PLEASE fix the random number generator [Archive]

dartie

06-07-2007, 12:31 PM

It's correct for you to say that we don't know for sure. However, I'll say that I'm satisfied the RNG is working adequately and that when people think it isn't, the perception is a product of the "clustering illusion" discussed in other threads.I know it can "feel" like you are getting shafted. Yesterday, I just needed one natural mana to make an adornment. I bought 4 T3 master scrolls because I know there's supposed to be a 25% chance of getting a mana as opposed to an infusion when I transmute fabled items. All 4 gave me infusions. I bought 2 more. 2 more infusions. I was running low on money, so I started buying one at a time and transmuting as I went.After I turned 12 M1s in a row into infusions with no manas, I gave up and bought a Natural Mana. Because of the hope and emotion I had invested in getting a mana, this experience was very frustrating. All by itself, it might lead one to conclude that the RNG isn't up to snuff. But of course, you really can roll a 4-sided die 12 times in a row without getting a 4. That's part of the nature of randomness. You aren't allowed to have one of these discussions without someone pointing out that your sample is statistically insignificant, just as my sample of 12 masters is statistically insignificant.But once you start operating on a much larger scale, you see that the percentages reported by the devs and confirmed by various members of the community really do hold up to scrutiny. When I need to replenish powders and fragments, I fill up a 32-slot box with treasured items from T1. I transmute those and then fill up the box with items from T2 all the way to T7. I have to go through this process at least once a week to keep my shop stocked. I've blown up thousands of items in sets of 32, and I've found that I tend to end up with something close to 8 powders(sometimes as few as 3, but once 15) for every 32 items I transmute. I can almost count on getting 7-9 powders when I transmute 32 items at a time.I'm extremely skeptical of many of the claims that the devs make (or, more frequently, have attributed to them) on these forums. But as for the RNG, I think it's working just fine.

archimidesX

06-07-2007, 12:31 PM

what you are asking for isn't to fix the random number generator...you are asking them to remove it entirely and put in some kind of predictability pattern...

einar4

06-07-2007, 12:31 PM

Actually the generator has to be what is called a pseudo-random number generator, as any deterministic algorithm is not going to be truly random. However, it is fairly advanced, all things considered. The issue is that, as one number theorist said, "Lady luck has a poor memory." The meaning is that with each toss of the coin, the odds are the same, 50-50. Even though a thousand throws, in hindsight, may show a random distibution of heads vs tails, on EACH toss, there is a 50-50 chance. So it doesn't matter if the last 5 tosses were all heads, the sixth toss is right back to a 50-50 shot of heads vs tails, and there is no prediliction for a change in the streak just because of past throws. Many amateur gamblers have learned this to their chagrin, increasing bets progressively as they sense the "streak" will be broken (and subsequently losing their shirts).

archimidesX

06-07-2007, 12:33 PM

<cite>Kizee wrote:</cite><blockquote><cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote>It definately is not random....it is very streaky.Having the same breast plate drop of Mayong 3 out of 4 times when the mob has a 24+ item loot table isn't really randomThe same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> roll a 30 sided dice a thousand times and see how much streak you see...flip a coin a thousand times see how many times in a row it lands on one side... random does not mean it won't have streaks...

Ama

06-07-2007, 12:38 PM

<cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote> Cusashorn I think what he means is fix it so it's random, but not that random. Maybe put in an identifier that states if X class item has been dropped it must move onto another class item. This way if a conjy glove set drops it moves on and forces it to pick from the 1 of X classes that are in the group. It would still be random but be more "Fair" which I know some people don't like. Though from what he's describing it sounds like the random event generator fouled up. Remember a problem similar to that when I was programming. Had a random number generator pull a number out but the program malfunctioned. It gave me the same random number I asked for in the beginning again for some reason.

kenm

06-07-2007, 12:52 PM

There's a difference between an occasional streak and flat out totally broken. Take this for example: Our guild server firsted the coercer BP off Mayong last month, absolutely no other guild on the server had gotten one through *countless* kills. Next week? Coercer BP. Week after? Coercer BP. In fact, in those three "streaky" kills two of them also dropped troub BPs as well. What the hell are the odds that absolutely nobody gets one for ages, and then we go and get the first three on the server in three consecutive kills? Recently we also had the same crap with Treah, server first the fury hat and then the same hat next kill. We constantly see the same loot over and over again on the majority of the mobs, it gets made especially clear something is [Removed for Content] with the two shared loot table vampires in MMIS. More often than not the exact same pair of gloves will drop both kills in a run. And whatever it is, you can be sure somebody already has it. Just last night in MMIS we transmuted every piece of loot except the Tactician's (which went to an alt), and Mayong's. Fun, fun. It's completely idiotic, even all through KoS massive amounts of stupid streaks were happening. Five Shadow Axe's in a row, countless Fists of Bashing in a row, tales of 10 Animist Tunic's in a row? Sheesh, more often than not every piece of loot comes in streaks! It's dead obvious that the random number generator is seeded from something not all that random at all.

Skylher

06-07-2007, 12:59 PM

what you can do is have the classes not kill that particular mob that already have their pieces. That is what we do to get by the so called RNG. We had zerker and SK bp's dropping constantly, so they either logged out or logged in an alt. Of course once our gaurds have their bp, not much we will be able to do about it, can't really kill it without the tanks LOL

Armawk

06-07-2007, 01:02 PM

Recently it was announced that mob drops at some levels (?) would not be random any more but would be biased towards the classes present in the group killing the mob. Food for thought.

Nolrog

06-07-2007, 01:22 PM

<cite>archimidesX wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote><cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote>It definately is not random....it is very streaky.Having the same breast plate drop of Mayong 3 out of 4 times when the mob has a 24+ item loot table isn't really randomThe same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> roll a 30 sided dice a thousand times and see how much streak you see...flip a coin a thousand times see how many times in a row it lands on one side... random does not mean it won't have streaks... </blockquote> True randomness will always have streaks. I had a professor once, that gave the assignment to flip a coin 100 times and record the results. The next class, he would look at the results and tell who had actually flipped the coin, and who just made it up. The ones that flipped the coin had runs of 5 or 6 in a row, while the ones that made it up did not.

Dasein

06-07-2007, 01:31 PM

While there may indeed be streaks in an RNG, for a player, there will not be enough iterations to move beyond a streak, so while you may see what amounts to randomness over the course of 1000 attempts, that's largely irrelevent to the player who might only get in a dozen attempts.

kenm

06-07-2007, 01:31 PM

<cite>Nolrog wrote:</cite><blockquote><cite>archimidesX wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote><cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote>It definately is not random....it is very streaky.Having the same breast plate drop of Mayong 3 out of 4 times when the mob has a 24+ item loot table isn't really randomThe same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> roll a 30 sided dice a thousand times and see how much streak you see...flip a coin a thousand times see how many times in a row it lands on one side... random does not mean it won't have streaks... </blockquote> True randomness will always have streaks. I had a professor once, that gave the assignment to flip a coin 100 times and record the results. The next class, he would look at the results and tell who had actually flipped the coin, and who just made it up. The ones that flipped the coin had runs of 5 or 6 in a row, while the ones that made it up did not.</blockquote>You know, I actually don't understand the coin example people use. If you flip a coin in the exact same way it's going to land on the same side every freaking time. /boggle

Siclone

06-07-2007, 01:33 PM

the op is right,,,,,right now there is nothing Random about it.we got the same dirge gloves from labs, 3 times, in 3 chests in a row. Give me a break.its broken,,,thats not random, you have better odds playing the lottery then that happening on a random generator.

Raveller

06-07-2007, 01:34 PM

<cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote> RNGs are not truly random. RNGs are programmed code and are given a seed number, which can, and in many cases should, be changed with regularity. RNGs actually do generate results in a predictable pattern, if you know how the RNG was programmed and what seed number it is using. You could also discern the pattern by annalyzing a large volume of its output.

Siclone

06-07-2007, 01:36 PM

<cite>Nolrog wrote:</cite><blockquote><cite>archimidesX wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote><cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote>It definately is not random....it is very streaky.Having the same breast plate drop of Mayong 3 out of 4 times when the mob has a 24+ item loot table isn't really randomThe same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> roll a 30 sided dice a thousand times and see how much streak you see...flip a coin a thousand times see how many times in a row it lands on one side... random does not mean it won't have streaks... </blockquote> True randomness will always have streaks. I had a professor once, that gave the assignment to flip a coin 100 times and record the results. The next class, he would look at the results and tell who had actually flipped the coin, and who just made it up. The ones that flipped the coin had runs of 5 or 6 in a row, while the ones that made it up did not.</blockquote>I wonder if you even understandthe topic,,,,,getting 5 heads on a row on a 50/50 chance is not what we are talking about.there are 24 classes, and what 12 solts ,,,I mean the random chance of getting 3 swashie breast plates in a row,,,,,,those odds are amazing, and its happening to allot of groups/guilds

Raveller

06-07-2007, 01:37 PM

<cite>shaunfletcher wrote:</cite><blockquote>Recently it was announced that mob drops at some levels (?) would not be random any more but would be biased towards the classes present in the group killing the mob. Food for thought.</blockquote>I would like to know which level mobs were changed. For myself, I never see a mob drop an item that either I or any of my alts can use, which tells me that if there is a class check of some kind being done that the bias is against the classes present (and may also be biased against all alts as well) rather than in favor of classes present.

Themaginator

06-07-2007, 01:41 PM

its random...its i mean its random...and...oh boy

Kaalenarc

06-07-2007, 01:58 PM

The thing is - it IS random. You killed Mayong 4 times "in a row" and got three of the same item. Wow, that makes the odds SEEM astronomical, until you factor in that your four kills over X number of days wasnt the only times he was killed. Consider this: On your server over the same period, maybe he was killed 10 times, or 20, or 7. If the odds of getting , say, the swashy breastplate are 1 in 24 - then you have an equal 1 in 24 chance. A run of three times in a row isnt such a stretch. Now , if ALL servers were complaining that everyone was getting 3 of 4 of the same item - THEN they might say that the RNG is somehow screwed. But, as it is, while it IS frustrating, it is NOT broken. Personally I like what they did with most of the EoF instances, making the loot drops random within the class range of the group. No more Paladin/Warden/Fury/Conjy/Wiz/Ranger groups getting the dirge class hat LOL

Dasein

06-07-2007, 02:00 PM

So while it may be random, this randomness is obviously not the optimal way to handle loot, especially for an encounter that will be done fairly infrequently (once a week, or so, as opposed to multiple times per day).

Kaalenarc

06-07-2007, 02:19 PM

I agree - Id MUCH rather see a "each person chooses from a list" type scenario.

Raveller

06-07-2007, 03:06 PM

Kaalenarc@Antonia Bayle wrote: <blockquote>I agree - Id MUCH rather see a "each person chooses from a list" type scenario. </blockquote>I really don't see that happening. EQ2 is designed from the ground up as a stamp collecting game, and a major feature of stamp collecting is the uncertainty of when you will obtain the desired stamp.

Iseabeil

06-07-2007, 03:31 PM

<cite>Kizee wrote:</cite><blockquote>The same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> Just have to chime in on that one. Im not 100% certain as the old forums are gone, but quite sure that both loams/metals and meat/hides had their results tweaked sometime long ago, as loams before tinkering was in much, much lower demand then the other possible harvest just as meat was less needed then hides, and were adjusted slightly in their return bias.As for the rest of the randomness.. The streaks that often happen are annoying, but I think much of the problem comes from the perception of what random is. The 'human' view of random is of something totally unpredictable (and in most cases fair), whilst real application of RNG doesnt follow that idea. To add upon that, making algorithms that are truly random is a science in itself, and the most cutting edge versions wont be found in a game but in more serious applications, wich leaves us with somewhat solid RNG that in most cases wont live up to what most people percieve as truly random.Personally, I think RNG is slightly overused in EQ2, as it dominates too many scenes where possible alternatives could have been thought up, but not sure they could change that by now even if they wanted to.

Kizee

06-07-2007, 03:45 PM

Kaalenarc@Antonia Bayle wrote: <blockquote>The thing is - it IS random. You killed Mayong 4 times "in a row" and got three of the same item. Wow, that makes the odds SEEM astronomical, until you factor in that your four kills over X number of days wasnt the only times he was killed. Consider this: On your server over the same period, maybe he was killed 10 times, or 20, or 7. If the odds of getting , say, the swashy breastplate are 1 in 24 - then you have an equal 1 in 24 chance. A run of three times in a row isnt such a stretch. Now , if ALL servers were complaining that everyone was getting 3 of 4 of the same item - THEN they might say that the RNG is somehow screwed. But, as it is, while it IS frustrating, it is NOT broken. Personally I like what they did with most of the EoF instances, making the loot drops random within the class range of the group. No more Paladin/Warden/Fury/Conjy/Wiz/Ranger groups getting the dirge class hat LOL</blockquote>This isn't as simple as flipping a coin with 2 sides. There are 24+ items in Mayongs loot table and to get the same freekin thing 3 out of 4 times is just plain wrong. Also, we have been clearing FTH since EoF came out and still haven't seen bloodcrusted band or the crit heal plate boots drop.To bad the change they did that makes it so an item doesn't drop when a certian class isn't there doesn't fix the problem. Now we get to see the same loot drop over and over again for the people that are in the raid.All I know is that I have been stuck at 3 pieces for a long long time....not because we can't kill the mobs but the [Removed for Content] things refuse to drop templar gear. It's getting to a point that I really don't feel like going to raids anymore since there is a slim to nil chance that something will drop.

Armawk

06-07-2007, 04:10 PM

<cite>Raveller wrote:</cite><blockquote><cite>shaunfletcher wrote:</cite><blockquote>Recently it was announced that mob drops at some levels (?) would not be random any more but would be biased towards the classes present in the group killing the mob. Food for thought.</blockquote>I would like to know which level mobs were changed. For myself, I never see a mob drop an item that either I or any of my alts can use, which tells me that if there is a class check of some kind being done that the bias is against the classes present (and may also be biased against all alts as well) rather than in favor of classes present. </blockquote> actually its only:"EoF Fabled and Legendary class armor distribution will now consider the makeup of those who participate in the encounter. " in gu34. I thought it was more than that.

