Square Enix's RNG

[gintoki88]

Anyone else feel like SE's RNG is a complete joke?

I have been crafting all night getting about 80-90% chance of HQ and swear I have failed about 50% of the time.

Meanwhile I barely get any HQ with 15-20% chance HQ.

Same goes for gathering. Several misses in a row on 80% chance.

They really need to update their RNG formula.

/rant

I ranted about this a few months a go

http://forum.square-enix.com/ffxiv/t...s-somewhat-off

[Rivienne]

No fancy math required here. If it's not 100% then there's always a chance you wont get HQ. Simple as that.

It isn't fancy math.. It is about statistics: which I didn't bring up, SE did by putting chance percentages next to these values in the first place. 25% chance means you have a 1/4 chance of getting it, and 3/4 of failing, the "fancy" math is just about what that translates into for more than a single chance.

Random is random(or pseudo-random in this case obviously). But once you apply a statistical value too it, like SE has done, then it can be expected to conform to that number within certain tolerance levels over larger sample sets, otherwise that number is meaningless.

I am not complaining here: I think things function okay-ish. Just that the numbers SE has given aren't accurate to what is going on from what I have seen thus far, but they do mostly conform to some pattern. Just not the one stated by the number SE gave. Subtract 10% (for any value between 10% and 99%) seems a good rule of thumb for me. (Though it likely isn't accurate to what is actually going on to cause that perception)

[Tiggy]

I am not complaining here: I think things function okay-ish. Just that the numbers SE has given aren't accurate to what is going on from what I have seen thus far, but they do mostly conform to some pattern. Just not the one stated by the number SE gave. Subtract 10% (for any value between 10% and 99%) seems a good rule of thumb for me. (Though it likely isn't accurate to what is actually going on to cause that perception)

As you say this is entirely your perception, and that perception is always biased. Specifically, human perception draws greater attention to things that fail or disappoint. As you say a 25% change means 1/4, but in statistics there is no expectation that every 4 attempts will provide exactly 1 success. Every given event is random and irrelevant to each other attempt. They are independent observations. As such runs of success or failures are entirely possible especially when you consider how many players are in game at once all causing the RNG to activate. In a great enough sample size it is entirely likely that the RNG does in fact fit within the tolerances expected of it. Everyone needs to chill out and realize their opinions are entirely biased and have absolutely no statistical accuracy on whether or not the RNG is broken or not. Someone, anyone, should start a spreadsheet and begin recording their observations. Once we have a few thousand instances at a specific percentage( Ie. at 99% for 1000 synths I had 900 successes and 100 failures for an actual fail rate of 10%) then we can say for sure if this is broken or not. Until that time this is all, 100%, subjective.

[coffeeking]

RNG is fine, in life you also don't get what you want. get used to it

[Rivienne]

As you say this is entirely your perception, and that perception is always biased. Specifically, human perception draws greater attention to things that fail or disappoint

Actually, I only inserted myself into this thread to insert some hard numbers of a reasonable quantity to be potentially meaningful. I performed a specific test to determine usefulness of unearth given the need for cordial/waiting for GP (not to find a flaw in the system): those results are large enough that they appear to show a skew in the system. You have deemed those numbers not enough (or not noticed them), that is fair, I already said they aren't enough to make a definitive conclusion: but that they are suspect.

I do in fact have hard experience with this type of system, so I am not exactly speaking in ignorance, but more... I am not sure why there is this huge push back against the possibility of this system being flawed. It is very hard to get a reasonable approximation of probability distribution right, and mistakes can easily happen.

Naturally people who are frustrated will complain: but that doesn't mean they don't have a potentially valid complaint. "Numbers or it didn't happen!" "More numbers or it didn't happen!" "ALL the numbers or it didn't happen."

Probability and statistics is about not needing all the numbers, just a sufficiently large sample size. I performed a test of a reasonable size to suggest there may in fact be a problem. I agree more numbers should be done independently to attempt to validate them. That doesn't mean my numbers, or my for the purpose of game play conclusions based on them, are not worth mention until then.

[Black91CRX]

I melded some materia for a friend the other day and failed a 35% chance about 8 times before giving up. The next person I melded for needed a 7% chance and I nailed it first try. It can be frustrating I know. But a 50% chance means you can fail EVERY time. Doesn't mean that 5/10 tries will succeed.

[Sin_Nombre]

I would actually like to know what random number generator they're using, and whether or not it has passed all the tests out there for random number generation. There are plenty of bad generators out there, and if they wrote the one they're using themselves, then there's a high probability it IS broken.

I've seen events happen more than a few times that simply should not have occurred, like failing over and over again with 98 and 99% chances for success. And by not occurring, I mean that the probability of these events is on the order of 1:1,000,000,000, and they occur all the time.

They've already had to fix issues with loot drops in coil, so it really makes you wonder what else is broken in this domain.

[ArkhamNative]

You can't draw any conclusions, none at all, from a single attempt. Not even 5-10 attempts. I shook 10 coins in my hand and tossed them on the table. 7 heads, 3 tails. Chance of a heads result? Around 50%.

The sting of failure is absolutely more intense than success. I NQ'd two separate synths at HQ:98% (class quest item) on different days this week. Table flip time. As for the mat synth HQs that came from HQ < 20%? Even at HQ:6%? Not even an in-chair dance.

I have seen that when I macro synth even around 100 base mats, the HQ results are close to the percentages shown. So I'm happy that it seems to be working correctly. When I gather, I have seen a 90% fail 2/5. I've also seen a 35% succeed 4/5. Usually it's ball park accurate. The only pattern is that it seems to be random.

[ruinedmirage]

I like to think of all RNG based concepts in this game as a 50/50 chance, whether loot, gathering, or crafting; either I got it or I didn't. Same as playing the lottery. You either win or you didn't. Numbers may have SOME relevance, but surely you can't tell if they're taking effect or not.

[Krystalynn]

Dies can be weighted. It's historically one of the most assured things a gambler around the world can know. Everything from slots, to roullette have often weighted or skewed results thrown in the seemingly random paradigm that they appear to be.

To think that SE does not, and has this perfectly random system in place is optimistic at best. None of us know the seed they use in their database. It would take months to gather the data, if not years before anyone could have a truly accurate result of what SE actually provides rather than what we expect.

"R"NG is the nature of any gambling practice. Go to Vegas and see just how random it truly is when you count the empty handed heads that walk in and out the door every day on every casino. Only the operator knows the truth.