RNG is not random.

[gpofcore]

The skillup formula is now known have a success rate of about 1/20 post 190 skill with a failure rate of 19/20. The chances of having x failures in a row is (19/20) to the x power. The problem can be seen in the smithing forum thread http://mboards.eqtraders.com/eq/showthread.php?t=16837 where one player had 400 failures in a row and another veteran tradeskiller had a worst case of 238 failures in a row.

The chance of 400 failures in a row is 1 in 213,000,000.
The chance of 238 failures in a row is 1 in 200,000.

Since this shouldn't actually happen it indicates that the RNG must not be generating true random numbers.

A single player can only get 480 skill points post 190. The norm for 480 skill points would be one run of 120 failures in a row. It's not realistic to assume that there have been millions of post 190 skillups in EQ's history.

[Lickity]

A single player can only get 480 skill points post 190. The norm for 480 skill points would be one run of 120 failures in a row. It's not realistic to assume that there have been millions of post 190 skillups in EQ's history.

Okay, I'll bite, I'm up for a good conspiracy, but first lets play with the numbers. For the sake of argument lets assume that everyone who has ever skilled post 190, took it all the way to 250. That is 60 skillups.

Zeralenn's 238 straight failures would be a 1 in 200,000 chance if a true random number generator was being used.

(from smithing thread where this is a similar topic)

Those odds seem large, but keep in mind, that this is a single skillup, and each player who can skill past 190, can go to 250, for a total of 60. Now 60 X lets say 100,000, as there are at LEAST that many GMs or some tradeskill.

60 x 100,000 = 60,000,000

Now all things straight, we have 60 million skillups, divide the chance for a streak of 238, and you have 300 people who are going to have the bad luck of 238 straight without a skill up.

Now, looking at our boards, we see a lot of people who get streaky losses like these, and it seems like an uneven distribution...but...look how few people report the same streaky successes. People only complain when they are hurt, and if they get one really stellar streak of wins, it might show up in a rave on the primal scream boards. The other thing to remember, is that we represent the hardcore tradeskilling community, meaning the people most likely to do multiple tradeskills, thereby increasing our individual odds of obtaining a critical streak of non-skillups. At this high of concentration, and only Ngreth could tell me the board population as far as I know, it is perfectly likely that we would end up with a lions share of 200 + combines without a skillup. Take into account that the people reporting their lack of skillups often have lower WIS/INT than the 355 the 5% is based off of, and you once again have a mitigating factor on our statistics.

I am an easy sell on a conspiracy, but this one lacks some meat.

[Tweedledea]

:dedhorse:

[Cubwynn]

The skillup formula is now known have a success rate of about 1/20 post 190 skill with a failure rate of 19/20. The chances of having x failures in a row is (19/20) to the x power. The problem can be seen in the smithing forum thread http://mboards.eqtraders.com/eq/showthread.php?t=16837 where one player had 400 failures in a row and another veteran tradeskiller had a worst case of 238 failures in a row.

The chance of 400 failures in a row is 1 in 213,000,000.
The chance of 238 failures in a row is 1 in 200,000.

Since this shouldn't actually happen it indicates that the RNG must not be generating true random numbers.

You are trying to extend a probability argument that isn't relevant. The only thing the RNG is determining on a combine (skill up-wise) is if that particular combine has generated a skill up. The chance is about 1:20 (per previous threads).

If you get a 20-sided die and roll 400,000,000 20's, what is the chance that the next roll will be a 20?
1 in 20.

This is the reason Las Vegas continues to rake in money... people don't know how to use logic correctly, and assume that probability means that they'll "get their due". In reality, people can have spectacularly good OR bad luck in games of chance. Such is the nature of randomness.

As long as they don't contaminate my Bingo packs with their bad luck cooties, that's okay with me...!!!

[gpofcore]

Sorry, didn't mean to start an argument but this is relevant now that the skillup formula is known. Before that all discussions about the RNG were speculative.

I'm not suggesting a conspiracy but simply that the RNG used for this isn't that good at generating consistently random numbers.

I agree that 1 in 200,000 is going to happen but 1 in 813 million? Adjusting for that persons actual conditions, 214 skill with 315 int at a 222 triv makes the actual chance of 400 failures in a row 1 in 659 million. And as pointed out this isn't even the worst case. These are big numbers in a fairly small universe of players. I'm not saying there is anything wrong with using an RNG that doesn't generate consistently random numbers but simply saying this is what we seem to have.

For those of you that have it, what was your biggest string of failures?

[kiztent]

You are trying to extend a probability argument that isn't relevant. The only thing the RNG is determining on a combine (skill up-wise) is if that particular combine has generated a skill up. The chance is about 1:20 (per previous threads).

Yes, and given that, it's possible to calculate the probablity of n failures in a row. Which was done. Correctly.

The chance that the next combine will yield a skill up remains 1 in 20, but that doesn't mean it's very improbable that you will get 400 failures in a row.

But just because improbably events happen, doesn't make the RNG not random. In fact, the improbable become pretty probable given enough time for truely random systems.

[Bindamel]

There does appear to be more here than meets the eye. Take a look at the odds of getting consecutive skillups for a moment. The first skillup in any streak is independent. What I mean by this is that the odds of the first skillup happening is 100%. The reason for this is the starting point of the streak is not arbitrary; we only start thinking about getting a streak of skillups when we get one.

Using a skillup chance of 5% (1/20):

The chance of getting two skillups in a row is 1/20, and of getting 5 in a row is 1/160000. 1/160000 on the failure side is 234 combines without a skillup.

I have never heard of anybody getting 5 skillups in a row at this level; I have heard of many people going 200 combines without a skillup.