Kaalenarc

06-07-2007, 04:20 PM

<cite>Kizee wrote:</cite><blockquote>Kaalenarc@Antonia Bayle wrote: <blockquote>The thing is - it IS random. You killed Mayong 4 times "in a row" and got three of the same item. Wow, that makes the odds SEEM astronomical, until you factor in that your four kills over X number of days wasnt the only times he was killed. Consider this: On your server over the same period, maybe he was killed 10 times, or 20, or 7. If the odds of getting , say, the swashy breastplate are 1 in 24 - then you have an equal 1 in 24 chance. A run of three times in a row isnt such a stretch. Now , if ALL servers were complaining that everyone was getting 3 of 4 of the same item - THEN they might say that the RNG is somehow screwed. But, as it is, while it IS frustrating, it is NOT broken. Personally I like what they did with most of the EoF instances, making the loot drops random within the class range of the group. No more Paladin/Warden/Fury/Conjy/Wiz/Ranger groups getting the dirge class hat LOL</blockquote>This isn't as simple as flipping a coin with 2 sides. There are 24+ items in Mayongs loot table and to get the same freekin thing 3 out of 4 times is just plain wrong. Also, we have been clearing FTH since EoF came out and still haven't seen bloodcrusted band or the crit heal plate boots drop.To bad the change they did that makes it so an item doesn't drop when a certian class isn't there doesn't fix the problem. Now we get to see the same loot drop over and over again for the people that are in the raid.All I know is that I have been stuck at 3 pieces for a long long time....not because we can't kill the mobs but the [I cannot control my vocabulary] things refuse to drop templar gear. It's getting to a point that I really don't feel like going to raids anymore since there is a slim to nil chance that something will drop.</blockquote> Hmm. I wonder if , of the 24 items - the % isnt balanced? Is it possible that say the templar gear has say a 5% change to drop and maybe tank gear has 12% or something? I dont think I?'ve ever heard that discussed. Have the DEVs ever said its an equal chance? Or is it weighted based on class ?

StormCinder

06-07-2007, 05:40 PM

I'm trying to sift through the posts here to find out who among us has seen the code and hence KNOWS that an RNG is actually in use?I mean what's the difference between "random" and "seems random?"Everyone seems to use the existance of the RNG as a starting point in these debates and goes on to discuss the merits of it. I contend that the algorithm is completely designed with no random number generation code whatsoever. Someone please prove me wrong.SC

archimidesX

06-07-2007, 06:20 PM

<cite>StormCinder wrote:</cite><blockquote>I'm trying to sift through the posts here to find out who among us has seen the code and hence KNOWS that an RNG is actually in use?I mean what's the difference between "random" and "seems random?"Everyone seems to use the existance of the RNG as a starting point in these debates and goes on to discuss the merits of it. I contend that the algorithm is completely designed with no random number generation code whatsoever. Someone please prove me wrong.SC</blockquote> it is a standard assumption based on common genre game theory. Random number generators are the heart of RPGs ranging from the good ole pen and paper(might as well chalk dice up to being an RNG) i mean i can't be for sure that there is an RNG used to assign loot, but i can make the assumption. Why would one design an algorithm that would be in place to screw some, yet provide others with what they want...because i can guarantee you for every person who sees absurd streaks in loot that there are those who don't see that same streaking. i mean i certainly can't prove your theory wrong, but i can say i disagree and contend that an RNG seems more likely...

StormCinder

06-07-2007, 06:35 PM

<cite>archimidesX wrote:</cite><blockquote><cite>StormCinder wrote:</cite><blockquote>I'm trying to sift through the posts here to find out who among us has seen the code and hence KNOWS that an RNG is actually in use?I mean what's the difference between "random" and "seems random?"Everyone seems to use the existance of the RNG as a starting point in these debates and goes on to discuss the merits of it. I contend that the algorithm is completely designed with no random number generation code whatsoever. Someone please prove me wrong.SC</blockquote> it is a standard assumption based on common genre game theory. Random number generators are the heart of RPGs ranging from the good ole pen and paper(might as well chalk dice up to being an RNG) i mean i can't be for sure that there is an RNG used to assign loot, but i can make the assumption. Why would one design an algorithm that would be in place to screw some, yet provide others with what they want...because i can guarantee you for every person who sees absurd streaks in loot that there are those who don't see that same streaking. i mean i certainly can't prove your theory wrong, but i can say i disagree and contend that an RNG seems more likely... </blockquote>If this theoretical RNG exists, and it is working for some and not for others, I would say that the problem lies in other areas of the code, and not the RNG.We have an ongoing argument in our office about RNGs. Some feel they are more efficient, others feel that it is lazy coding. A good programmer can write efficient, random-seeming code without the crutch of an RNG (can you tell which side I'm on?)SC

TuinalOfTheNexus

06-07-2007, 06:56 PM

It would make more sense from a programming perspective to be using pre-generated values; if you think of the number of random numbers you need per second for even just a 24 person raid that's a lot of CPU time if you're calculating a fresh value each time. Probably still viable but it's easier just to pregenerate a few MB of numbers then read from the list.I think whatever process it is, it's functional and not bugged. Random doesn't mean fair or even distribution. It's entirely probable if you run MMIS enough times you'll get double drops. Since they changed the code to have classes on the raid influence the loot it's even more likely than it used to be. It's poor game design, caused by a desire to slow guilds in EoF due to lack of high end content, and a move away from the armor mold system back to the dark ages of raid loot (EQ1 pre-Velious). If you want to complain about anything it's the dev time wasted on things like Neriak when the current raid content is really lacking.

Eriol

06-07-2007, 06:57 PM

Get them to store the output of all the randoms on all the servers for a weekend (so you have the same influences to the PRNG as you normally would, and not just a "bare run"<img src="/smilies/8a80c6485cd926be453217d59a84a888.gif" border="0" alt="SMILEY" /> to a file. Then run the <a href="http://en.wikipedia.org/wiki/Diehard_tests" target="_blank" rel="nofollow">Diehard Tests</a> on them. Do it for 4 weekends just to make sure you didn't have a bad one. I somehow doubt they'll come even close to passing. Then implement a <a href="http://en.wikipedia.org/wiki/Mersenne_twister" target="_blank" rel="nofollow">Mersenne Twister</a> as the PRNG and we can at least be somewhat certain that we're not getting screwed. Edit: I mean for the rolling purely. Even though it's a fast one, I doubt it's fast enough for all the randomness necessary in the game.

Wossname

06-07-2007, 07:51 PM

Or use a real cryptographic quality system with an entropy pool that gets stirred regularly with unpredictable data. All the open source Unix's have an implementation. The run time costs really aren't much. Personally I suspect it is more likely to be a glitch in the code that chooses loot from the classes present.

Liral

06-07-2007, 08:03 PM

Some of the rep;ies in here are downright frightening in their lack of common sense. You're the types casinos love because you think that because the wheel has coem up black 5 times in a row red just HAS to come up on the next spin. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" />

Liral

06-07-2007, 08:07 PM

<cite>einar438 wrote:</cite><blockquote> Actually the generator has to be what is called a pseudo-random number generator, as any deterministic algorithm is not going to be truly random. However, it is fairly advanced, all things considered. The issue is that, as one number theorist said, "Lady luck has a poor memory." The meaning is that with each toss of the coin, the odds are the same, 50-50. Even though a thousand throws, in hindsight, may show a random distibution of heads vs tails, on EACH toss, there is a 50-50 chance. So it doesn't matter if the last 5 tosses were all heads, the sixth toss is right back to a 50-50 shot of heads vs tails, and there is no prediliction for a change in the streak just because of past throws. Many amateur gamblers have learned this to their chagrin, increasing bets progressively as they sense the "streak" will be broken (and subsequently losing their shirts). </blockquote>Casinos make a fortune off of people who don't realize things like this. I worked in casinos for years and have seen that mindset in play so many times.

Decad

06-07-2007, 09:49 PM

<cite>Liral wrote:</cite><blockquote>Some of the rep;ies in here are downright frightening in their lack of common sense. You're the types casinos love because you think that because the wheel has coem up black 5 times in a row red just HAS to come up on the next spin. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" /></blockquote>Well, I don't have my Probability and Statistics textbook available (just moved and can't find my textbook box), but I do believe that pretty much everything in this thread would fall under a normal distribution.Thus as somepoint, as we get further away from the mean value, it becomes more and more unlikely that something should happen. Even almost to the point where you can safely say it's not going to happen. You have to think of a normal distribution as a bell curve that starts infinitely close to 0, shoots up to some value, and then returns infinitely close to 0.As far as roulette goes, it might be that after 10 straight black spins, you could safely say it's going to be red. I forgot the formula, but one could figure it out. Problem is, it will never be 0.Now applying this logic to the OP's problem, one could argue and figure out that getting the same item 3x over could be quite probable as it does not lie that far away from the mean value of it's associative probably range.However what might prove interesting is to see what would the normal distribution would look like for getting the same 3 items 3 times over. Figuring that one out might prove that there is something wrong.And then again, one of the first problems that we figured out in my probability class was, "What is the probability of 2 people in the room (was like 38 people) having the same birthday?" The answer was like 76%. And in fact we have 2 pairs of people sharing the same birthday. Who would've thunk?

TuinalOfTheNexus

06-07-2007, 11:04 PM

Decadre@Najena wrote: <blockquote>Thus as somepoint, as we get further away from the mean value, it becomes more and more unlikely that something should happen. Even almost to the point where you can safely say it's not going to happen. You have to think of a normal distribution as a bell curve that starts infinitely close to 0, shoots up to some value, and then returns infinitely close to 0.As far as roulette goes, it might be that after 10 straight black spins, you could safely say it's going to be red. I forgot the formula, but one could figure it out. Problem is, it will never be 0.Now applying this logic to the OP's problem, one could argue and figure out that getting the same item 3x over could be quite probable as it does not lie that far away from the mean value of it's associative probably range.However what might prove interesting is to see what would the normal distribution would look like for getting the same 3 items 3 times over. Figuring that one out might prove that there is something wrong.</blockquote>I think you might need to re-read the textbook <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />It's funny I think no field of maths has as much misconception as probability. I think it's probably down to human nature; the very notion of "luck" seems inherent to us.After 10 straight black spins the odds of red are - gasp - exactly the same as on the first spin. If you think about it logically - a Roulette table has no memory. It's simply throwing a ball into a wheel. You're as likely to get 1000 straight reds as you are any other combination. And the poster who says they worked in a casino is also right in saying this is the #1 folly for gamblers (well, roulette is a game you're bound to lose anyway). People always look for and notice what seem to be patterns; it's a subtle trick and one reason roulette is so hypnotic is it allows for so many patterns to seemingly appear; number sequences, colours, locations on the table... None of which in fact have any bearing on the next spin. The only way you beat a roulette table is by cheating, or if you have an ability to predict roughly where the ball will fall on the wheel the moment it's released based on the rotation speed and release velocity. That or if you get lucky the first time you play, and walk away and never come back. In fact the "smartest" way to play roulette is to bet all the money you'll ever put on the game on a single spin - which gives you approximately a 47.5% chance of winning. Every spin after that decreases that chance.Translating this to Everquest 2, just because you got a Star Darkened Longbow off Mayong last time you killed him, doesn't mean he's less likely to drop one next time. If one of the named in MMIS drops Guardian gloves, the next one is equally likely to drop them.

Niffoni

06-07-2007, 11:41 PM

I'm just glad to see there are still some people who understand the concept of probability. Each time you roll the dice, everything resets. It seems someone is calling for "true randomness" when what they really want is "fairness". There is an enormous difference. If you get 50 Guardian BPs in a row, it is no more "strange" than getting a nice spread. Any combination of drops is just as likely as another.

Jrral

06-08-2007, 12:23 AM

<cite>Liral wrote:</cite><blockquote>Some of the rep;ies in here are downright frightening in their lack of common sense. You're the types casinos love because you think that because the wheel has coem up black 5 times in a row red just HAS to come up on the next spin. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" /></blockquote>True, but on the other hand if one person was sitting at the roulette wheel and it consistently, spin after spin for hours on end, gave him twice as many blacks as reds, the casino would be shutting down that wheel and tearing it apart to find out why it wasn't working right. And they'd probably be frisking that player up to and including a cavity search to find any devices he might be using. Because the chances of the wheel showing that kind of pattern for that long purely by random chance are just too low. The whole basis for a casino is that, while the behavior of the last few events doesn't let you predict what the next one will be, the behavior of a very large number of events over time mirrors the statistical odds very very closely. That is, if I spin the wheel 5 times I may get 4 reds, but if I spin it 100,000 times I'm going to get very close to 47,368 reds, 47,368 blacks and 5264 greens (remember zero and, because American casinos are greedy, double-zero) because over the long run all the streaks and clumps and "non-random" behaviors tend to even themselves out. Every game in a casino (where you play against the house, at least) is built on that assumption, and if it weren't true casinos wouldn't still be in business. I've seen this in the EQ2 RNG myself. Two accounts owned by the same person, harvesting raws. Now, from night to night the results vary, some nights are good and some are bad. But consistently, across every character on both accounts, every zone, every tier, all hours of the day and days of the week for an entire year, one account consistently does 3-4x as well harvesting as the other. With that large a sample, the confidence that these two samples represent statistically different populations is well above the 99% mark. Which is a big red flag to me that the RNG has some statistical problems.

