do believe the allocation, or "loot", tables need to be adjusted to fix the rarity of items with a similar drop-rate of others.

Edited May 27, 2013 by DBugII

[Shion963]

+1 good explanation. Are you a statistics master, by chance? ;p

 

AFAIK there are at least 3 tiers of rarity (Common, Uncommon, Rare) with common having more weight compared to rare.

So taking a ? alert as example (and ignoring the area, credit, and timelimit fields) it'll have to pick out whether it has any ? at all, then pick out its rarity, and THEN a reward from that rarity's list. That's like 1/4 * 1/10 * 1/(however many items are in that particular rareness).

[Excitonex]

I don't understand how you determine that the RNG has a larger chance of rolling a 5 than a 0 or 9. On average sure, but it doesn't seem to make sense that it would choose the items from the "middle" of the table when the computer has no way to differentiate between the middle, the start, and the end of the table.

[JollierThanThou]

I invite you to put your RNG hypothesis to the test. Try going to tryclj.com and entering the following code:

(repeatedly 100 #(rand-nth ["Guns" "Beer" "KFC" "Walmart" "Money"]))

Let me know how the results stack up against your expectations!

[Excitonex]

I'd say the algorithm they use to determine randomness is suspect. Using a RNG from another company, using a different program, proves very little. It does show that this is a possibility but it in no way proves that this is the RNG system that DE uses for Warframe.

[JollierThanThou]

I'm not trying to prove much. Just that the OP's blanket statement about RNGs is misinformed and possibly hurting his experiences in all games.

[DBugII]

I'm not trying to prove much. Just that the OP's blanket statement about RNGs is misinformed and possibly hurting his experiences in all games.

Actually, it's based on the testing of several different RNG engines from both videogames and other sources. If my hypothesis is indeed wrong, then the only remaining explanation is that DE's RNG is, in fact, flawed. Since they have more credibility, and the fact their RNG seems to do the same thing I've seen from others, I have doubts that it's just the RNG that is at fault here unless these results were promoted intentionally or out of human error.

Like RNG, allocation is just another part of a whole system. As I also stated, lists consisting of 5 items scattered into small list is cute and easy to work with, but it does not serve to fully explain the repercussions of allocations which use lists far, far greater.

 

I don't understand how you determine that the RNG has a larger chance of rolling a 5 than a 0 or 9. On average sure, but it doesn't seem to make sense that it would choose the items from the "middle" of the table when the computer has no way to differentiate between the middle, the start, and the end of the table.

I'm saying there's a better chance of RNG rolling a [3, 4, 5, 6] than 0 or 9.

Edited May 27, 2013 by DBugII

[JollierThanThou]

Actually, it's based on the testing of several different RNG engines from both videogames and other sources. If my hypothesis is indeed wrong, then the only remaining explanation is that DE's RNG is, in fact, flawed. Since they have more credibility, and the fact their RNG seems to do the same thing I've seen from others, I have doubts that it's just the RNG that is at fault here unless these results were promoted intentionally or out of human error.

Like RNG, allocation is just another part of a whole system. As I also stated, lists consisting of 5 items scattered into small list is cute and easy to work with, but it does not serve to fully explain the repercussions of allocations which use lists far, far greater.

 

I'm saying there's a better chance of RNG rolling a [3, 4, 5, 6] than 0 or 9.

I'm very surprised with your experiences, then. I've used a lot of RNGs and played a lot of video games. The RNG from Golden Sun is the only measurably nonrandom one I've seen.

[CrazyCanadian24]

This thread... It is amazing.

[DBugII]

I'm very surprised with your experiences, then. I've used a lot of RNGs and played a lot of video games. The RNG from Golden Sun is the only measurably nonrandom one I've seen.

I'm just going to leave this here.

[SeraphicRadiance]

This is quite awesome. Unfortunately, there are few things truly randomly generated that work in that manner, and it's fact that values for different items are specifically assigned in Warframe. In an unbiased environment, you would be correct :D.

Edited May 27, 2013 by SeraphicRadiance

[DBugII]

This is quite awesome. Unfortunately, there are few things truly randomly generated that work in that manner, and it's fact that values for different items are specifically assigned in Warframe. In an unbiased environment, you would be correct :D.

That's kind of what I'm hoping Warframe would adopt to better the looting system and RNG used currently.

[SeraphicRadiance]

That's kind of what I'm hoping Warframe would adopt to better the looting system and RNG used currently.

 

I would be in full support of this system. Following a stone set system rather than individually fiddling with drop values would be much better.