super-rare drops linked to # in existence on server?

[Navia]

Unlike other games where the RNG is queried hundreds of times per frame, I doubt that FFXI needs a bleeding fast function for random numbers. Just have an independent RNG for each zone server, should be plenty of horsepower. If they have one that has noticeably streakiness, they're doing it wrong.

In another game I worked on we had a RNG per region server but it still ran several thousand times in the span of a few seconds - it drives mob ai and pathing choices, synthing checks, spawning, loot table generation, combat rolls and other stuff.

[Norelco]

Pretty much the more experience a person has with programming, the more in the "it's all totally random" camp they are likely to be.

I'm a software engineer and I was thinking the exact opposite. I think the more you know about programming the more you realize that things are never truly random.

[aurik]

Well, what do you mean by streaky? I'm sure even your fastest random number generator still generates an even distribution of output values.

Even if I'm right, though, the other part of your post is extremely illuminating and I hope everyone who posted in this thread except aurik and a few others really reads and thinks about it.

Pretty much the more experience a person has with programming, the more in the "it's all totally random" camp they are likely to be. It's the noobs that know nothing about programming who come up with the elaborate stupid theories. Was it mikesjustice who first advocated the humongous database that contains every single NM drop and craftable item cross-referenced with each player theory?

-

That's my initial gut feeling, aurik, but after thinking about it for a sec I guess with all the combat damage generation, mob movement and stuff, I guess it's conceivable that the random number calculations might make a difference. I'm not sure, though, Navia or anyone who's actually worked on an online game would know better.

Most mob movement code is deterministic, not random (this way, the client doesn't need nearly the amount of bandwidth; it can predict most of the movement and just get periodic updates from the server). Combat RNG usage is pretty sparse, on the whole. It's nothing like a realtime fps where, say, 40-50 projectile trajectories are calculated each frame, based on a normal distribution...

[shotgn]

Hmm, that's a good point.

Let's say 5000 random number generations a second, ballpark 100 operations per generation = 500000 operations a second. 0.5 Mhz processing needed. I'd have to agree, it doesn't seem like an issue.

[Nystul]

I think most of the problem when people get into these kinds of discussions lies in the fact that the common layperson's perspective of "random" is much different from the mathematical and statistical definition of random.

For example, may would consider their itunes player set on random to be truly random if it were to never play the same song twice, and if it were to never play a song by the same band back to back. To them that would be a nice even distribution.

Statistical and mathematical random, however, says that you SHOULD have same song back to back and songs by the same band back to back.

I don't think anyone would cry foul if you flipped a coin IRL and it landed heads 8 out of 10 times. It's not supremely likely but it's very very possible. However, when it comes to this game if people don't see a 50/50 split or something very close they think something is up. Random as far as this game's programming is concerned means that streaks are inevitable. Having a constant alteration of heads/tails/heads/tails is actually NOT random. Having streaks of heads/heads/heads and tails/tails/tails that ultimately over a large number of samples adds up to roughly the same number of individual heads and tails IS random.

I'm not saying people are stupid but rather it's typically a difficult concept to grasp as it is sort of counterintuitive and abstract.


**edit**
someone smarter than I could probably whip up some thing that says a 1 in 20 drop of an item would have to go through X number of no drops before it would achieve some sort of statistical anomoly. I'd wager that the number of no drops in a row that would be allowed under a 1/20 distribution would be much higher than folks would guess.

[alt]

Improbable does not equal impossible. Most of us would consider winning the lottery a nearly impossible odds to beat. However there have been multiple people who have won the lottery twice, beating odds that are something like a million to one to just win it once. Given enough chances, anything that can happen will happen.

[Seyrr]

Quick binomial distribution with p = .05? Then to get 0 drops to be outside 2 standard deviations... someone may want to check my math, but for a quick and dirty estimate I'm getting that you'd need to be in the neighborhood of 100 trials before it started being weird (expected 5, standard deviation about 2.18.)

Edit: And in the context of the post that asked, 80 drops without would not be weird, statistically (expected 4, standard deviation about 1.95.)

[Nystul]

I'm doing Ttests right now umm...if you can tell me what the data set would look like I could run it through excel but typically speaking my handle on statistics is a tenuous one at best in terms of practical application. COMPUTER DO MATH RAWR!

[Medivac]

the only thing about consistancy in this game, is if your jp orny

[shotgn]

I don't think anyone would cry foul if you flipped a coin IRL and it landed heads 8 out of 10 times. It's not supremely likely but it's very very possible. However, when it comes to this game if people don't see a 50/50 split or something very close they think something is up.

Yeah, people become retards when they play this game. Now that's a theory I believe.

[Everlasting]

Seems on most servers if RMT claim KB they wipe.

Maybe this is SE being nice to us.

[neophytal]

Seems on most servers if RMT claim KB they wipe.

Maybe this is SE being nice to us.

SE? nice? wha?