CoLD MeTaL

06-08-2007, 12:54 AM

because this game is written in c++, the RNG is could be based on the MAC address you connect through, thus 1 character experiences a completely different "Random" experience than another.In computers, there IS NO PURE RANDOM, only random-like simulator. This isn't statistics, it's comp sci, last c++ I dealt with used the MAC address and the system timer (among other things) to simulate a random number generator, and statistically, on 1 machine, and 1 connection, its close. You say, "Well it's close", yes, but there IS ALWAYS A pattern. you may not be able to detect the pattern, ie it appears random to humans, but one exists, it must for the algorythm to function.

erin

06-08-2007, 01:27 AM

Jrral@Unrest wrote: <blockquote><cite>Liral wrote:</cite><blockquote>Some of the rep;ies in here are downright frightening in their lack of common sense. You're the types casinos love because you think that because the wheel has coem up black 5 times in a row red just HAS to come up on the next spin. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" /></blockquote>True, but on the other hand if one person was sitting at the roulette wheel and it consistently, spin after spin for hours on end, gave him twice as many blacks as reds, the casino would be shutting down that wheel and tearing it apart to find out why it wasn't working right. And they'd probably be frisking that player up to and including a cavity search to find any devices he might be using. Because the chances of the wheel showing that kind of pattern for that long purely by random chance are just too low. The whole basis for a casino is that, while the behavior of the last few events doesn't let you predict what the next one will be, the behavior of a very large number of events over time mirrors the statistical odds very very closely. That is, if I spin the wheel 5 times I may get 4 reds, but if I spin it 100,000 times I'm going to get very close to 47,368 reds, 47,368 blacks and 5264 greens (remember zero and, because American casinos are greedy, double-zero) because over the long run all the streaks and clumps and "non-random" behaviors tend to even themselves out. Every game in a casino (where you play against the house, at least) is built on that assumption, and if it weren't true casinos wouldn't still be in business. I've seen this in the EQ2 RNG myself. Two accounts owned by the same person, harvesting raws. Now, from night to night the results vary, some nights are good and some are bad. But consistently, across every character on both accounts, every zone, every tier, all hours of the day and days of the week for an entire year, one account consistently does 3-4x as well harvesting as the other. With that large a sample, the confidence that these two samples represent statistically different populations is well above the 99% mark. Which is a big red flag to me that the RNG has some statistical problems. </blockquote>You were doing great until that last paragraph. I have to call serious doubt on the idea that you have harvested a statistically significant amount.

Josgar

06-08-2007, 02:22 AM

Didn't they make it so that the loot table depends on the classes in the group so that people dont get stuck with random junk?

Themaginator

06-08-2007, 05:09 AM

how do you prove that you are right? just because you get the same thing 3 times in a row? the chances are so good for that happening in this situation its rediculous...please guys stop this stop trying to disprove the existence of a random number generator, every MMO uses this for loots every game that has loots uses it for loots, just dont do this please, this does not need to be here, and i know you think that SOE is ZOMG EVIL CORPORATION DOWN WITH THE MAN LAWL I HATE BIG BUSINESS El Oh El, and you wouldn't believe them weather they confirmed it or not you'd still think it was some sort of conspiracy, i know how you guys think, it just doesnt matter who says it, youll say its wrong just because you dont want to believe it so whats the point i ask...what the hell is the point of this thread

Twoboxer2

06-08-2007, 05:28 AM

RNG - Random Number Generator - a near oxymoron lol.Anyhow, the RNG's output has always been (perceived to be) streaky. Best way to see it perhaps is to look at the number of people who have analyzed Caustic vs Hemotoxin poison. It should be relatively easy for Hemotoxin to out-damage Caustic, but it doesn't unless artificially propped up. Because the RNG is streaky, the two poisons initial proc roughly the same number of times, but all too often the procs occur BANG BANG. So, Caustic out-damages it more often than it should.The belief the RNG is streaky goes way back. Advice has always been if you are /random 100 for a drop, get ready fast but do not roll first. If the person before your rolls a high number, roll immediately. If a low number, wait.As for getting 3 Swashie BPs . . . don't bring so many Swashies to the raid <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />And btw, can I have your loot?

callahan

06-08-2007, 07:17 AM

So many people don't realise what a dirty Hack most games are, for RNG, AI, Physics, you name it. I've worked on games where the RNG was a fixed sequence of 1000 numbers, where the total distribution was random, but if you took a 100 number chunk, it definitely wouldn't be Random. This works on principal that whats goes around comes around, not that your next Roll is truly random, like flipping a coin. Most CPUs will kick out the same sequence of Random numbers, and I always preferred using System Time as the Seed for the RNG. I've no idea what EQ2 uses, but getting 4 rare harvests from 1 node, should be statistically in the realm of winning the Lottery, and I know a lot of people who have. (I'm a 3d/technical artist, and glad I don't have to deal with this BS <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" /> )

MrWolfie

06-08-2007, 07:59 AM

<cite>callahan44 wrote:</cite><blockquote>I've no idea what EQ2 uses, but getting 4 rare harvests from 1 node, should be statistically in the realm of winning the Lottery, and I know a lot of people who have. </blockquote> Not in the least because you could only get three rare harvests from the same node <img src="/smilies/8a80c6485cd926be453217d59a84a888.gif" border="0" alt="SMILEY" />BTW OP, if Mayong is among those mobs whose loot drops were "fixed" to drop for characters present, then just don't take any Swashies next time. And if he is one of those mobs, there isn't a x24 chance to drop an item unless you're raiding with only one each of every class.And if his loot hasn't been "fixed". Then one of the replies further up has it right, just cos you got 3 of the same item doesn't mean everyone else does. Random means random beyond your experiences. If he'd been killed 100 times or more and never dropped a particular item for *anyone* on your server (or indeed, on every server) then the loot table might be broken.Perhaps if all you high-end guilds were friendly with each other, then you'd invite people from other guilds to come loot stuff that you already have. But, of course, the chances of you guys getting along with anyone else are so remote...

callahan

06-08-2007, 08:28 AM

<cite>MrWolfie wrote:</cite><blockquote>Not in the least because you could only get three rare harvests from the same node <img src="/smilies/8a80c6485cd926be453217d59a84a888.gif" border="0" alt="SMILEY" /> </blockquote>Doh, probably why I'm always harvesting a node after its despawned. <img src="/smilies/69934afc394145350659cd7add244ca9.gif" border="0" alt="SMILEY" />

Thunderthyze

06-08-2007, 09:28 AM

<cite>FightGame wrote:</cite><blockquote>Well, they say it's random, but we don't know for sure, do we? There is something wrong.</blockquote> It's not random....you've all seen the cartoon. It's a bunch of Frogloks in a room deciding on the most annoying formula for loot drops.Well, maybe not....but it's the same affect.I'm prepared to accept that drops/harvests DO fall under the umbrella of RNG statistical insignificance even when they appear to be deeply unfair. I think the point is that perhaps using an RNG is not necessarily the right way to allocate loot. Personally I would prefer to see an allocation method that takes into account previous results and group makeup. The problem is that randomness is only "fair" if applied to an infinite population. When we play EQ2, no one even begins to approach the population sample required and as such the results we see are unfair, be they streaky for the good or the bad.

Kizee

06-08-2007, 09:52 AM

<cite>Twoboxer2 wrote:</cite><blockquote>RNG - Random Number Generator - a near oxymoron lol.Anyhow, the RNG's output has always been (perceived to be) streaky. Best way to see it perhaps is to look at the number of people who have analyzed Caustic vs Hemotoxin poison. It should be relatively easy for Hemotoxin to out-damage Caustic, but it doesn't unless artificially propped up. Because the RNG is streaky, the two poisons initial proc roughly the same number of times, but all too often the procs occur BANG BANG. So, Caustic out-damages it more often than it should.The belief the RNG is streaky goes way back. Advice has always been if you are /random 100 for a drop, get ready fast but do not roll first. If the person before your rolls a high number, roll immediately. If a low number, wait.As for getting 3 Swashie BPs . . . don't bring so many Swashies to the raid <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />And btw, can I have your loot?</blockquote>Seems silly to start sitting people out of raids to "fix" a problem with their coding. I don't mind getting stuff people can use and seeing some other items drop but it seems like we get the same [Removed for Content] items every week.

Decad

06-08-2007, 03:59 PM

<cite>TuinalOfTheNexus wrote:</cite><blockquote>Decadre@Najena wrote: <blockquote>Thus as somepoint, as we get further away from the mean value, it becomes more and more unlikely that something should happen. Even almost to the point where you can safely say it's not going to happen. You have to think of a normal distribution as a bell curve that starts infinitely close to 0, shoots up to some value, and then returns infinitely close to 0.As far as roulette goes, it might be that after 10 straight black spins, you could safely say it's going to be red. I forgot the formula, but one could figure it out. Problem is, it will never be 0.Now applying this logic to the OP's problem, one could argue and figure out that getting the same item 3x over could be quite probable as it does not lie that far away from the mean value of it's associative probably range.However what might prove interesting is to see what would the normal distribution would look like for getting the same 3 items 3 times over. Figuring that one out might prove that there is something wrong.</blockquote>I think you might need to re-read the textbook <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />It's funny I think no field of maths has as much misconception as probability. I think it's probably down to human nature; the very notion of "luck" seems inherent to us.After 10 straight black spins the odds of red are - gasp - exactly the same as on the first spin. If you think about it logically - a Roulette table has no memory. It's simply throwing a ball into a wheel. You're as likely to get 1000 straight reds as you are any other combination...</blockquote>Stop right there. Once you know the probability for a single independent event, you can then calculate the probability of getting your first red on the 5th spin of the wheel. You can calculate the the probability of have 5 cocurrent independent spins resulting all in a black result. It is very possible to do, and can be backed up. We did it in class. (Edit: we did it with coin flips to check the numbers.)Not to create a flame fest.... (Edit too).Now again, IIRC a geometric distribution series would give you the probability of you getting the first black on the Nth spin of the wheel. <a href="http://en.wikipedia.org/wiki/Geometric_distribution" target="_blank" rel="nofollow">http://en.wikipedia.org/wiki/Geomet...ic_distribution</a>Now googling around now, I think we can use the Binomial Distribution to determine the chances of getting the same loot 3 times overs. <a href="http://en.wikipedia.org/wiki/Binomial_distribution" target="_blank" rel="nofollow">http://en.wikipedia.org/wiki/Binomial_distribution</a>The binomial distribution calculates the number of successes over N trials. All you need to is the number of trials, number of successes, and the probability of a success.The catch is we don't exactly know the drop rates of each item in the loot table, so we would have to assume that it's a probability of 1/24. So looking at the formula listed in the above Wiki link;n = 3k = 3p = 1/24the (1 - p)^(n-k) doesn't really matter since the exponential will be 0 in this example making the whole thing equal to 1.in fact the first part actually means n! (factorial) over (in a sense) k! (factorial). the 3! over 3! cancels out to 1 too.Therefore we get (1)(1/24)^3(1). So all we need to do really is (1/24) ^ 3.So, the odds of looting the same object 3 times in a row is .0074%Now knowing the odds of getting the same loot 3 times in a row is .0074%, we can calculate for arguments sake the chances of looting 3 seperate and independent items 3 times in a row from their respective loot tables that each item comes from (also assuming all 3 items have a 1/24% chance of dropping). The chances would be .00004%.Now my wording may be a little off here, for that I apologize. I'm trying to rush this before I have to head back to work.Edit: BTW the chances of getting a black spin is 50%, while the chance of getting 10 straight blacks is 0.1%Edit: Quick and Easy online calculator... <a href="http://stattrek.com/Tables/Binomial.aspx" target="_blank" rel="nofollow">http://stattrek.com/Tables/Binomial.aspx</a>

Jrral

06-08-2007, 04:32 PM

Erin@Blackburrow wrote: <blockquote>You were doing great until that last paragraph. I have to call serious doubt on the idea that you have harvested a statistically significant amount. </blockquote>As I said, this sample involved characters harvesting for at least an hour or so a day every day for a year (note the "at least", it could go as high as 5-6 hours). When I did some rough calculations, the most optimistic, favorable set still gives me a confidence of less than 50% that both characters represent the same population. To get to 85% confidence, the point that's my cut-off for usable results, I had to posit an active population so high that overland zones and tradeskill instances would routinely be being split into multiple instances. I don't see that kind of splitting often (the last times I saw an overland zone split was Antonica back on Errolisi Day and before that GFay just after the EoF release). I'm also minded of a known problem in a lot of Unix system RNGs. The RNGs themselves are fairly good, but they have a major problem in use: the entropy's primarily in the high bits, the low few bits exhibit a very deterministic pattern based on the seed. This interacts badly with code like "r = rand() % 20 + 1;".

Jesdyr

06-08-2007, 05:22 PM

Jrral@Unrest wrote: <blockquote>*snip* major problem in use: the entropy *snip* </blockquote> That was all you needed to say .. it is why the RNG in this game is all weird. Some EoF named MOBs use to drop in a very predicatable pattern (they were basicly scrolling though the list of classes). My guess is that the current system is using the same thing only "rounding" up to the nearest class in the group. Say the drop woudl have been for class 3 but there is no one in the group that is class 3 4 or 5 .. So the item dropped is for class 6 ... next kill would have been for class 4 but since the group has not changed you get an item for class 6... THis is all based on 2nd hand info from someone I know has a good handle on how to QA a system without actually seeing the code as well as information gathered from other people.

Korpo

06-08-2007, 05:49 PM

Keep telling yourself that the RNG is broken, and maybe someday someone (with two braincells to rub together) will believe you. Prove it to yourself though: <ul><li>Turn on logging</li><li>Make a macro to do /ran 10; ran 10; ran 10 (x50 or something)</li><li>Push that macro button a few hundred times</li><li>Run a statistical analysis on your data</li><li>Realize that the numbers ARE random and start working on your next excuse as to why you don't get what you want </li></ul>

Decad

06-08-2007, 06:25 PM

Erin@Blackburrow wrote: <blockquote>Jrral@Unrest wrote: You were doing great until that last paragraph. I have to call serious doubt on the idea that you have harvested a statistically significant amount. </blockquote> How big of a sample size do you think someone would need. Major polling companies only need a polling size of anywhere from 750 to 1000 people to fairly (within 3 - 5%) predict who's going to win the U.S. Presidential Election. And how many millions of people vote? (and how many don't, but that's an argument for another day.)

