Forum Settings
       
Reply To Thread

Let's do some science (Atma Japanese Theory)Follow

#1 Apr 13 2014 at 8:11 PM Rating: Excellent
Sage
***
1,500 posts
Basically I and some friends decided to go out and test this Atma theory, please take a look at our findings...

Experiment definition:
We ran 6 sets of tests with 4 people. Each set consisted in an hour long fate farming in a specific zone according to the theory presented in Reddit.
Note: to account for the lack of information about the time zone(Server time, Local time or Japanese time) we run 3 sets on Japanese time (which is also local time for 2 of the test subjects) and 3 sets on PST time (the time zone of our server)

We made sure that every fate started and ended within the hour we were farming. The main objective was to farm as many fates as possible, hence some times we simply engage the fate and ran to the next one, although we tried to get gold as often as possible it wasn't the priority.

Data:
- Test subjects:
Here is what you need to know about each of them:
A: Resides in Japan and has already collected a set of 12 Atmas.
B: Resides in Latin America (CST time) and already collected a set of 12 Atmas.
C: Also resides in Lain America (CST time) and has yet to collect a set of 12 Atmas.
D: Resides in Japan and has yet to collect a set of 12 Atmas.
All test subjects play in Hyperion (a PST based American server)


- First set:
We farmed 2:00 am to 2:59 am (Japan local time) in Southern Thanalan a total of 21 fates each.
A: obtained nothing during this time.
B: obtained nothing during this time.
C: obtained Atma of the SCORPION in the 5th fate.
D: obtained nothing during this time.
Total number of fates during this set: 84 (21X4)
Total number of atmas obtained: 1


- Second set:
We farmed 3:00 am to 3:59 am (Japan local time) in Upper La Noscea a total of 23 fates each.
A: obtained Atma of the WATER BEARER in the 9th fate.
B: obtained nothing during this time.
C: obtained nothing during this time.
D: obtained nothing during this time.
Total number of fates during this set 92 (23X4)
Total number of atmas obtained: 1


- Third set:
We farmed 4:00 am to 4:59 am (Japan local time) in East Shroud a total of 25 fates each.
A: obtained nothing during this time.
B: obtained nothing during this time.
C: obtained Atma of the GOAT in the 21th fate.
D: obtained nothing during this time.
Total number of fates during this set 100 (25X4)
Total number of atmas obtained: 1


- Forth set:
We farmed 2:00 am to 2:59 am (PST time) in Southern Thanalan a total of 19 fates each.
A: obtained nothing during this time.
B: obtained nothing during this time.
C: obtained nothing during this time.
D: obtained nothing during this time.
Total number of fates during this set 76 (19X4)
Total number of atmas obtained: 0


- Fith set:
We farmed 3:00 am to 3:59 am (PST time) in Upper La Noscea a total of 21 fates each.
A: obtained Atma of the WATER BEARER in the 7th fate.
B: obtained Atma of the WATER BEARER in the 19th fate.
C: obtained nothing during this time.
D: obtained nothing during this time.
Total number of fates during this set 84 (21X4)
Total number of atmas obtained: 2


- Sixth set:
We farmed 4:00 am to 4:59 am (PST time) in East Shroud a total of 20 fates each.
A: obtained nothing during this time.
B: obtained nothing during this time.
C: obtained nothing during this time.
D: obtained Atma of the GOAT in the 21th fate.
Total number of fates during this set 80 (20X4)
Total number of atmas obtained: 1


Data compilation:
					Atmas	Fates	% drop 
Total					6	516	1.2% 
	------------------------------------------------------- 
	Japan time			3	276	1.1% 
		-------------------------------------------- 
		Southern Than		1	84	1.2% 
		Upper La Noscea		1	92	1.1% 
		East Shroud		1	100	1.0% 
	------------------------------------------------------- 
	PST time			3	240	1.3% 
		-------------------------------------------- 
		Southern Than		0	76	0.0% 
		Upper La Noscea		2	84	2.4% 
		East Shroud		1	80	1.3%


Conclusion:
Ideally more data is required in order to minimize probabilistic anomalies, although it's our stance that 500+ fates and 24 hours man hours of data collection are enough to see some sort of impact in the drop rate, however as you can see in the data itself we found nothing of the sort.

