Is the Gathering RNG broken for high tier? (or at least the % message)

I'm getting very long strings of the same result even when close to 50%.

Wonder if other people are getting this as well?


Lots of unstatistical fail chains of RNG happening for me. I might be a statistical anomaly (as usual) but I suspect some of the underlying RNG is off compared to the % chance that is written in the message.

For instance yesteday mining an Iron node in Graff Gems I got a string of 12 fails, with what the message said was a 52% chance of success. That is about 1½ in 10 000 chance. And I'm getting a lot of those, with 4-8 fails in a row for >50% chances according to the text.
I'm not even close to getting the same string of successes on those Iron/Copper nodes. 3-4 tops.
In that run when I looked through the chat logs I had 32 Iron/Copper node attempts. Out of those 12 was successes and 20 was fails. For a ~50% chance, that is pretty low turnout.

But only for high tier, for the low tier mining gathering like the crystals (20xp) I had a probability of 94% and only failed once out of 50ish attempts. Which is within statistical norm but still a little bit too good.

 

Not sure where you get the 1.5 chance in 10,000 from. If it's a true random chance, it is a 1 in 2 chance every time (ie.50% which is close to the 52% chance you have stated). Each event is independent, just like when you flip a coin. Each time you gather is independent from each other attempt, they are depeendant on each other.

Now they could be design to increase the chance of success ifyou fail, but from what I see they are. It's just a string of very bad luck.

 

Statistics and probability are two sides of the same coin.

Yes each event is independent, hence why I was asking for others to chime in if I am a statistical anomaly or not.
If others are also getting long strings like that, then their RNG is broken.
A 4-5 long string is commonplace, a 5-10 string would be rare, a 12 string should only happen once a blue moon.
Since my statistical sample is too small I am asking for more observations.


Lets see if my math was sound for the in 10000.
The probability for a string of 12 consecutive failures when it is a 52% chance of success.
My calculation was this : .48^12 = 0.000149587
How else would you calculate such probability?


My hunch is that the text message we get and the actual chance of the "roll" doesn't match up.

 

getting anything from 5 fails in a row, single success to 12 repeats.. 69.39% chance spectral mines nodes. seems ok.

 

My record are with garlic, 96% chance of success according to the game and average tries is about 4 but ome time I stop counting at 10 times

 

For debug/test/informative purposes, it'd be interesting to know the success percentage even in case of success.

 

It depends on what level node you're attempting because Yes, a 50% chance is 1 in 2 odds... But you could alway get "heads" and never get "tails" over and over, because each roll is new... I was just in the new shardfall and every node gave me a 52.8% chance and I failed like 10 times in a row on every node, I never harvested anything, no ore, no wood, no cotton and to top it off it took forever for the timer to count down (and I pumped a lot of points into reducing the timer)... In the valeways, I have nearly zero timer its that quick because they're low level nodes, whereas in the new novia swamps, I failed with a 95% and 86% chance a few times, but trying again with a success...

 

Doesn't matter if the roll is new every time. The standard deviation of failing 10 times in a row is orders if magnitudes from the norm. Failing 10 times in a row is so minuscule it should statistically happen only once every 1000 nodes. Getting it on 3+ nodes in a row like you did tells us that the RNG is not a brand new roll of 52.8% every time.

I will return to Graff tonight and confirm if I still have the statistical anomaly string.

 

Completely agreed. Right now I can only go on the XP value for successes.

 

I was wondering this as well. In one scene, I chopped several (~10) trees and each reported 44.something % success. Every single tree required 3+ attempts, I was paying attention since I wanted to verify this. Spoon is correct in saying that even if every attempt has ~1/2 chance of succeeding, repeated long strings of failures is an anomaly (I'm studying statistical models as a postgrad and know something about this). On average, those trees should require ~2 attempts. Either the percentage is reported incorrectly, or the RNG is flawed or not used properly. It is quite unlikely that this would happen so often for several players.

As a side note, reporting so accurate percentages to the player is something to abandon later. At the moment, they are good for testing and verifying, but at launch, it would be better to report something like 'you have 4 out of 10 chance to succeed'. It doesn't need to be precise to have a rough idea.