Decad

06-08-2007, 07:26 PM

<cite>Korpo wrote:</cite><blockquote>Keep telling yourself that the RNG is broken, and maybe someday someone (with two braincells to rub together) will believe you. Prove it to yourself though: <ul><li>Turn on logging</li><li>Make a macro to do /ran 10; ran 10; ran 10 (x50 or something)</li><li>Push that macro button a few hundred times</li><li>Run a statistical analysis on your data</li><li>Realize that the numbers ARE random and start working on your next excuse as to why you don't get what you want </li></ul></blockquote>Well here's a quick and easy one I just did on the Najena server in my house in Neriak, not to spam anyone.1 - 532 - 463 - 424 - 575 - 446 - 637 - 438 - 429 - 4210 - 48480 /ran 10's done in a macro 5 at a time. Needed some time to jot the numbers down on a pad. Perhaps anywhere from 3 - 5 secs.variance of 48.4 and σ = 6.95701 (to determine σ, go 1/2 down this link to the example: <a href="http://en.wikipedia.org/wiki/Standard_deviation" target="_blank" rel="nofollow">http://en.wikipedia.org/wiki/Standard_deviation</a>)Observations since I did this manually with a pen and paper.Times numbers twice in a row: 23 (of those 18 I marked what number they were. 1 - 3x, 3 - 1x, 4 - 2x, 5 - 4x, 6 - 2x, 7 - 1x, 8 - 1x, 9 - 3x, 10 - 1x) The number 4 appeared twice 3 times in a row.Another thing I noticed later on, and too later on to track. Something I might track some other time. The last 10 times a number appeared twice in a row among the 5 new numbers that appeared in the chat window, the number appeared again within one of the other 3 numbers left 7 times. (I was doing "/ran 10" 5 times in a macro.) Basically that means 7 out of 10 times, of the 5 numbers appears, one numerical value was represented 3 times. Twice in a row and at least one position away.Additionally, the numbers were EXCEPTIONALLY "streaky". At least as far as I'm concerned. You see how the 1, 4, and 6 columns are higher. I would see patterns where numbers would build up a large lead like those have, and then you wouldn't see that number again for a long while. Early on, the #3 had a large lead and by the end I don't recall even putting a mark in that column for a long time. It wasn't at all an even progression on the paper from the left to the right. I would often get outputs like this 4 - 6 - 2 - 6 - 2 OR 1 - 10 - 1 - 2 - 10. Interestingly enough, I only got one output that had a "straight" like progression involving at least 3 numbers... something like 5 - 7 - 8 - 9 - 2Anyway, they numbers do look random when put down like they are above. I think if anyone wants to prove that the RNG is broken needs to analyze how the numbers come out over time, as my little experiment in my eye tended to show that only a select few numbers come out repeatedly for a short burst. The next time I do it, I'll probably also track the number of time numbers appear in small clusters within a certain range. That number could probably be of use to someone who is a lot better at Probability Distributions, and determinating if there is in fact a problem.

Korpo

06-08-2007, 08:49 PM

I'll use your numbers since I'm not in the game right now, but for next time it's MUCH easier to clear your logs, turn on logging, then use any of the various text manipulation programs out there to automatically count the rows that contain "Bill rolls a 3!" or whatever the text is. That way you are sure that you get the count right, and you can do tests of 50,000 rolls if you want. Anyway: The null hypothesis is that there's no relation between number rolled and frequency of that number being rolled. Basically, whether every number has an equally likely shot at coming up. Here's our data, along with the chi-square value for each value. Chi-square is calculated as ((observed - expected)^2))/expected). ( Bah, I give up trying to post data in something like a table format, the total chi-square is 10.08 ) Since this is a sample with ten different possible outcomes, it's said to have nine degrees of freedom. We look up the P value in a chart such as <a href="http://www.ento.vt.edu/~sharov/PopEcol/tables/chisq.html" target="_blank" rel="nofollow">this</a> for an experiment with nine degrees of freedom and find that the critical value is 16.92. If our total chi-square is greater than the critical value, we reject the null hypothesis. Since our observed value is quite a bit less than the critical value, we have no reason to reject the hypothesis. A winner is math! If you don't want to do the math yourself, you can have Excel do it for you. Use the chitest function as described on <a href="http://www3.georgetown.edu/departments/psychology/resources/researchmethods/statistics/8495.html" target="_blank" rel="nofollow">this page</a>. As for "streakyness"... that's expected. Random numbers do not come in as 123456789012345678901234567890, that's not random. Flip a coin or roll a die or anything like that and you'll see streaks all over the place.

Fortai

06-09-2007, 01:57 AM

Go here, and select 100 numbers, between 1 and 24, allowing repeats. You will see it giving 3 of the same number in a row often. (at least it did for me). It happens. Until you know how the "RNG" in the game is designed, it's all just speculation. <a href="http://www.mdani.demon.co.uk/para/random.htm" target="_blank" rel="nofollow">http://www.mdani.demon.co.uk/para/random.htm</a>

Controlor

06-09-2007, 02:05 AM

<cite>archimidesX wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote><cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote>It definately is not random....it is very streaky.Having the same breast plate drop of Mayong 3 out of 4 times when the mob has a 24+ item loot table isn't really randomThe same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> roll a 30 sided dice a thousand times and see how much streak you see...flip a coin a thousand times see how many times in a row it lands on one side... random does not mean it won't have streaks... </blockquote>To actually elaborate on this point. Toss a coin 150 times. I believe (i am going from memory but i have the actual statistics written down somewhere) that at least 1 6-7 streak of head will appear. And in fact long streaks are expected vs smaller ones. The chance of getting your highest streak of 3 in a 150 coin toss is about 2% (if not less). And having it alternate between 1 head and 1 tail is like 0.0025% or something (again i do have the stats written down just dont wanna riffle through my probability notes atm).

Twoboxer2

06-09-2007, 05:22 AM

<cite>Kizee wrote:</cite><blockquote><cite>Twoboxer2 wrote:</cite><blockquote>RNG - Random Number Generator - a near oxymoron lol.Anyhow, the RNG's output has always been (perceived to be) streaky. Best way to see it perhaps is to look at the number of people who have analyzed Caustic vs Hemotoxin poison. It should be relatively easy for Hemotoxin to out-damage Caustic, but it doesn't unless artificially propped up. Because the RNG is streaky, the two poisons initial proc roughly the same number of times, but all too often the procs occur BANG BANG. So, Caustic out-damages it more often than it should.The belief the RNG is streaky goes way back. Advice has always been if you are /random 100 for a drop, get ready fast but do not roll first. If the person before your rolls a high number, roll immediately. If a low number, wait.As for getting 3 Swashie BPs . . . don't bring so many Swashies to the raid <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />And btw, can I have your loot?</blockquote>Seems silly to start sitting people out of raids to "fix" a problem with their coding. I don't mind getting stuff people can use and seeing some other items drop but it seems like we get the same [I cannot control my vocabulary] items every week. </blockquote>I don't disagree lol, but the "new" loot selection approach was in response to a (legitimate) whine about getting Bruiser*, Monk*, and SK* gear when there are never any at a raid <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" /> SOE then restricted loot drops to those classes present.Had they stopped there, all might have been fine (until the next whine). But they thought about it and shaped the drop-odds to match not only the occurrence of a class in the raid, but also the frequency of that class in the raid. Which is why the guess I made - "Bring less Swashies" - apparently struck a nerve with you lol.* For those with limited senses of humor, that was a joke. Sort of.

Decad

06-09-2007, 11:58 AM

Fortai@Oasis wrote: <blockquote>Go here, and select 100 numbers, between 1 and 24, allowing repeats. You will see it giving 3 of the same number in a row often. (at least it did for me). It happens. Until you know how the "RNG" in the game is designed, it's all just speculation. <a href="http://www.mdani.demon.co.uk/para/random.htm" target="_blank" rel="nofollow">http://www.mdani.demon.co.uk/para/random.htm</a> </blockquote>ANDControlor wrote: <blockquote><cite>archimidesX wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote><cite>Cusashorn wrote:</cite><blockquote>The purpose of being a RANDOM number generator is that it's completely RANDOM. To fix the generator would be giving it a predictable pattern. It's RANDOM for a reason.</blockquote>It definately is not random....it is very streaky.Having the same breast plate drop of Mayong 3 out of 4 times when the mob has a 24+ item loot table isn't really randomThe same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> roll a 30 sided dice a thousand times and see how much streak you see...flip a coin a thousand times see how many times in a row it lands on one side... random does not mean it won't have streaks... </blockquote>To actually elaborate on this point. Toss a coin 150 times. I believe (i am going from memory but i have the actual statistics written down somewhere) that at least 1 6-7 streak of head will appear. And in fact long streaks are expected vs smaller ones. The chance of getting your highest streak of 3 in a 150 coin toss is about 2% (if not less). And having it alternate between 1 head and 1 tail is like 0.0025% or something (again i do have the stats written down just dont wanna riffle through my probability notes atm). </blockquote>I know and understand streakiness. I apologize for being somewhat not upfront about it, but I think part of my amazement by it was how it reared it's head in-game when I did my little experiment.I'd say each macro hit had a 3-5 sec pause in between. Perhaps at times even more. And it was "streaky" (yeah I'm going to say it) to a point where I can see why a lot of people complain about certain group mates always getting the loot. I'm talking stretches where the numbers 2, 4, 7, 8, and 10 might appear as a whole less than 10 times during an output of 50+ numbers. I think I kind of fell into a situation where I forgot that each "/ran 10" is in itself an independent random event, so it didn't really matter than it took place 1 nanosecond apart from the other events, or 1 hour later. The same propensity for streakiness to be present is there irregardless of time passed.However, you put that in terms of 6 players in a group, and I can envision streakiness causing a lot of problems in the eyes of a lot of players. A group gets together and during that time, there are say 35 loot drops. You see where streakiness will cause problems.So, after a good nights sleep, I'm starting to wonder if what people really are saying, is that "random number generators are broken enough because of their natural propensity to be streaky, that they shouldn't be used in Games." Perhaps, that is the stand some people should be taking.Maybe we need to start looking at a situation where instead of lotto loot, we need to start having a "weighed" chance to loot. So in a full group, everyone has a 1/6th chance to win loot if everyone wants it at the start. When a group mate wins a loot, they are given a penalty which increases everytime they win something. Could be a nasty math algorithm to work out, especially since it needs to be pretty dynamic based on the ever changing size of the group looking to loot items.

Fortai

06-09-2007, 01:38 PM

AND streaks happen, giving people the impression that it isn't random when it really is. Like I said, until anyone knows how the RNG is actually designed, it's pretty much speculation at this point.

Lilj

06-10-2007, 09:11 AM

<cite>Iseabeil wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote>The same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> Just have to chime in on that one. Im not 100% certain as the old forums are gone, but quite sure that both loams/metals and meat/hides had their results tweaked sometime long ago, as loams before tinkering was in much, much lower demand then the other possible harvest just as meat was less needed then hides, and were adjusted slightly in their return bias.As for the rest of the randomness.. The streaks that often happen are annoying, but I think much of the problem comes from the perception of what random is. The 'human' view of random is of something totally unpredictable (and in most cases fair), whilst real application of RNG doesnt follow that idea. To add upon that, making algorithms that are truly random is a science in itself, and the most cutting edge versions wont be found in a game but in more serious applications, wich leaves us with somewhat solid RNG that in most cases wont live up to what most people percieve as truly random.Personally, I think RNG is slightly overused in EQ2, as it dominates too many scenes where possible alternatives could have been thought up, but not sure they could change that by now even if they wanted to.</blockquote>Isabeil, you are quite right on this one. Ore/loam is not on a 50/50 ratio, it was changed. I don't remember the numbers but I think it was more like Ore60/Loam40. This was before tinkering came into the game. The pelts/meat ratio is also changed to favor larger ratio of pelts, but I don't remember the ratio. An interesting read here about the illusion of patterns: <a href="http://en.wikipedia.org/wiki/Clustering_illusion" target="_blank" rel="nofollow">http://en.wikipedia.org/wiki/Clustering_illusion</a>

Decad

06-10-2007, 01:26 PM

<cite>Liljna wrote:</cite><blockquote><cite>Iseabeil wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote>The same thing happens when you are harvesting. you should have a 50 50 chance at loam vs. metal where after hours of harvesting I always end up with many more stacks of metal. </blockquote> Just have to chime in on that one. Im not 100% certain as the old forums are gone, but quite sure that both loams/metals and meat/hides had their results tweaked sometime long ago, as loams before tinkering was in much, much lower demand then the other possible harvest just as meat was less needed then hides, and were adjusted slightly in their return bias.As for the rest of the randomness.. The streaks that often happen are annoying, but I think much of the problem comes from the perception of what random is. The 'human' view of random is of something totally unpredictable (and in most cases fair), whilst real application of RNG doesnt follow that idea. To add upon that, making algorithms that are truly random is a science in itself, and the most cutting edge versions wont be found in a game but in more serious applications, wich leaves us with somewhat solid RNG that in most cases wont live up to what most people percieve as truly random.Personally, I think RNG is slightly overused in EQ2, as it dominates too many scenes where possible alternatives could have been thought up, but not sure they could change that by now even if they wanted to.</blockquote>Isabeil, you are quite right on this one. Ore/loam is not on a 50/50 ratio, it was changed. I don't remember the numbers but I think it was more like Ore60/Loam40. This was before tinkering came into the game. The pelts/meat ratio is also changed to favor larger ratio of pelts, but I don't remember the ratio. An interesting read here about the illusion of patterns: <a href="http://en.wikipedia.org/wiki/Clustering_illusion" target="_blank" rel="nofollow">http://en.wikipedia.org/wiki/Clustering_illusion</a> </blockquote> I liked the part about the SATs being checked to make sure runs don't exist because students will change an answer because they don't think streaks exist.

liveja

06-10-2007, 03:28 PM

<cite>Kizee wrote:</cite><blockquote>Last week in MMIS we got 2 conj gloves off 2 different mobs and this week we get 2 ranger gloves off the same 2 mobs.</blockquote> Yea, things be that way, some times.It's not a reason to claim the RNG is "broken" & in need of fixing, tho. It's just a matter of you & your guild need to suck it up, drive on, & do it again.Too bad, so sad, welcome to MMO World, enjoy your stay.