The drop rate increment(if any) seems to be insignificant.
____________________________
"Maybe it means: you're the evil man, and I'm the righteous man, and Mr. Nine-millimeter here, he's the shepherd protecting my righteous *** in the Valley of Darkness." - Jules.
FFXIV: Mabel Rand (Gugnir)
FFXI: Kenage, retired.
K&K forever!,
#2 Apr 13 2014 at 10:20 PM Rating: Good
***
3,441 posts
I've voiced my opinion before, and though somewhat unpopular it might seem, after playing the game longer now, every day I am more convinced their RNG could use some tweaks.

It likes to get stuck on ludicrous streaks that should not be happening, let alone how often they happen. While lucky/unlucky streaks can always happen, having them happen every single day is a tad bit hard to swallow as just a streak.

As everybody knows, a computer cannot produce a true random number. It has to use some function of Time and an algorithm. These vary from program-to-program and are never 100% random. Some programmers use good ones, some.... not so good.

I'll give a few examples from tonight:

Every day, or at least most days, I do the Grand Company Craft List. I try to HQ everything in this list, because, well, double rewards obviously. I know that I'm not going to HQ -everything-, but I do a surprising number of them. Well... I can just tell when the game is going to ***** me over (and when the game is going to give me ridiculous success streaks).

Earlier tonight:

1). 3 Failures to HQ at 85%+. One of those a 96%.
2). Several Synths in which I failed two 90% Basic Touch (with Steady Hands up) attempts in the same synth.
3). In Gathering, on two occasions, I managed to gather 3 HQs in a row (15% each).
4). HQ'd two 18% HQ chance items in a row.

Each one of these carries a pretty small probability, and all of them happening in just one night is kinda like winning the lottery.

Problem: This happens every single night I play FFXIV.

So my conclusion is thusly:

The RNG algorithm being used in FFXIV is either borked, OR, there's hidden variables they are not showing us a'la FFXI (the whole Day of the week/Compass direction/etc crap that people STILL don't know for sure to this day).
#3 Apr 13 2014 at 11:44 PM Rating: Good
****
5,729 posts
Seriously.

I get the idea of random chance. Failing two 90% chance things in a row will happen occasionally. It shouldn't be happening all the freaking time though. I've never actually done tests and math, but I am positive that the odds of success on certain crafting things aren't what they say they are.
____________________________
75 Rabbit/75 Sheep/75 Coeurl/75 Eft/75 Raptor/75 Hippogryph/75 Puk
75 Scorpion/75 Wamoura/75 Pixie/75 Peiste/64 Sabotender
51 Bird/41 Mandragora/40 Bee/37 Crawler/37 Bat

Items no one cares about: O
Missions no one cares about: O
Crafts no one cares about: O
#4 Apr 14 2014 at 5:10 AM Rating: Decent
***
1,556 posts
Please enlighten me about your knowledge of programming and how random number generators actually work. Smiley: rolleyes

Also, let me know which PRNG SE uses for FF14 functions/subroutines.

You can start by explaining this simple program to me:

m_w = <choose-initializer>;    /* must not be zero, nor 0x464fffff */ 
m_z = <choose-initializer>;    /* must not be zero, nor 0x9068ffff */ 
  
uint get_random() 
{ 
    m_z = 36969 * (m_z & 65535) + (m_z >> 16); 
    m_w = 18000 * (m_w & 65535) + (m_w >> 16); 
    return (m_z << 16) + m_w;  /* 32-bit result */ 
}


Pretty common pseudo RNG. If you're not familiar with assembly code, you probably won't get far (despite being written in a higher level programming language).

This is a good article on PRNGs:

https://www.schneier.com/fortuna.pdf

While Fortuna is a cryptographically secure PRNG, you will quickly realize that video games do not require cryptographically secure PRNGs to get the job done.

But I feel like this is getting off topic. Please take off the tin foil hats and stop trying to find patterns in randomness. Not true randomness? Sure. Discernible by humans outside of fields like cryptography? No.

