The RNG.. What's behind it?

[Eshelon]

I'm just currious if there are any threads or web sites that have really delved into the anomoly that is the EQ Random Number Generator. As any computer programmer can tell you, there is no such thing as a completely random number. Computers are not capable of such a feat. They simply use a seed or multiple seeds (one user given the other pulled from the server's clock) and a mathematical formula to generate a number that "seems" random.

This has got to have something to do with why the RNG seems to be so streaky.. Many times the laws of probability won't hold up with EQ's RNG. Standard deviation is violated etc.. Question is why? Is there a known answer or is it forever a SoE secret?

[Slippyblade]

It seems to me that it was just a poorly implemented beast. Frankly, the rand() library function returns a greater deviation than does the RNG in-game...

[Iannyen]

The devil. The devil is behind the RNG.

[Hethral]

Remember, there's a few thousand people pulling random numbers off the same generator all the time. One in a million chances happen to someone every few seconds.

[Xenephex]

The devil is behind the RNG.

That's silly. It's actually the Tri-Lateral Commission and the Masons.

[Inyidd]

The devil is behind the RNG.

That's silly. It's actually the Tri-Lateral Commission and the Masons.

Same thing.

--Inyidd

[Lothay]

The devil is behind the RNG.

That's silly. It's actually the Tri-Lateral Commission and the Masons.

What? I thought it was the Boy Sprouts and the Orbital Mind-Control Lasers! Or the Gnomes of Zurich....

~Lothay

[Inyidd]

Close. You're thinking of the Illuminati. They're the devil's biggest competitor....well, besides government, I mean.

--Inyidd

[Gwynna]

There are lots of possible reasons why the pseudo random number generator could be returning streaky combines. One possibility is that they are seeding the pRNG with the computer's timer before every number is generated (should be seeded only once, before the first number is generated). It's an amateurish mistake that a lot of software engineers make.

Perhaps they've implemented their own algorithm for a pRNG that doesn't work well. Or they could have simply used the default pRNG for their compiler, which might not work well for the number ranges being generated. Historically, many of the default pRNGs in some very popular compilers have had terrible number distribution. It's not until you start using compilers (or interpreters) intended specifically for simulation that you tend to get "decent" pRNG algorithms and choices for which ones to use.

The fact is, though, Sony has decided that either the problem doesn't exist or it's not worth fixing, and that's unlikely to change unless someone can report a duplicatable exploit involving the prediction of number sequences. It's unlikely that that will happen, though, and if you can't predict the numbers well enough to exploit it (I'm sure many have tried), then perhaps it's "random enough", even though it's frustrating for us to watch a long string of fails in a row.

I will be the first to agree that there's a problem. When a truly random number generator is asked to spew out a series of numbers in the 1-100 range, eventually it will spit out a string of 10 "100's" in a row. If it didn't it wouldn't be truly random. However, it's pretty obvious that the in-game pRNG is far more "streaky" than true randomness would allow. FAR more streaky.

If I were developing the game I'd look into it, but that's because my perfectionistic streak is overriding my need to give the shareholders their best-return on investment, and that's a bad decision for a producer to make from the standpoint of those shareholders. In any large company, a good producer (not just of games, but of anything) is not the one that generates a perfect product, but one that generates the most profitable product. Droves of people aren't quitting over the problems with the pRNG, so why should they fix it?

Given Sony's track record here, it's a pretty safe bet that it will never change, and I don't really blame them. The software engineers themselves might even know the problem exists and WANT to look into it, but when a bug is identified the producers have to determine how many developer hours (dollars) it's worth to get it fixed. If the problem can't be fixed within this budget, then it simply isn't. If an exploit were found and published involving the pRNG, then obviously the value to Sony for fixing this bug would rise. Until that happens (and my guess is it never will), don't expect them to even look at it. There are always bugs that are higher on their priority list, or other projects that developer resources could be more profitably used for.

My advice: close your eyes and suck it up, because it's unlikely this particular weakness in the code will ever change.

[eqNichola]

Whenever this topic comes up, I always point people at this rigorous statistical study of EQ's RNG, complete with formal hypotheses, evidence, and conclusions.

/e wanders off whisling a little tune...

[Benaveriel]

Actually, everyone "feels" that the EQ RNG is streaky, and more so than it should be.

I've not read any proof that the EQ RNG is worse than any other RNG, or even stating it as something more significant than a feeling. Any truly random process will have streaks in it, but people tend to notice streaks when they happen much more than the absence of them.

However, I so vaguely remember someone posting a statistical analysis of the RNG using /ran which showed, using a statistically significant sample size, that it was actually much more truly "random" than expected.

Unfortunately, I can't find this again, it was a long time ago. I can't even remember where I read it though it might even have been on the EQ traders Corner Ezboard.

It did cover /ran and not pass/fail attempts at anything, but until I see a study on it, proving that it is really isn't any good, I will carry on accepting that the streakiness is largely down to people's perceptions.

[KyrosKrane]

Nichola, that's just mean. You really had me going there...

I played a lot of Magic Online a while back, and similar accusations (of a biased RNG) were levelled at Leaping Lizard, the developers. Their response was exemplary of how such a situation should be handled. They published the actual algorithm that they used, including the relevant academic references proving that it is indeed a random system -- or at least, sufficiently random. They conducted millions of tests using sample sizes in the millions, the results of which were quite conclusive that for all intents and purposes, their RNG really was random. They published all of this for the beta testers, reviewers, and nay-sayers to review.

It really is the most effective way to prove the effectiveness or superiority of the RNG. Show how your system works, point to references validating it, and show actual test results of runs with huge sample sizes using your RNG. Publish it all openly so anyone can inspect it for themselves.

