90% sucess rate and still missing

[Astarica]

The whole 'sample size' is really a pet peeve of mine. If you really got 12 misses in a row with 90%, that's not a small sample. The chance of that happening is 1 trillion in 1. It is a virtual statistical impossibility. If you have proof this actually happened I can almost certainly agree with you that the RNG is indeed messed up. Though I'm sure failing 12 times in a row at 90% didn't actually happen and is only an exaggeration.

At any rate the whole RNG being good or not completely misses the point. Let's say we have a coin flip simulator and all my RNG really does is gives the result opposite of the last roll made by the server, so it just alternates between heads and tails forever. Must be a pretty bad RNG right? No, because after you rolled once, you don't know if there are 0, 1, or 100 guys who since then requested a coin flip to my server, so to you there is no pattern despite the RNG is 100% predictable.

[Tarrick]

The whole 'sample size' is really a pet peeve of mine. If you really got 12 misses in a row with 90%, that's not a small sample. The chance of that happening is 1 trillion in 1. It is a virtual statistical impossibility. If you have proof this actually happened I can almost certainly agree with you that the RNG is indeed messed up. Though I'm sure failing 12 times in a row at 90% didn't actually happen and is only an exaggeration.

When you come to terms with the fact that the order in which random events happen doesn't matter, you will see that it's a small sample. If there were 1 million successful consecutive events prior to that, no one would be here complaining about them. They would, however, be a necessary component of the analysis of the RNG.

Though I'm sure failing 12 times in a row at 90% didn't actually happen and is only an exaggeration.

This is definitely the more likely scenario.

[Astarica]

When you come to terms with the fact that the order in which random events happen doesn't matter, you will see that it's a small sample. If there were 1 million successful consecutive events prior to that, no one would be here complaining about them. They would, however, be a necessary component of the analysis of the RNG.

Um, failing 12 in a row has exactly one ordering (12 in a row). The chance of this event happening is a trillion in 1. To observe such an event means either you're extremely (un)lucky or there's something wrong with the system, with the latter being the far more plausible explanation given the extreme odds involved. It's likely this event never occured in the history of FF14 thus far, and while that's not proof it cannot happen, one should be extremely suspicious if it did.

[Tarrick]

Um, failing 12 in a row has exactly one ordering (12 in a row). The chance of this event happening is a trillion in 1. To observe such an event means either you're extremely (un)lucky or there's something wrong with the system, with the latter being the far more plausible explanation given the extreme odds involved. It's likely this event never occured in the history of FF14 thus far, and while that's not proof it cannot happen, one should be extremely suspicious if it did.

And yet, even in your other posts you discuss that the sequence of the randomly generated values are not limited to a single character or event. So is this really 12 fails in a row? If the RNG is a shared resource then no it isn't, and it even approaches truly random when you have other unpredictable entities accessing that resource. Unless someone comes in with a statistically significant sample, then there's no reason to even consider it.

[emblasochist]

And yet, even in your other posts you discuss that the sequence of the randomly generated values are not limited to a single character or event. So is this really 12 fails in a row? If the RNG is a shared resource then no it isn't, and it even approaches truly random when you have other unpredictable entities accessing that resource. Unless someone comes in with a statistically significant sample, then there's no reason to even consider it.

To this end, I will record two sample sizes of 100 gathers each, one with a stated success rate of 83% success and one with a stated 81% success rate. I'll record only the nodes with + gathering yield, no bonus or + gathering attempts, but, before I do so, I believe that the stated percentages are pretty reasonable estimates...

[O-Deka-K]

To this end, I will record two sample sizes of 100 gathers each, one with a stated success rate of 83% success and one with a stated 81% success rate. I'll record only the nodes with + gathering yield, no bonus or + gathering attempts, but, before I do so, I believe that the stated percentages are pretty reasonable estimates...

I work with random numbers all the time. I have to read reports on random numbers and determine if they follow given percentages "correctly" or not. That being said, 100 samples is not even close to being a significant sample size. At an 83% success rate, I would not be surprised at all if you get 65 successes or 95 successes. If you want to compare 81% to 83% for accuracy, then I'd say that you would need over ten thousand samples (each), and that's still on the low side. I'm completely serious. And no, I don't expect anyone to do this.

The deviation is greater than you think, and yes, it annoys me too.

[Xbob42]

I work with random numbers all the time. I have to read reports on random numbers and determine if they follow given percentages "correctly" or not. That being said, 100 samples is not even close to being a significant sample size. At an 83% success rate, I would not be surprised at all if you get 65 successes or 95 successes. If you want to compare 81% to 83% for accuracy, then I'd say that you would need over ten thousand samples (each), and that's still on the low side. I'm completely serious. And no, I don't expect anyone to do this.

The deviation is greater than you think, and yes, it annoys me too.

You don't expect people to do this? You're new to MMOs, aren't you? Someone will do this for FUN.

[Astarica]

And yet, even in your other posts you discuss that the sequence of the randomly generated values are not limited to a single character or event. So is this really 12 fails in a row? If the RNG is a shared resource then no it isn't, and it even approaches truly random when you have other unpredictable entities accessing that resource. Unless someone comes in with a statistically significant sample, then there's no reason to even consider it.

The existence of other unlikely event doesn't affect any particular event you're trying to observe. Yes the RNG of FF14 most likely handed out an event with odds of 1 trillion to 1 at some point, but the event we're trying to observe (say, crafting) most certainly didn't happen a trillion times, so we should be skeptical if such an event happened. The observance of a 1 trillion to 1 event itself is statistically significant, which is why claims to observe such an event are almost always false.

[Ilmoran]

If you really got 12 misses in a row with 90%, that's not a small sample. The chance of that happening is 1 trillion in 1. It is a virtual statistical impossibility.

The thing is, rolling a 10 sided die 12 times and getting 12 1's in a row is no less likely than rolling the specific sequence: 1, 3, 7, 7, 5, 2, 4, 3, 9, 0, 8, 2. The chance is also 1 trillion in 1. The difference is you actually care when you get 12 1's, you don't care when you get what you perceive to be a random sequence. In truth, 12 1's is a statistically insignificant sample size to prove that a 90% chance isn't really 90%.

That said, most computerized random number generators are actually pseudorandom number generators (they for the computer to do it, there must be a formula or algorithm; usually, it involves a seed value and a function of time from a set reference point), and are prone to streaking, but even out over large samples.

[Astarica]

No inexplicable event can be explained by anything other than extreme luck or a flaw in the system. If something has 1 trillion to 1 chance of occuring, you're not supposed to observe it unless you've unbelievably luck or the system is flawed. You can't explain this away with 'small sample'. If it happened, it's either crazy luck or the system is broken, with the latter being the far more likely outcome. Of course these events are almost always exaggerations, but if they happened you can say with near confidence that the RNG is not only flawed, it's flat out wrong.