Disclaimer: RNG in computers is generally a field associated with computer science. My field is engineering so I'm no expert when it comes to PRNGs but the more complex functions discussed in comp. sci. are usually the ones used in cryptography and not a video game. I wouldn't be surprised if SE used something as basic as the multiply with carry that was posted above or the linear congruential algorithm.

What some of you fear is periodicity in PRNGs. However...

http://www.random.org/randomness/
Quote:
PRNGs are efficient, meaning they can produce many numbers in a short time, and deterministic, meaning that a given sequence of numbers can be reproduced at a later date if the starting point in the sequence is known. Efficiency is a nice characteristic if your application needs many numbers, and determinism is handy if you need to replay the same sequence of numbers again at a later stage. PRNGs are typically also periodic, which means that the sequence will eventually repeat itself. While periodicity is hardly ever a desirable characteristic, modern PRNGs have a period that is so long that it can be ignored for most practical purposes.


Anyways, if you'd like to look at examples of bad PRNGs, just load up iTunes and stick it on shuffle. One of the worst implementations of a PRNG algorithm/function that I've seen to date. :X

Edited, Apr 14th 2014 9:53am by HitomeOfBismarck
#5 Apr 14 2014 at 6:47 AM Rating: Excellent
At least with HQ synthing, if you can get to 100% quality, you will HQ end of story. And you never have to worry about breaking a synth with rare/expensive ingredients even if you don't HQ because you have absolute control over the final product (unless you accidentally hit the wrong button.)

That alone is a huge improvement over FFXI's "we might let you HQ this synth. We probably won't. We might let you fail this synth entirely despite it being 80 levels below your skill. Just because we can."

Edited, Apr 14th 2014 8:47am by Catwho
#6 Apr 14 2014 at 7:20 AM Rating: Good
*
72 posts
FFXIV is LUA programming, so I'd start with LUA's math.random() function. No magic here folks =)

math.random() generates pseudo-random numbers uniformly distributed. Supplying argument alters its behaviour: 
math.random() with no arguments generates a real number between 0 and 1. 
math.random(upper) generates integer numbers between 1 and upper. 
math.random(lower, upper) generates integer numbers between lower and upper. 
> = math.random() 
0.0012512588885159 
> = math.random() 
0.56358531449324 
> = math.random(100) 
20 
> = math.random(100) 
81 
> = math.random(70,80) 
76 
> = math.random(70,80) 
75 
upper and lower must be integer. In other case Lua casts upper into an integer, sometimes giving math.floor(upper) and others math.ceil(upper), with unexpected results (the same for lower).
#7 Apr 14 2014 at 10:48 AM Rating: Good
***
2,550 posts
mess3 wrote:
FFXIV is LUA programming, so I'd start with LUA's math.random() function. No magic here folks =)

math.random() generates pseudo-random numbers uniformly distributed. Supplying argument alters its behaviour: 
math.random() with no arguments generates a real number between 0 and 1. 
math.random(upper) generates integer numbers between 1 and upper. 
math.random(lower, upper) generates integer numbers between lower and upper. 
> = math.random() 
0.0012512588885159 
> = math.random() 
0.56358531449324 
> = math.random(100) 
20 
> = math.random(100) 
81 
> = math.random(70,80) 
76 
> = math.random(70,80) 
75 
upper and lower must be integer. In other case Lua casts upper into an integer, sometimes giving math.floor(upper) and others math.ceil(upper), with unexpected results (the same for lower).


So the point is that the randomization is controlled by an upper and a lower. For instance, a random whole number between 1 and 2 would return 50% 1's and 50% 2's, while a random whole number between 1 and 10 would return 10% 1's, 10% 2's, etc. It all depends on the upper and lower limits. Something tells me that we are sitting in a 1 to 2% bracket. Can you imagine > = math.random(1,100000)? Smiley: oyvey



Edited, Apr 14th 2014 11:48am by Valkayree
Reply To Thread

Colors Smileys Quote OriginalQuote Checked Help

 

Recent Visitors: 321 All times are in CST
Anonymous Guests (321)