Kizee

06-10-2007, 04:19 PM

<cite>livejazz wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote>Last week in MMIS we got 2 conj gloves off 2 different mobs and this week we get 2 ranger gloves off the same 2 mobs.</blockquote> Yea, things be that way, some times.It's not a reason to claim the RNG is "broken" & in need of fixing, tho. It's just a matter of you & your guild need to suck it up, drive on, & do it again.Too bad, so sad, welcome to MMO World, enjoy your stay.</blockquote> LOL. The point I am trying to make is that it happens more than SOME times. In fact it happens that way more offen than not.

liveja

06-10-2007, 05:36 PM

<cite>Kizee wrote:</cite><blockquote><cite>livejazz wrote:</cite><blockquote><cite>Kizee wrote:</cite><blockquote>Last week in MMIS we got 2 conj gloves off 2 different mobs and this week we get 2 ranger gloves off the same 2 mobs.</blockquote> Yea, things be that way, some times.It's not a reason to claim the RNG is "broken" & in need of fixing, tho. It's just a matter of you & your guild need to suck it up, drive on, & do it again.Too bad, so sad, welcome to MMO World, enjoy your stay.</blockquote> LOL. The point I am trying to make is that it happens more than SOME times. In fact it happens that way more offen than not.</blockquote>The point I'm trying to make is that your "point" isn't evidence that anything is wrong. It's nothing but stories, which may or may not even be true. So, the best way to deal with the issue, is to simply deal with it.Or, don't. Your choice.

Jrral

06-10-2007, 09:27 PM

<cite>livejazz wrote:</cite><blockquote>The point I'm trying to make is that your "point" isn't evidence that anything is wrong. It's nothing but stories, which may or may not even be true. So, the best way to deal with the issue, is to simply deal with it. </blockquote>The only problem is, there's too many of those stories. As I noted in my first post, my experience is not a single night. It's that, over the course of nearly a year, characters on one account consistently, night after night, without exception, out-harvest those on another by a factor of 2-3 times. The same thing with crafting: crafting characters on one account are consistently getting events several times as often and have to fight to maintain durability on items, while characters on another account can breeze through to pristine with minimal effort. Now, if the RNG were OK as you claim it is, then it would not be behaving that consistently over that long a period of time. Characters would have good days and bad days, but it wouldn't consistently be the same character having the same kind of day every day. The sample's just too large, if the RNG is behaving as you say then a year's worth of sample had better be converging on the expected distribution and it isn't. As noted in the casino example, just because a roulette wheel hit black this time doesn't mean it'll hit red next time. But if, over the course of several nights play, that wheel consistently lands on red twice as often as it lands on black, what are the odds that wheel's actually good and this just happened by chance?

Decad

06-10-2007, 09:53 PM

Jrral@Unrest wrote: <blockquote><cite>livejazz wrote:</cite><blockquote>The point I'm trying to make is that your "point" isn't evidence that anything is wrong. It's nothing but stories, which may or may not even be true. So, the best way to deal with the issue, is to simply deal with it. </blockquote>The only problem is, there's too many of those stories. As I noted in my first post, my experience is not a single night. It's that, over the course of nearly a year, characters on one account consistently, night after night, without exception, out-harvest those on another by a factor of 2-3 times. The same thing with crafting: crafting characters on one account are consistently getting events several times as often and have to fight to maintain durability on items, while characters on another account can breeze through to pristine with minimal effort. Now, if the RNG were OK as you claim it is, then it would not be behaving that consistently over that long a period of time. Characters would have good days and bad days, but it wouldn't consistently be the same character having the same kind of day every day. The sample's just too large, if the RNG is behaving as you say then a year's worth of sample had better be converging on the expected distribution and it isn't. As noted in the casino example, just because a roulette wheel hit black this time doesn't mean it'll hit red next time. But if, over the course of several nights play, that wheel consistently lands on red twice as often as it lands on black, what are the odds that wheel's actually good and this just happened by chance? </blockquote>Well, I may be mistaken on this since I'm not exactly taking the time to think it out all that much, but... all those events are independent of one another.In an attempt to explain it better, take for example the throw the die 500 times and record the result experiment that has been discussed. It really doesn't matter if you do all 500 throws in one night within an hours time, OR if you do a throw once a night for 500 days on the same table and use the same dice to play other games with during that period. Each throw you do has the same probability results since the person throwing the dice doesn't change, the dice doesn't change... essentially as long as any variable that might exist in the act of throwing the dice doesn't change, you get the same sort of random collection of numbers with the same streaks, same "perceived" patterns, etc.Now this is assuming that everyone (all accounts) is on the same level playing field. Problem is, we don't know how the random numbers are generated. A story that I like to bring up from time to time, and if we had access to the old forums I could try to find a link to, relates to this.My first programming class in college, I had a lab project to do a game that needed a random number generator. Started the lab at school and got that backend working. For the purposes of this story, essentially I had a functioning as needed random number generator in my project. Sometime later while at home (you know, near the deadline), I'm working on the project and I was having problems. So I start debugging it, and I notice that everytime I run it, the random numbers I get are EXACTLY the same. I would get like 1, 7, 3, 5, 9, .... every single time I ran it. So I take it to school, and it works fine. Go to a friends house who had Visual Studio, and it works fine there. Make a long story short, professor is baffled and I end up lugging in my PC to hook up at school because he doesn't believe me, and sure enough it spits out 1, 7, 3, 5, 9... Never figured it out. It was seeded off of the system clock, but the clock worked fine.So, if someone finds out that random numbers are generated off of something on the clients end, theoretically speaking there could be an issue, as I've been a witness too. If it's on the server end, you my friend I'm sorry to say will have a tough time proving mathematically in terms of probability, as well as programming wise since everyone's random numbers are coming from the same place, that there's a legimate problem.You could try arguing that because of streakiness, that using a random number generator causes too wierd of a problem... Something I think might be the perceived problem that even I fall into.

thebunny

06-10-2007, 10:05 PM

Jrral@Unrest wrote: <blockquote>As noted in the casino example, just because a roulette wheel hit black this time doesn't mean it'll hit red next time. But if, over the course of several nights play, that wheel consistently lands on red twice as often as it lands on black, what are the odds that wheel's actually good and this just happened by chance? </blockquote>I think the problem is that there's more than one wheel here. If your guild was the only one using the RNG over a period of a month, and you still witnessed the same behavior, then yes I would blame the RNG. But who knows exactly what events are using the RNG at the same time that the "streak" is happening. In other words, while you might perceive seeing 6 twice in a row, it may actually go something like 6, 2, 9, 5, 6, where the middle three events are completely separate from your events. Imagine if the same RNG is used to determine every single loot drop in the game - that's a ton of events in between your two "streak" events.Like others have said in this thread, I think it would make more sense to argue for putting some rules on loot drops (like the recent class specific change) than to assume that the RNG is broken and needs to be fixed.

thebunny

06-10-2007, 10:06 PM

Decadre@Najena wrote: <blockquote>So, if someone finds out that random numbers are generated off of something on the clients end, theoretically speaking there could be an issue, as I've been a witness too. If it's on the server end, you my friend I'm sorry to say will have a tough time proving mathematically in terms of probability, as well as programming wise since everyone's random numbers are coming from the same place, that there's a legimate problem. </blockquote>It would really surprise me if random numbers were being generated on the client. For security reasons, they would almost have to be generated at the server.

StormCinder

06-10-2007, 10:17 PM

Fortai@Oasis wrote: <blockquote>AND streaks happen, giving people the impression that it isn't random when it really is. Like I said, until anyone knows how the RNG is actually designed, it's pretty much speculation at this point. </blockquote> AMEN!SC

Decad

06-10-2007, 11:31 PM

<cite>thebunny wrote:</cite><blockquote>Decadre@Najena wrote: <blockquote>So, if someone finds out that random numbers are generated off of something on the clients end, theoretically speaking there could be an issue, as I've been a witness too. If it's on the server end, you my friend I'm sorry to say will have a tough time proving mathematically in terms of probability, as well as programming wise since everyone's random numbers are coming from the same place, that there's a legimate problem. </blockquote>It would really surprise me if random numbers were being generated on the client. For security reasons, they would almost have to be generated at the server.</blockquote> In a sense, I wholeheartily agree with you.But's theres also a part of me that doesn't know a whole lot about networking and the processing power of servers. This part of me wonders there is probably a random number generated everytime an attack is made, a spell is cast to hit as well as it's subsequent resist check, every time a node is harvested from, someone falls too far, a skill check is made, a crafting success check is made, etc. Multiple all those events by everyone on the server, plus all the mobs that those people might be involved in fight with, factor in all the database calls that probably occurs whenever someone updates a quest, or simply opens their inventory. Then factor in all the script calls that have to be made for quest events, NPC hail events, etc. This part of me wonders if the server can handle all of that? If not, or even if so, did the developers make a concession and seed the random number generator on the client end?Nah, that be dumb. Besides there is still going to be CPU time needed to interpret the data anyway whether it's from the client end or server end, and how much extra CPU power can a silly random number generator use up?but...........................

Bromir

06-11-2007, 02:23 AM

Could be interesting to see more of the official math on this .. YoursBrorim

Noaani

06-11-2007, 09:38 AM

Siclone wrote: <blockquote>I wonder if you even understandthe topic,,,,,getting 5 heads on a row on a 50/50 chance is not what we are talking about. there are 24 classes, and what 12 solts ,,,I mean the random chance of getting 3 swashie breast plates in a row,,,,,,those odds are amazing, and its happening to allot of groups/guilds</blockquote>Statistics 101.The odds of getting a given classes BP from Mayong on any given kill = 1/12 (24 classes, 2 drop per kill, is code in place to avoid 2 of the same dropping for 'most' situations).Odds of 2 dropping after 2 kills = 1/144Odds of 3 dropping after 3 kills = 1/1728Statistically, at some point, that has to happen. It would not be random if it did not, but rather a predictable pattern. Events with a 1/1,000,000 chance of happening still happen, so 1/1728 is not indicative of a broken mechanic.Hell, if you got a 4th one in a row, thats still only 1/20736, so you still have a long way to go.On a side note, should you 'not' get that same drop on your next kill, your odds would fall back in to what would be conisdered a 'normal' range, not odd, not broken, but normal. Wow, it must be broken.

redde

06-11-2007, 09:46 AM

Wow, 6 pages of posts on this??! The RANDOM generator generates a number RANDOMLY! The chance of class armor dropping from the EoF set is 1/(number of different classes in the raid). Say you have every class (unlikely): The chance of the breastplate dropping for any class is 1/24. Next time you come in, you have 24 classes again, the chance of the breastplate dopping for any class is 1/24. Every time you go in there will be a 1/24 chance of each thing dropping, it's completely independant of what you got last time. Don't moan because you got the same thing three times in a row. I'd rather not have a lower chance of getting my breastplate because last time someone did the raid, mine dropped & someone else won it.

Chefren

06-11-2007, 10:15 AM

Decadre@Najena wrote: <blockquote>thebunny wrote: But's theres also a part of me that doesn't know a whole lot about networking and the processing power of servers. This part of me wonders there is probably a random number generated everytime an attack is made, a spell is cast to hit as well as it's subsequent resist check, every time a node is harvested from, someone falls too far, a skill check is made, a crafting success check is made, etc. Multiple all those events by everyone on the server, plus all the mobs that those people might be involved in fight with, factor in all the database calls that probably occurs whenever someone updates a quest, or simply opens their inventory. Then factor in all the script calls that have to be made for quest events, NPC hail events, etc. This part of me wonders if the server can handle all of that? If not, or even if so, did the developers make a concession and seed the random number generator on the client end?Nah, that be dumb. Besides there is still going to be CPU time needed to interpret the data anyway whether it's from the client end or server end, and how much extra CPU power can a silly random number generator use up?but...........................</blockquote>Random numbers have to be generated on the server or someone from the dev team and their boss needs to be fired. Syncing random numbers from the clients to the server and back to whatever clients happen to need to know the result would be both inefficient and a nightmare to debug. That would not be a concession to anything except stupidity. Generating random numbers even in huge amounts is still not very demanding compared to for example saving the game state to disk (I would imagine).

liveja

06-11-2007, 10:42 AM

Jrral@Unrest wrote: <blockquote>The only problem is, there's too many of those stories. </blockquote>This is the Internet. Anecdotal stories, no matter how frequently they're told, are the very worst sort of "evidence". They are inherently not to be trusted. IMHO, anecdotes are as likely to be outright lies told by trolls, as they are to be anything else.I'm going to go out on a limb & say there isn't one person posting in this thread (including myself) that has the faintest idea what we're really talking about. That's not an atmosphere in which SOE should do anything at all.