This could be a bias in the random number generator, although I remember Scott Hartsman mentioning that it was checked; or a misinterpretation/misrepresentaton of the formulae along the way. Or maybe people are just way too embarrassed to admit when they get 5 skillups in a row.

[kiztent]

Actually, it's worth noting that 5% is the chance of a skill up for y=2, a successful combine and some insanely high primary stat.

It's a maximum in other words.

[Itzena]

But just because improbably events happen, doesn't make the RNG not random. In fact, the improbable become pretty probable given enough time for truely random systems.

Another point is that people don't come here to post that they got the average number of skillups over a range of combines - only the extremes (good and bad) generally get threads started about them...or even remembered for that matter.

[Lilosh]

The skillup formula is now known have a success rate of about 1/20 post 190 skill with a failure rate of 19/20. The chances of having x failures in a row is (19/20) to the x power. The problem can be seen in the smithing forum thread http://mboards.eqtraders.com/eq/showthread.php?t=16837 where one player had 400 failures in a row and another veteran tradeskiller had a worst case of 238 failures in a row.

The chance of 400 failures in a row is 1 in 213,000,000.
The chance of 238 failures in a row is 1 in 200,000.

Since this shouldn't actually happen it indicates that the RNG must not be generating true random numbers.

Thanks for the laugh.

The "Random number generator is failing to be predicatable, and hence can't be random" argument always makes me laugh.

Random means that any given combine has X% chance of a sucess or skillup. Trying to apply that to a small set of 100,1000, or 10,000 combines will be flawed at best.

Truly random numbers, as far as computers are capable of random numbers, can include the possibility of flipping a coin 100 times and getting all heads. It's possible, and with millions upon billions of people rolling and flipping, it's downright probable.


-Noish

[gpofcore]

From http://mboards.eqtraders.com/eq/show...tempts+skillup

In the fan faire write-up (or somewhere similar, but recently) one of the Developers was quoted as saying something along the lines of ...

"When I joined the team I took the RNG algorithm and ran it for a billion cycles. It -is- broken, in that it's PERFECTLY random, with totally equal distribution.

[Itek]

I Do Not Frigging BE-LIEVE that this has come up A-GAIN!!

Humans are shockingly bad at noticing patterns. They commonly do NOT see patterns where there ARE patterns and almost ALWAYS see them where there are NONE. Over and over and over again.

Listen to the words "it's broken in that it's PERFECTLY random" ...

Then go back and read "a BILLION" again and contemplate how LONG it will take you to test that one combine at a time.

*Itek waits patiently for the poster to return*
*Itek invents and uses a cryo-sleep chamber to assist in the waiting*

ONE GUY had a streak of 400 in a row making him the most unlucky guy to hit the 213 million to one negative outcome. Now let's examine how many of us have hit 400 combines. Oh, wait, let's check instead (as we all have) how many combines we, the players of EQ, have hit in total. (There are 8000 users of this board, I'd be safe in saying each averages 8000 combines [haha] and that they represent 1/30th of the playerbase or less.)

That's 1.92 BILLION combines.

What are the odds of a 1 in 213 million to one probability occurring in 2 billion rolls? Dang near bloody CERTAIN.

People win the PowerBall(tm) lottery nearly like clockwork every two months or so. The odds are roughly 120 million to one.

Listen... carefully....

Pseudo-Random-Number-Generation is a "mature technology" ... It's been around forever and there are rarely innovations which dramatically improve performance that would be visible to end users. The RNG of EverQuest(tm) is probably more tuned to performance (considering the number of "random" request the servers must submit in a day) but it's distribution is certainly more than adequate for any practical purpose.

What would be proof of a "broken" RNG? Well if the published formula were radically proven to be non-correlative.

Didn't follow that? Ok...

If the average was supposed to be 20 and after 20 people did 20 skill ups each on the same formula with the same stats and all the results were OUTSIDE 3 standard deviations from the expected norm... then you could BEGIN to have a case.

BUT!!!

And I can't stress this enough....

What, precisely, could be done about it?

The RNG is the RNG. They certainly aren't going to take "your word for it" that it's broken and institute a new one. You can't change the RNG any more than you can change the color of the sky. It is what it is. Literally the ENTIRE GAME is based upon it. If ONE THING in EQ has to work each and every time without fail... it's the call to "random."

We can scream until we are blue in the face "the RNG is broken" ... but the hard fact is that it's ...

1) Not broken

2) Probably one of the best p-RNG's around on a technical basis

3) Extensively tested

4) Watched like a newborn baby panda

5) A ... indeed THE ... reality of the game

6) Here to stay

[Lilosh]

<3 Itek <3

[Lickity]

What I said in more words, I hope this is crystal clear now. Can we sticky this so we don't have it come up again.

[Kieroth_whiteleaf]

Well, given that it's not broken.. would someone please break it next time I'm working on skillups so that I can actually get a decent streak going?

I'd really like to break 200 in something someday. Smithing preferred. I think I'm on about 30 trying to hit 194.

Ugh. Hell levels are teh suck.

*I* still think the RNG is broken, because it's not directly benefiting *ME* :P

EDIT: the one thing I *would* like to see on the issue is that a lot of people have seemed to notice short-term streakiness - I've noticed it a bit and I see other people commenting on this. I'd be interested to know if any devs have tried running a good sized test sample (like a billion numbers or so) in batches of 1000 or so, and see if the distributions average out in the short term for the most part. I've seen more than a few posts where people have done large-scale testing of it (like the above quote with the dev running a billion) but I don't think I've ever seen it tested for short term stuff.
Just a thought.

SFG