I still admire LL's confidence and panache in doing this. Every time someone came along claiming the deck shuffler wasn't random, I took great delight at pointing them to the info and telling them to find the flaw or quit complaining.

[Xenephex]

I am one of those who, after a skill-up on a combine, will click, click, click as fast as my finger will move in order to get that next combine so I can get another skillup. Having said that....

1. Even if they were using a perfect RNG, most humans would perceive it to be streaky (everyone except maybe Russell Crowe).

2. Anomalous streaks (especially long streaks of failures) will tend to get reported.

3. 3D graphics programming, game programming and everything else that is required to produce a MMORPG are heavy-duty, hard-core topics. The developers who wrote EQ are obviously heavy-duty, hard-core programmers, who clearly take a lot of pride in their work. As much as we like to complain, they have done a remarkable job in making this program as functional and as stable as it is.

Guys like this LOVE stuff like random number generators. They are probably perfectly capable of having an on-going discussion for weeks on this one subject (believe me - I've heard it). The chances that these guys produced a slip-shod or streaky RNG for EQ strike me as next to nil.

4. They are never going to tell us how their RNG was implemented, for obvious reasons.

5. No matter how good their RNG is, some people are occasionally going to get very lucky, and some people are occasionally going to get really screwed. It has to happen.

6. Bottom line: If I had to bet, I would bet that they have implemented a very good, fair as possible, RNG. I would also bet that we (including me) will all continue to perceive it otherwise.

[Itek]

*groan*

The pRNG in EQ is probably not significantly better or worse than most default computer RNGs. They are Pseud-RandomNumberGenerators because computers are really bad at math.

0 + 0 = 0
0 + 1 = 1
1 + 0 = 1
1 + 1 = 0

That's -exactly- as far as most non-techno-geeks need to delve into "why does the RNG not work PRECISELY as well as a real RNG" ... computers are really bad at math. Making them better at math is what Computer Scientists get paid a LOT of money to do.

If you can develop a really nice, smooth bell-curve RandomNumberGenerator the world will beat a path to your door. Or more specificly Bally Gaming will. (they make a LOT of computer based slot machines, making a better slot machine would make $$ in the real world)

Let's face it. SOE has to make a decision.

1) Spend time and money paying programmers to write a better RNG, which may NEVER produce measureable results.

2) Spend time and money paying programmers to write EQ2 and LDoN. Which have ship dates and measureable profits that can be handily relied on.

Think think... think think think... (Itek does his Whinny the Pooh impression)

Now let's say the pRNG took the character name as a seed. Then "Lothay" might have a much easier time than "Itek" for various reasons. That would be BAD, and should be changed.

But, ever since the "/rand loot roll cheating hack bug fix" several years ago it's been safe to presume that the RNG is strictly server-side and reliably unbiased.

Say you are working on a tradeskill in the range of skill 80. You can expect roughly 1 skill in 5 attempts. So you do a stack of 20. You get 4 skill up. Another person says "I got no skill, rotten streaky RNG" and another person says "I got 8 skill, yea me"

Now for the reality check.

All those that can change the RNG raise their hands.
*cue the crickets chirpping*

Let's move on then shall we. We all wish the RNG were ... looser. Giving us more successes, more skill ups, letting us hit more often, crit for more, dodge better, see more plat drops, have the "rare" item drop just when we need it, and see our /rand 100 come out on top...

Ever heard of a place called "Las Vegas" ... ?? When walking in the door you should ask yourself "how can they have such a nice fancy building when they give away money all day long?" The answer is, they do not. EQ exists to make money. The way it does so is by getting people to pay for their subscriptions every month. The way they do that is for the game to be enjoyable for the maximum amount of time. They aren't going to FIX the RNG because it's not broken. It just gives the House (SOE) an advantage.

But the guys that said "the devil" were close. Stick a mouse in a cage and give it a "random reward" feeder which randomly kicks out pellets. The mouse will work itself to DEATH for the ever increasingly random reward. Give it an "always dispenses" feeder and it will only eat when it's hungry.

[BrellRakshasas]

As any computer programmer can tell you, there is no such thing as a completely random number. Computers are not capable of such a feat.

Meaningless statement. Computers are capable of generating pseudo-randoms that are statistically indistiguishable from true random numbers for any desired sample size. Further, with the simple addition of an analogue white-noise sampler, true random number generation is possible.

Now, has the RNG been analyzed? We don't know. Is the seed being pulled from a good semi-random source? We have no idea about the answers to this question.

However, it's pretty obvious that the in-game pRNG is far more "streaky" than true randomness would allow. FAR more streaky.

Is it? It's not obvious to me that that's the case.

I do statistical analysis for a major international financial company for my day to day job. While I don't do work with analysis of computer algorithms relating to randomness, I do understand the mathematics and what that implies in terms of observed effects.

I seriously doubt many here (including me) are sufficiently in-tune with what "random" means that we can just intuitively know that the "streakiness" we see is in-fact outside of what we should expect.

People generally, psychologically expect randomness to look somethign like this:

101010101010110010110100100101100110

But that isn't what "random" looks like at all (in fact, that is a very patterned non-random string).

Random usually really looks more like this:

11100100000000001110101010111001001111

You see long runs of the same number. You see "streakiness."

Now, toss in additional complicators like having an 70 or 80% chance of success and I'm not entirely sure that anyone who hasn't studied the statstical properties of random numbers in great depth and detail (well beyond a mere graduate degree in statistics) is going to find anything about this (or any RNG) to be "obvious" from merely observing the limited (and probably self-selected) sample set we are privy too.