Jrral

06-11-2007, 10:13 PM

<cite>thebunny wrote:</cite><blockquote>I think the problem is that there's more than one wheel here. If your guild was the only one using the RNG over a period of a month, and you still witnessed the same behavior, then yes I would blame the RNG. But who knows exactly what events are using the RNG at the same time that the "streak" is happening. In other words, while you might perceive seeing 6 twice in a row, it may actually go something like 6, 2, 9, 5, 6, where the middle three events are completely separate from your events. Imagine if the same RNG is used to determine every single loot drop in the game - that's a ton of events in between your two "streak" events.</blockquote>If that's the case then the RNG is broken. One property of a well-designed RNG is that any subset of it's output is as random as the entire output. In short, if I take 500,000 spins of a roulette wheel and choose 50,000 of them, about half the set I choose should be each color. It shouldn't matter if I choose randomly, choose every other spin out of the first 100,000, choose the last 50,000 or choose every tenth spin of the whole set, all those samples should (assuming they're large enough in and of themselves) converge to the expected distribution. If they aren't, you've uncovered non-randomness in the underlying data: either the output itself is non-random or individual outputs aren't statistically independent of each other (in which case all the math everyone here's been using to "prove" the RNG must be OK goes right out the window because those statistical tests assume independence). Note that it may not be the RNG itself that's broken, it may be a combination of the RNG and the way it's output is used. See, for example, the Unix system rand() functions where the entire 32-bit output is good but the low few bits are highly non-random. Also note that the most common RNG algorithms have a 9-shaped output sequence: an initial tail that's a unique non-repeating sequence, merging eventually into a repeating cycle. This is why cryptographic protocols (which are highly sensitive to the goodness of the random numbers they use) always re-seed their RNGs periodically.

thebunny

06-12-2007, 10:32 AM

Jrral@Unrest wrote: <blockquote>If that's the case then the RNG is broken. One property of a well-designed RNG is that any subset of it's output is as random as the entire output. In short, if I take 500,000 spins of a roulette wheel and choose 50,000 of them, about half the set I choose should be each color. It shouldn't matter if I choose randomly, choose every other spin out of the first 100,000, choose the last 50,000 or choose every tenth spin of the whole set, all those samples should (assuming they're large enough in and of themselves) converge to the expected distribution. If they aren't, you've uncovered non-randomness in the underlying data: either the output itself is non-random or individual outputs aren't statistically independent of each other (in which case all the math everyone here's been using to "prove" the RNG must be OK goes right out the window because those statistical tests assume independence). </blockquote>So what you're saying is that regardless of what subset of numbers you pick out of the result set, that subset should be just as random as the whole result set? That's physically impossible; unless every single number in the result set is different, there is no way to ensure that every subset of the result set is "random". Randomness does not mean that a result set has an equal number of every possible output, rather that every possible output has an equal probability of being selected without any discernable pattern. Sure, if your result set size is infinite, then all outputs will be equally represented. But we're talking about fairly small result sets here.My point was that observing streaks does not necessarily imply that the RNG is broken. For one thing, those streaks may not actually be streaks at all, rather simply picking outputs from a much larger result set. For another, a random number set is not void of streaks, and if the generator is fixed to avoid streaks then it violates one of the rules of randomness.

Korpo

06-12-2007, 12:52 PM

<cite>livejazz wrote:</cite><blockquote>Jrral@Unrest wrote: I'm going to go out on a limb & say there isn't one person posting in this thread (including myself) that has the faintest idea what we're really talking about. That's not an atmosphere in which SOE should do anything at all.</blockquote>I know how to do a chi-square on a bunch of numbers. It doesn't matter though, just like every single thread posted in the last two years on the subject of "OMGZ TEH RNG IS BORKEDERR!", people will continue pushing forward on the assumption that it's broken, despite overwhelming evidence that it's not. People believe what they want to believe, especially stupid people.

Decad

06-12-2007, 02:45 PM

<cite>livejazz wrote:</cite><blockquote>I'm going to go out on a limb & say there isn't one person posting in this thread (including myself) that has the faintest idea what we're really talking about. That's not an atmosphere in which SOE should do anything at all.</blockquote>Well, I have a bachelors in Computer Science, and had to sit through 2 semesters of Probability classes....but even my admission that probably means I don't really know anything

Valdaglerion

06-12-2007, 03:26 PM

Its random but they recently changed it to only drop for the classes within the group/raid, correct?This would mean that if you have conjs/swash/fury as the makeup of the group the RNG is only considering those classes making it a 33.3% chance of dropping an item for that class from the mobs loot table.At least that is my understanding at this point.

Wossname

06-12-2007, 05:26 PM

Having read the thread there seem to be a number of entangled issues that we, as players, can't necessarily see as separate. At the most basic level this can be split into two parts: 1. The RNG (more correctly the pseudo-Random Number Generator) 2. What happens to the random numbers. The quality of the random number generator is partly in the distribution of numbers produced i.e. it should be unbiased and partly in the unpredictability of the output. A simple ramp producing sequential integers is unbiased but completely predictable. Other generators, depending on implementation, have varying properties of bias and predictability. The Unix rand() function is known to be very poor but cryptographic grade pRNG's that periodically re-seed using data from interrupt timing etc are good enough to secure financial transactions. What is done with the pRNG output matters as well. Consider the following trivial example, where random() is a perfect RNG which outputs floating point numbers between 0 and 1: result = round(random()*5)+1 This will produce a set of numbers between 1 and 6, you might consider this ideal for distributing loot in a group. It's not, 1 and 6 have half the probability of coming up that the rest do: 0.0 - 0.5 => 1 0.51 - 1.5 => 2 1.51 - 2.5 => 3 2.51 - 3.5 => 4 3.51 - 4.5 => 5 4.51 - 5.0 => 6 Of course, this is a deliberately broken example (frighteningly it comes up in real code sometimes) but it does show that even a perfect RNG can be wrecked by poor usage of the output. Back to EQ2, which problem do we have? I don't know, I'd guess both. True RNG's exhibit streaks but aren't predictable. pRNG's can exhibit streaks too but they can also become predictable. Bad pRNG's are completely predictable sequences with a long repetition period (<a href="http://random.mat.sbg.ac.at/~charly/server/node3.html" target="_blank" rel="nofollow">Linear Congruential Generator</a> note the references to Unix rand()). If the seed isn't completely independent of the character/mob then it is too easy to predict the anecdotal reports of certain characters doing far better for harvesting. Problems like Mayong Mistmoore's drop predictability is probably more down to the usage of the random numbers, especially given the recent changes to only drop loot that someone can in the raid can use.

Valdaglerion

06-12-2007, 06:01 PM

<cite>Wossname wrote:</cite><blockquote>1. The RNG (more correctly the pseudo-Random Number Generator) [SNIP] What is done with the pRNG output matters as well. Consider the following trivial example, where random() is a perfect RNG which outputs floating point numbers between 0 and 1: result = round(random()*5)+1 This will produce a set of numbers between 1 and 6, you might consider this ideal for distributing loot in a group. It's not, 1 and 6 have half the probability of coming up that the rest do: 0.0 - 0.5 => 1 0.51 - 1.5 => 2 1.51 - 2.5 => 3 2.51 - 3.5 => 4 3.51 - 4.5 => 5 4.51 - 5.0 => 6 </blockquote> Not sure why you would use a rounding function. When designating a whole number it would be better to designate the variable as a whole integer and use the RND to do this -Int((6 - 1 + 1) * Rnd + 1)This produces a random number between 1 and 6 which could be used for distributing loot. I ran this sequence 5000 times to benchmark a random selection analysis. Here are the results, left side is the number chosen and the right side denotes how many times the number was chosen out of the 500 rolls.<ul><li>1 -859</li><li>2- 864</li><li>3- 817</li><li>4- 847</li><li>5- 819</li><li>6- 794</li></ul>Dont know why you consider that "psuedo". The results seem fairly balanced and no discernible patterns in the available selections. Sure there are times when it is more probable that numbrs are going to be selected more times subsequently when using only 6 numbers, its the nature of RNG programs. Use the same sequences to pick numbers between 1 and 15,000,000 and see what happens. That is the same logic used in lotteries and keno games (where the casinos cant afford to have discernible patterns).I dont see an imbalance in the RNG being used by SOE, its simply luck of the draw. As in any application where a RNG is present the more you play the more balanced you will see the results to be. For casual players it is entirely possible to have a bad night where you arent lucky with the RNG and the same thing happen the next time you play. This is not as evident with more "hard core" players because they already know that it all breaks out in the long run.With that said and done we do have a Fae in our guild that is extremely lucky, I swear she is using loaded dice <img src="/smilies/e8a506dc4ad763aca51bec4ca7dc8560.gif" border="0" alt="SMILEY" />

Jrral

06-12-2007, 11:35 PM

<cite>Valdaglerion wrote:</cite><blockquote>Dont know why you consider that "psuedo". The results seem fairly balanced and no discernible patterns in the available selections. </blockquote>It's called "pseudo" because it is. To a human it looks random, but in reality it's a completely 100% deterministic sequence. It always produces exactly the same output for any given initial seed. This is a property of any computer algorithm, because computers are deterministic machines. You can get hardware that generates true random numbers, but it's expensive and involves radiation sources. A PRNG is still useful, though, because of two properties: it's output has statistical properties that are "close enough" to those of a truly random sequence, and given only the output it's infeasible to determine what seed generated it. As it turns out, for a lot of purposes you don't need the output to be random, merely unpredictable and evenly distributed. I'm planning on doing a more accurate test. 2 characters. Character A is on my poor-performing account but is level 57 with skills maxed for his level. Character B is on the better-performing account but will only be leveled up far enough to get skills to 90 (the T2-T3 boundary). Both will harvest the same area of Antonica at the same time for the same length of time each night for 2 weeks. I'll alternate characters each day, so variations due to day of the week and long-term trends over the whole 2 weeks should balance out. I'll keep track of number of nodes of each type harvested and try to keep it as close as possible between the 2 characters. Null hypothesis will be that the harvest results of the two characters (the samples) represent the same population. Accept or reject, I'll still make some coin on the broker. <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />

Wossname

06-13-2007, 04:58 PM

<cite>Valdaglerion wrote:</cite><blockquote><cite>Wossname wrote:</cite><blockquote>1. The RNG (more correctly the pseudo-Random Number Generator) [SNIP] What is done with the pRNG output matters as well. Consider the following trivial example, where random() is a perfect RNG which outputs floating point numbers between 0 and 1: result = round(random()*5)+1 This will produce a set of numbers between 1 and 6, you might consider this ideal for distributing loot in a group. It's not, 1 and 6 have half the probability of coming up that the rest do: 0.0 - 0.5 => 1 0.51 - 1.5 => 2 1.51 - 2.5 => 3 2.51 - 3.5 => 4 3.51 - 4.5 => 5 4.51 - 5.0 => 6 </blockquote> Not sure why you would use a rounding function. When designating a whole number it would be better to designate the variable as a whole integer and use the RND to do this -Int((6 - 1 + 1) * Rnd + 1)This produces a random number between 1 and 6 which could be used for distributing loot. I ran this sequence 5000 times to benchmark a random selection analysis. Here are the results, left side is the number chosen and the right side denotes how many times the number was chosen out of the 500 rolls.<ul><li>1 -859</li><li>2- 864</li><li>3- 817</li><li>4- 847</li><li>5- 819</li><li>6- 794</li></ul>Dont know why you consider that "psuedo". The results seem fairly balanced and no discernible patterns in the available selections. Sure there are times when it is more probable that numbrs are going to be selected more times subsequently when using only 6 numbers, its the nature of RNG programs. Use the same sequences to pick numbers between 1 and 15,000,000 and see what happens. That is the same logic used in lotteries and keno games (where the casinos cant afford to have discernible patterns).I dont see an imbalance in the RNG being used by SOE, its simply luck of the draw. As in any application where a RNG is present the more you play the more balanced you will see the results to be. For casual players it is entirely possible to have a bad night where you arent lucky with the RNG and the same thing happen the next time you play. This is not as evident with more "hard core" players because they already know that it all breaks out in the long run.With that said and done we do have a Fae in our guild that is extremely lucky, I swear she is using loaded dice <img src="/eq2/images/smilies/e8a506dc4ad763aca51bec4ca7dc8560.gif" border="0"></blockquote> Jrral has kindly explained why it is pseudo-random. You seem to have missed the part of my post where I specifically mention it is a broken example to prove a point: even a perfect entropy source such as thermal noise can seem to produce biased end results if the output of the RNG is used badly. The round() function as used in my example is a bad idea. The test you performed tests one of the two aspects of a pRNG: bias. The numbers you present, and your commentary, tell us your sample is reasonably unbiased. However, they tell us precisely nothing about the predictability of the sequence in the sample. As has been mentioned previously there are <a href="http://en.wikipedia.org/wiki/Diehard_tests" target="_blank" rel="nofollow">proper statistical tests</a> (I hate citing wikipedia but it is a simple explanation) which do a far better job of evaluating randomness than a human looking at a sequence. Streaks can happen in a random sequence but they must not be predictable. The OP's problem is apparent predictability, I'm just trying to point out that there are three possible aspects to consider when looking at the pRNG problem: bias, predictability and usage.

Hoark

06-13-2007, 06:31 PM

Whether there is a RNG, a PRNG, or some other algorythim designed to give the illusion of a randomly generated piece of loot is irrelevant. The real issue that the OP is raising is: When we the players kill a tough mob, we want to be rewarded, and when the mob drops a piece of loot that rots or gets transmuted, it feels like our effort was wasted. Possible solution: My proposal is to allow the raid leader to choose a rewad fom the loot table, much like we can choose rewards for some quest completions. A window would pop up, the raid leader would pick a reward from that mobs loot table, and then the standard loot window would pop up with the selected loot.Considering lockout timers, a raid boss can be killed a maximunm of once evey 5days 20hours. Let's say for the sake of argument that Raid_Guild is a hardcore raiding guild. They only have 24 members total to gear up, and they are good enough to kill Raid_Boss once a week, every week.Even with my proposed system it will take them four months of raiding to gear up, assuming they all show up to every raid and take out Raid_Boss every time. IMO, after four months of dedicated raiding every member SHOULD be able to have that uber item they are after. With the way loot is done now (luck of the draw), it will take far longer for that hardcore force to gear up and God help the casual raiding guilds out there!