 

While I am not disagreeing that something is hokey with the RNG for these events, Spoon would be right if her results were the only ones in the game. They are not and therefore are only the smallest slice of the rest of the sample set. Imagine all of us looking at a kaleidoscopic view of the scattered results from this generator, each of us limited to our own little splinter of the whole, and the generator spits out 50,000 "dice" that bounce about and come to rest within. Now we are all looking for the same type of result, but let's say that under Spoon's view, nine out of the ten dice she sees have rolled a 92+, and so fail. In that scenario, even a dozen players could come out with what seem to be really improbable results. This doesn't indicate that the generator is broken, just that the results for that slice seem to be skewed. I've failed on eight sequential skinning attempts that should have had a 91% chance to succeed, only to mine a half dozen nodes successfully on the first shot that should have only had a 32% chance of success. All it means is that I was extremely unlucky for a stretch, and then uncommonly lucky for the next series.

 

Tachys, agreed that it *is* possible that this is just normal. Worth checking however. For me, it wasn't just this one test of 10 trees, I did the test because I started to feel that this particular tier (~44% on production level 2) was taking way too many attempts. While other tiers seemed normal.

 

???
Nope.

If I had made such claims then maybe, but since I didn't I really don't see where you are coming from.

Please do the math before attempting such broad dismissals.

Right now lets say that we have 2000 players trying out hard core gathering. So far in the last week they might have done some 500 tier 5 attempts each.
Then lets double that to put in a really wide safety net.
That is 2M attempts with the safety net. Among the population that would give a reasonable chance for someone getting a 21 fail streak.
However if SmokerKGB's account is correct he has had a 30+ streak. That is orders of magnitudes larger than what even the embellished sample size above gives.
A 30+ streak would take a sample size of
1 / 0.472^30 = 6 049 786 395
that would be 100k players doing 60k tier 5 attempts each.

And if SmokerKGB's account is correct then the streak is much longer than 30.

It is possible, but simply not probable. Check out what Standard Deviation and Orders of Magnitude means in these conexts.

If SmokerKGB's account is correct then your explanation meets Occams and comes out with no hairs, no skin, no meat and no bones.

So if you want to point out some flaw here then I'd go with SmokerKGB's account and not my hunch.

 

I'm well aware of the math involved, and that we're dealing with a RNG, which in all honesty are flawed in every iteration ever made, because they aren't true RNG's. People have complained about computer-based RNG's since they were created, and while it may have produced improvements over the decades, it never stops this from rearing its head.

Is his result extremely improbable, hell yes.

Should they look into it? Probably.

But beyond that, I'm not worried, because I have seen both sides of the same coin flip on this one multiple times in this release while testing out other bugs, even results nearly as improbable as SmokerKGB's.

 

It depends on which crowd you're talking to... People who play the "odds" or people who watch the "statistics" because they're really talking about 2 different things... The "odds" of something happening is a future event, and "statistics" are using data of past events...

I'm not quite sure "what" the 52.8% is talking about, it could be just if I will get 1 or 2 ore, or 3 or 4 ore, or the added bonus of tin or nickel, or just a flat success rate, since the Dev's will not confurm any of this, I can only observe my results... I would say there are multiple rolls going on in the background and they are set a little different for each resource, ore being the hardest, wood, cotton, garlic and finally hides, they all seem to have different rolls going on and nodes in different areas are different levels, like tha shardfall, all of those were high level (you can tell by looking at how long the timer is while harvesting, which actually has nothing to do with success), these nodes took a very long time to harvest, even though a fail was the results.

They even have a roll after harvesting, wheather or not the node remains for another harvest attempt, I had one iron node that gave me 10 ore total after 3 attempts, as it didn't dissolve after each attempt (I pumped a lot of points into that too)...

The shardfall I think is broken as I tried every node 10 times each (and that's all I could do because they stopped responding after 10), and I tried at least 20 nodes before giving up on all the nodes, trees, ore, cotton all acted the same... I couldn't go out of combat mode either, so I couldn't regenerate HPs, I had to leave the island, only had so many potions...

The 2 sided coin is a good example of equal odds, there is the same odds of getting Heads as there is for getting Tails, both have a 1 in 2 chance, but when flipped continuously there is niether an equal number of Heads to Tails... Go figure...