Valdaglerion

06-13-2007, 06:41 PM

Jrral@Unrest wrote: <blockquote>It's called "pseudo" because it is. To a human it looks random, but in reality it's a completely 100% deterministic sequence. It always produces exactly the same output for any given initial seed. This is a property of any computer algorithm, because computers are deterministic machines. You can get hardware that generates true random numbers, but it's expensive and involves radiation sources. A PRNG is still useful, though, because of two properties: it's output has statistical properties that are "close enough" to those of a truly random sequence, and given only the output it's infeasible to determine what seed generated it. As it turns out, for a lot of purposes you don't need the output to be random, merely unpredictable and evenly distributed. Ok, I will give you this one because I made the assumption they are seeding with a chaos algorithm. Seeding from something as simple as the client system clock would be a joke. Although, now I am curious as well as to what they are seeding the RNG with. I have noticed consistent luck between some characters but it has been stated in other places on the forums that having a higher skill level than the tier you are harvesting in affects the RNG for rares, etc. Seems there are multiple variables which affect the output which are to be expected with percentage advantages but the biggest question I think here is really the seeding process of the RNG.On a side note - this is actually an interesting thread... I'm planning on doing a more accurate test. 2 characters. Character A is on my poor-performing account but is level 57 with skills maxed for his level. Character B is on the better-performing account but will only be leveled up far enough to get skills to 90 (the T2-T3 boundary). Both will harvest the same area of Antonica at the same time for the same length of time each night for 2 weeks. I'll alternate characters each day, so variations due to day of the week and long-term trends over the whole 2 weeks should balance out. I'll keep track of number of nodes of each type harvested and try to keep it as close as possible between the 2 characters. Null hypothesis will be that the harvest results of the two characters (the samples) represent the same population. Accept or reject, I'll still make some coin on the broker. <img src="/smilies/3b63d1616c5dfcf29f8a7a031aaa7cad.gif" border="0" alt="SMILEY" />Not sure how well this will be due to the reasons mentioned above concerning skill level, etc. Will be interesting none the less.</blockquote>

Valdaglerion

06-13-2007, 06:47 PM

Hoark@Oasis wrote: <blockquote>Possible solution: My proposal is to allow the raid leader to choose a rewad fom the loot table, much like we can choose rewards for some quest completions. A window would pop up, the raid leader would pick a reward from that mobs loot table, and then the standard loot window would pop up with the selected loot.</blockquote> Not a bad suggestion at all. But all mobs have multiple types of loot within their table. Currently you have a RNG roll to determine what type of loot from the table you have won (Legendary, Fabeled, Treasured, etc), another to determine the number of item(s) you have won if applicable and another to determine the specific items from that type set you are awarded. Even with a selection window you would still be subject to the RNG for the initial determinations which brings you back to the RNG issue.

Hoark

06-13-2007, 08:15 PM

<cite>Valdaglerion wrote:</cite><blockquote>Hoark@Oasis wrote: <blockquote>Possible solution: My proposal is to allow the raid leader to choose a rewad fom the loot table, much like we can choose rewards for some quest completions. A window would pop up, the raid leader would pick a reward from that mobs loot table, and then the standard loot window would pop up with the selected loot.</blockquote> Not a bad suggestion at all. But all mobs have multiple types of loot within their table. Currently you have a RNG roll to determine what type of loot from the table you have won (Legendary, Fabeled, Treasured, etc), another to determine the number of item(s) you have won if applicable and another to determine the specific items from that type set you are awarded. Even with a selection window you would still be subject to the RNG for the initial determinations which brings you back to the RNG issue.</blockquote>Is that true of all mobs though? I could be mistaken, but don't some EoF raid mobs only drop class specific set gear? Either way, I could live with the RNG determining which loot table comes up, so long as we could pick the reward from that table. This would eliminate the problem that the OP brought up and ensure that the effort it takes to bring down that nasty raid mob is rewarded.

Jrral

06-13-2007, 10:42 PM

<cite>Valdaglerion wrote:</cite><blockquote><blockquote>Not sure how well this will be due to the reasons mentioned above concerning skill level, etc. Will be interesting none the less.</blockquote> </blockquote>Well, one of the ideas is that I've tried to even out the environmental factors as much as possible, so the only difference is the characters themselves. Then I stacked the deck as much as possible in favor of the account my experience says performs poorly compared to what I'd expect. That character is higher level, has higher skills (which the devs have said should result in better harvest results and more rares) and is well above the cap for the tier he's harvesting in (which again should result in more rares). If, given all that, he still can't out-perform my second account, where the character will have skills maxed but just barely at the point they could move to the next tier, over the time period involved then IMO "luck of the dice" just doesn't explain it. Anecdotal: I've run about 8 characters through Queen's Colony on my first account. I've harvested a grand total of 3 rares total between all of them on the Isle getting all their skills up (none of them were less than level 7 before leaving), all after their skills were maxed. Compare this to the first character I rolled on the second account, who collected 8 rares (not imbues, true rares) in 40 minutes of harvesting with skills not yet maxed! It could be coincidence, that's too short a time-frame for statistical conclusions, but I find it curious that in a year on my first account not one of my characters has ever had a run of good luck like that. Not once. Not even close. That's the kind of outlier that raises my eyebrows and makes me re-check my data to make sure it's really an outlier.

Valdaglerion

06-14-2007, 03:33 PM

Jrral@Unrest wrote: <blockquote><cite>Valdaglerion wrote:</cite><blockquote><blockquote>Not sure how well this will be due to the reasons mentioned above concerning skill level, etc. Will be interesting none the less.</blockquote> </blockquote>Well, one of the ideas is that I've tried to even out the environmental factors as much as possible, so the only difference is the characters themselves. Then I stacked the deck as much as possible in favor of the account my experience says performs poorly compared to what I'd expect. That character is higher level, has higher skills (which the devs have said should result in better harvest results and more rares) and is well above the cap for the tier he's harvesting in (which again should result in more rares). If, given all that, he still can't out-perform my second account, where the character will have skills maxed but just barely at the point they could move to the next tier, over the time period involved then IMO "luck of the dice" just doesn't explain it. Anecdotal: I've run about 8 characters through Queen's Colony on my first account. I've harvested a grand total of 3 rares total between all of them on the Isle getting all their skills up (none of them were less than level 7 before leaving), all after their skills were maxed. Compare this to the first character I rolled on the second account, who collected 8 rares (not imbues, true rares) in 40 minutes of harvesting with skills not yet maxed! It could be coincidence, that's too short a time-frame for statistical conclusions, but I find it curious that in a year on my first account not one of my characters has ever had a run of good luck like that. Not once. Not even close. That's the kind of outlier that raises my eyebrows and makes me re-check my data to make sure it's really an outlier. </blockquote>I dont know what more to say. I know that on my primary toon, I have vastly different experiences in harvesting even one the same toon. This may be a totally off the wall idea but something to consider as well since we dont actually know how the RNG is coded.Consider for a moment -Mob A has a loot table that is loaded into a server variable array. The RNG is used to deplete items from that loot table until a predetermined event occurs (certain number of items are won or the UBER item is won) and then the variable array is reset.If this were the case it could explain a lot. The more players working a specific area would be depleting the tables more often and resetting them which would subject you to a more random selection because your percentages are not only based on your performance but the performance of others which is completely unknown to you.Yes, this may be crazy I know but after thinking about it, it seems semi plausible. I know I always have better luck at off times when there are few other people in an area and I run the area multiple times by myself. It could also explain why people have reported instanced harvesting zones as better places to find rares (you are the only person being considered to deplete those loot tables), same with adventuring zones that are instanced. Oh well, just another thought.

Jrral

06-14-2007, 10:42 PM

<cite>Valdaglerion wrote:</cite><blockquote>If this were the case it could explain a lot. The more players working a specific area would be depleting the tables more often and resetting them which would subject you to a more random selection because your percentages are not only based on your performance but the performance of others which is completely unknown to you.</blockquote>That's one reason I'm doing the harvesting at the same time each night, and for long enough for each character to harvest on each day of the week: so that variations like zone population on weeknights vs. weekends evens out between them. But one thing I notice is that the results run the opposite from what you suggest. The more people are active in the zone, the worse my results seem to be. Others I've talked to have said the same thing without me mentioning what my observations were. It acts as if there's a zone-wide rate cap on some things, and as activity picks up the individual drops are scaled back in an attempt to keep the total rate in the zone from exceeding that cap. Interestingly, this would make for an effective if crude anti-farming measure.

Kizee

06-15-2007, 12:56 AM

Seriously [Removed for Content].Got our 4th swash BP tonight with our 3rd grim earing in 5 times.....and another set of conj gloves.Yeah...the RNG is working real good. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" />I am so sick of seeing the same items over and over and over again every week.F I X I T ! ! !

Themaginator

06-15-2007, 02:55 AM

<cite>Kizee wrote:</cite><blockquote>Seriously [Removed for Content].Got our 4th swash BP tonight with our 3rd grim earing in 5 times.....and another set of conj gloves.Yeah...the RNG is working real good. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" />I am so sick of seeing the same items over and over and over again every week.F I X I T ! ! !</blockquote> yeah...fix random probability...rewrite the laws of math and the universe SOE...Durrrr

kenm

06-15-2007, 03:27 AM

Sashtan@Mistmoore wrote: <blockquote><cite>Kizee wrote:</cite><blockquote>Seriously [Removed for Content].Got our 4th swash BP tonight with our 3rd grim earing in 5 times.....and another set of conj gloves.Yeah...the RNG is working real good. <img src="/smilies/2786c5c8e1a8be796fb2f726cca5a0fe.gif" border="0" alt="SMILEY" />I am so sick of seeing the same items over and over and over again every week.F I X I T ! ! !</blockquote> yeah...fix random probability...rewrite the laws of math and the universe SOE...Durrrr </blockquote>There is no freaking way loot distribution is truly random. It flat out HAS to be seeded from something with a very high chance of being non-random, something dumb like the first person to zone in. We get necro and guardian gloves up the [Removed for Content], we've never seen a Berserker or Guardian breastplate... yet we've gotten 3 Coercer, 2 Necro, 2 Troub. One week we actually got an identical chest as the last one minus I think only the master. (I think it was Troub, Coercer, Ring, Brell Altar... two weeks in a row. What the heck are the odds of that?) So it's technically possible for something to be random and have that amount of stacked similar loot? Wow we must have bad luck. So does every one of my friends guilds. Be it countless berserker helms in a row (I still don't have one, so it's clearly different streaks for every guild) or 10 animist tunics in a row, everybody always has these totally freaking stupid streaks where we're forced to transmute 75% of the loot we get. Raiding zones for nothing but transmute fodder is just dumb, as is having to sit out people so they don't cause the mob to drop loot they already have.

mamasan

06-15-2007, 09:02 AM

The coin tossing and there being streaks, well, that's quite simple, theres only 2 sides to a coin so its a 50-50. But when theres 24 items and you get the same 3 in a row, thats not a 50-50, now is it? Thats a (1 in 24) x (1 in 24) x (1 in 24). Whats that? 1 in a million? But seems to happen constantly.

Valdaglerion

06-15-2007, 01:01 PM

Jrral@Unrest wrote: <blockquote><cite>Valdaglerion wrote:</cite><blockquote>If this were the case it could explain a lot. The more players working a specific area would be depleting the tables more often and resetting them which would subject you to a more random selection because your percentages are not only based on your performance but the performance of others which is completely unknown to you.</blockquote>That's one reason I'm doing the harvesting at the same time each night, and for long enough for each character to harvest on each day of the week: so that variations like zone population on weeknights vs. weekends evens out between them. But one thing I notice is that the results run the opposite from what you suggest. The more people are active in the zone, the worse my results seem to be. Others I've talked to have said the same thing without me mentioning what my observations were. It acts as if there's a zone-wide rate cap on some things, and as activity picks up the individual drops are scaled back in an attempt to keep the total rate in the zone from exceeding that cap. Interestingly, this would make for an effective if crude anti-farming measure. </blockquote>Actually, this is precisely what I meant. Instead of you being the only person the probability affects (you have a 1 in 100 chance of obtaining a rare) that probability is constantly being modified by the number of players in the zone so that instead of having a 1 in 100 chance, if there are 50 people in the zone harvesting you may only have a 1 in 50 chance of having a 1 in 100 chance due to the common frequency of resets of the looting tables. They would be resetting faster thus resetting the probability. Make sense or am I confusing it more?

Valdaglerion

06-15-2007, 01:19 PM

<cite>mamasan wrote:</cite><blockquote>The coin tossing and there being streaks, well, that's quite simple, theres only 2 sides to a coin so its a 50-50. But when theres 24 items and you get the same 3 in a row, thats not a 50-50, now is it? Thats a (1 in 24) x (1 in 24) x (1 in 24). Whats that? 1 in a million? But seems to happen constantly.</blockquote> Acutally no, the probability of repeats greatly increases with a smaller number set. For example, here are the results of a RNG with 24 numbers for 20 runs. So this could be the probability of drops for 20 runs into a raid zone if 24 items existed in the loot table. The numbers I am giving you are only those which were chosen multiple times:<ul><li>2 - 2</li><li>9 - 2</li><li>19 - 3</li><li>21 - 2</li></ul>Now, consider the recent changes where these mobs are only dropping loot for the classes in the raid/group. If you take 17 classes in your raid out of the possible 24 the numbers change even further (as expected). Here are those results:<ul><li>4 - 4</li><li>7 - 3</li><li>8 - 2</li><li>9 - 2</li><li>10 - 2</li><li>11 - 2</li><li>13 - 2</li></ul>Now you have 2 numbers which were chosen 3 or more times and a greater number of 2+ selections.Basic math dictates that the smaller the number of possible selections the greater the chance for reocurrence. take for instance, flipping a coin. You have a 50/50 chance in the outcome. The possibility of selecting heads multiple times in a row is highly probable because each time that outcome has the equal chance of happening. I feel your frustration in spending the time doing these zones and getting the same drops but the problem isnt with the math, its more likely in its implementation, if at all. It is in fact probable that you could do the zone with every class every time and be done in 24 times or an undetermined number of times due to the random nature of it. Just because you have a 1 in 24 chance of something happeneing does not mean that if you do it 24 times, everything will happen once.

YummiOger

06-15-2007, 01:49 PM

Hmm ..I Like the idea of a Open Loot table like for quest rewards. RNG could generate 4/5 Loot drops, then generate the option for 1/2/3 picks from that loot table. Raid Sees the whole Loot Table and how many to be picked. Raid Leader picks 1/2/3 items, DKP is bid on those items, peeps click the conferm window on the won loot. that would SERIOUSLY stop the 3 BPs in a row transmutes.Also the KoS loots are to specific IMO. especially the armor. Relic was too EZ to get (cuz 1 relic made 4 armor parts), but then restricting it to a 1:1 base class .. Heavyhanded to say the least. The Archtype armor in Labs was the most balanced loot set iv seen. 1 Armor fit 2 classes in the same Archtype. Well .. then there is the problem of the Class only abilities. so what to do?..I think KoS armor should be a Mixture of BOTH Relic and Archtype armor. Kill Mayong and maby a Divine Pulsing Breastplate will drop. Divine BP would be quest into both Inq/Templar Class Breatplates. 1 Drop useable by 2 classes. reduces the loot table by 12 drops and DOUBLES loot ability. Yet ur chance of seeing it is still only 8.3%. Small enough to make collecting a set difficult, but not frusteratingly improbable.

Themaginator

06-15-2007, 04:00 PM

Malchore

06-15-2007, 04:24 PM

ToiletBomb

06-15-2007, 05:07 PM

I dunno for everyone claiming random is random ... Well this random was made by someone. Secondly this random is now confirmed played with when it does the class check. I do not see why it is at all unacceptable to call this out. How can we be sure that the class check isn't indeed screwing the random over. This isn't probability, this isn't the laws governing mathematics and the universe, this is the laws governing EQ2s random number generator. As has been pointed out there are very obvious and serious issues with programming RNGs. They are simulated random. The other thing to remember which could very well be constant is raid setup. The more serious guilds are most likely running a very similiar if not exact same setup every go at some of these mobs. This further points that it is not unreasonable to question a possible bug with how the loot is selected. For example, what if the RNG simply rolls until a class present item wins. Great, the class present gets an item that is actually of use. BUT, what if for example there is a rounding up. So the same guild who never rolls with a brawler, or a defiler, or a whatever and these rolls are the ones winning. What if they are simply defaulting to another class based on the raid tree, or the time stamp, or as others mentioned a constant seed. This isn't mathematics at work it is the design of the generator. It is completely unfair to say this is 100% random because as we can see they have allready biased the system towards classes present. This obviously shows they can indeed skew it and without knowing how they skew it specifically it is unfair to say it is guaranteed random. There very well could be an issue in the way the check takes place for classes present. Of those getting incredibly streaky results (yes we hear of guilds that have luck on certain pieces all the time) how much does your raid setup change from visit to visit? Are you raiding at the same time every time. Is group setup pretty similiar? IE, MT group always group 2 and therefore always same people etc. Bottomline, we have no idea how this RNG check is taking place and we do KNOW for a fact that the game fudges this number as is to take into account classes present. Just the fact that this check happens before or after the roll or even if there are multiple rolls if something "drops" that shouldn't to re-drop it will change the outcome significantly. Just something to consider in my opinion.

Korpo

06-15-2007, 05:34 PM

Rienlos@Crushbone wrote: <blockquote>I dunno for everyone claiming random is random ... Well this random was made by someone. Secondly this random is now confirmed played with when it does the class check. This random has been mathematically proven to be well within the limits of what can be considered random. I do not see why it is at all unacceptable to call this out. How can we be sure that the class check isn't indeed screwing the random over. This isn't probability, this isn't the laws governing mathematics and the universe, this is the laws governing EQ2s random number generator. Which has been mathematically proven to be well within the limits of what can be considered random. As has been pointed out there are very obvious and serious issues with programming RNGs. They are simulated random. None of the people that pointed out these "very obvious and serious issues" has any clue what they are talking about. They have provided no proof, unlike those that have mathematically proven that the RNG in this game is well within the limits of what can be considered random. Etc., etc., etc. more nonsense. See above. Just something to consider in my opinion. </blockquote>You are welcome to your opinion. My opinion that the moon is made of green cheese -- despite overwhelming evidence to the contrary, despite proof after proof after proof, despite people that know better explaining over and over again that it's not -- is just as valid as yours that the RNG is broken.

sah

06-15-2007, 05:44 PM

The problems with the RNG have nothing to do with the loot system for class armor. The problems have existed before the new loot system was in place and they've actually been around since even before EOF...and it is not just limited to class armor in EOF....If you're really interested in RNG's this is one of my former professor's websites with a ton of links about randomness: <a href="http://www.cs.berkeley.edu/~daw/rnd/" target="_blank" rel="nofollow">http://www.cs.berkeley.edu/~daw/rnd/</a>

Kizee

06-15-2007, 06:03 PM

<cite>YummiOger wrote:</cite><blockquote>Hmm ..I Like the idea of a Open Loot table like for quest rewards. RNG could generate 4/5 Loot drops, then generate the option for 1/2/3 picks from that loot table. Raid Sees the whole Loot Table and how many to be picked. Raid Leader picks 1/2/3 items, DKP is bid on those items, peeps click the conferm window on the won loot. that would SERIOUSLY stop the 3 BPs in a row transmutes.Also the KoS loots are to specific IMO. especially the armor. Relic was too EZ to get (cuz 1 relic made 4 armor parts), but then restricting it to a 1:1 base class .. Heavyhanded to say the least. The Archtype armor in Labs was the most balanced loot set iv seen. 1 Armor fit 2 classes in the same Archtype. Well .. then there is the problem of the Class only abilities. so what to do?..I think KoS armor should be a Mixture of BOTH Relic and Archtype armor. Kill Mayong and maby a Divine Pulsing Breastplate will drop. Divine BP would be quest into both Inq/Templar Class Breatplates. 1 Drop useable by 2 classes. reduces the loot table by 12 drops and DOUBLES loot ability. Yet ur chance of seeing it is still only 8.3%. Small enough to make collecting a set difficult, but not frusteratingly improbable.</blockquote>The only reason that the relic was too easy to get was because it dropped off trash. If SoE made it so you would only get a mold from killing a named mob then it would work out nice.I seriously hope they are doing something different with kunark sets.

Themaginator

06-16-2007, 05:27 AM

Rienlos@Crushbone wrote: <blockquote>I dunno for everyone claiming random is random ... Well this random was made by someone. Secondly this random is now confirmed played with when it does the class check. I do not see why it is at all unacceptable to call this out. How can we be sure that the class check isn't indeed screwing the random over. This isn't probability, this isn't the laws governing mathematics and the universe, this is the laws governing EQ2s random number generator. As has been pointed out there are very obvious and serious issues with programming RNGs. They are simulated random. The other thing to remember which could very well be constant is raid setup. The more serious guilds are most likely running a very similiar if not exact same setup every go at some of these mobs. This further points that it is not unreasonable to question a possible bug with how the loot is selected. For example, what if the RNG simply rolls until a class present item wins. Great, the class present gets an item that is actually of use. BUT, what if for example there is a rounding up. So the same guild who never rolls with a brawler, or a defiler, or a whatever and these rolls are the ones winning. What if they are simply defaulting to another class based on the raid tree, or the time stamp, or as others mentioned a constant seed. This isn't mathematics at work it is the design of the generator. It is completely unfair to say this is 100% random because as we can see they have allready biased the system towards classes present. This obviously shows they can indeed skew it and without knowing how they skew it specifically it is unfair to say it is guaranteed random. There very well could be an issue in the way the check takes place for classes present. Of those getting incredibly streaky results (yes we hear of guilds that have luck on certain pieces all the time) how much does your raid setup change from visit to visit? Are you raiding at the same time every time. Is group setup pretty similiar? IE, MT group always group 2 and therefore always same people etc. Bottomline, we have no idea how this RNG check is taking place and we do KNOW for a fact that the game fudges this number as is to take into account classes present. Just the fact that this check happens before or after the roll or even if there are multiple rolls if something "drops" that shouldn't to re-drop it will change the outcome significantly. Just something to consider in my opinion. </blockquote>i promise you srand() is a very random piece of code when seeded correctly (im guessing this is what is used in the EQ2 code or another language's equivalent), random number generators are very easily made...even with parameters in for the sets.

IKilled007

06-16-2007, 11:47 AM

All of this nonsense about "streaks" in regards to allegedly random events has led me to offer a thought experiment.Suppose before a raid, I am so confident I know which items will drop that I predict them? Suppose I predict them in advance with 75% or higher accuracy? I can actually plan for specific outcomes which are supposedly random. Now if it's the case that the future can be correctly guessed with 75% or better accuracy, isn't that incompatible with randomness?Furthermore, at what point do you as a scientist consider your system flawed? Suppose you did a coin flipping exercise to demonstrate statistical probability to your class of students. You flip a coin 10 times and all 10 it lands on tails. You flip it another 100 times and always it's heads. You flip it 10,000,000,000 times and always it's heads? Suppose you programmed a computer to simulate a coin flip... You tell the computer something like let x = [rnd]2 if x=1 then j$="heads" if x=2 then j$="tails" print j$ and then repeat infinitely. Now suppose every single flip for 10,000,000,000 flips was heads (1)? Would you still think the process was random? Would you not inquire into how the computer generated the number?Don't be fooled.

thebunny

06-16-2007, 12:21 PM

IKilled007 wrote: <blockquote>Suppose before a raid, I am so confident I know which items will drop that I predict them? Suppose I predict them in advance with 75% or higher accuracy? I can actually plan for specific outcomes which are supposedly random. Now if it's the case that the future can be correctly guessed with 75% or better accuracy, isn't that incompatible with randomness?</blockquote> No. If you could scientifically prove that the RNG follows a particular pattern, and thus predict with 100% accuracy (using some algorithm or other scientifically acceptable method) the next item selected over a large sample size, then yes you could prove that the RNG is broken. Simply guessing and being correct (especially given the extremely small sample size) is not sufficient to prove that there is a particular pattern to the RNG.Keep in mind that, for a finite sample size, randomness does NOT imply an equal distribution of results, but rather an equal probability for each result to occur on any given roll. Perceptions aside, unless you can scientifically prove that is not the case, there is no way you can say with any certainty that the RNG is broken.

Jrral

06-16-2007, 01:41 PM

Sashtan@Mistmoore wrote: <blockquote>i promise you srand() is a very random piece of code when seeded correctly (im guessing this is what is used in the EQ2 code or another language's equivalent), random number generators are very easily made...even with parameters in for the sets. </blockquote>Actually, while rand()'s total output is usually fairly decent, in normal use it can suck badly. That's because the normal implementation produces 32 bits with good total entropy, but the low 3 or 4 bits from that algorithm tend to have a repeating pattern and generally not be very random. Certain systems (FSF's current glibc versions, for example) have fixed this, but most commercial Unixes still use this old implementation of rand(). And common code for generating random numbers from 1 to N is "j = ( rand() % N ) + 1;", which uses only the low bits from rand(). Look up the man page for rand/rand_r/srand on a modern Linux system, the Notes section has a bunch of references and texts on the algorithms involved and their strengths/weaknesses. NB: that linear-congruential algorithm you were taught back in first-year CS? Good enough for class assignments, not really good if you care about statistical properties.

Themaginator

06-16-2007, 01:46 PM

Jrral@Unrest wrote: <blockquote>Sashtan@Mistmoore wrote: <blockquote>i promise you srand() is a very random piece of code when seeded correctly (im guessing this is what is used in the EQ2 code or another language's equivalent), random number generators are very easily made...even with parameters in for the sets. </blockquote>Actually, while rand()'s total output is usually fairly decent, in normal use it can suck badly. That's because the normal implementation produces 32 bits with good total entropy, but the low 3 or 4 bits from that algorithm tend to have a repeating pattern and generally not be very random. Certain systems (FSF's current glibc versions, for example) have fixed this, but most commercial Unixes still use this old implementation of rand(). And common code for generating random numbers from 1 to N is "j = ( rand() % N ) + 1;", which uses only the low bits from rand(). Look up the man page for rand/rand_r/srand on a modern Linux system, the Notes section has a bunch of references and texts on the algorithms involved and their strengths/weaknesses. NB: that linear-congruential algorithm you were taught back in first-year CS? Good enough for class assignments, not really good if you care about statistical properties. </blockquote>i'm not sure about other languages but in C++ you can seed it with the current time in milliseconds, which makes it pretty random (atleast from the times I've tested I've rarely come up with the exact same results i had before.) And yeah i must admit I'm only second year CS, so I'm quite sure even with a time seed eventually it will have some recurring properties.

Jrral

06-16-2007, 01:47 PM

<cite>thebunny wrote:</cite><blockquote>Keep in mind that, for a finite sample size, randomness does NOT imply an equal distribution of results, but rather an equal probability for each result to occur on any given roll. Perceptions aside, unless you can scientifically prove that is not the case, there is no way you can say with any certainty that the RNG is broken.</blockquote>Yes, but if each result has an equal probability on each roll then for a large sample size the distribution will approximate closely an equal distribution. Note that there is a statistical test for whether or not the RNG is broken. Take two characters that appear anecdotally to be getting different results, get fairly large and roughly equal samples from each, then test the null hypothesis that those two samples represent the same underlying population.

Jrral

06-16-2007, 06:21 PM

Sashtan@Mistmoore wrote: <blockquote>i'm not sure about other languages but in C++ you can seed it with the current time in milliseconds, which makes it pretty random (atleast from the times I've tested I've rarely come up with the exact same results i had before.) And yeah i must admit I'm only second year CS, so I'm quite sure even with a time seed eventually it will have some recurring properties. </blockquote>Language and seed don't matter in this case. It's the algorithm behind traditional rand implementations that's the problem. A random seed just means you get reliably different non-random sequences in the low bits. The books noted on the Linux rand(3) cover the problem, why it happens and how to avoid it. Note: if you're on a modern Linux system with current glibc, you won't see the described behavior. Again, the man page notes that glibc threw out traditional rand(3) and replaced it with a wrapper around random(3) which uses a much better algorithm that doesn't have the low-entropy low bits problem.

ToiletBomb

06-17-2007, 04:50 AM

<cite>Korpo wrote:</cite><blockquote>Rienlos@Crushbone wrote: <blockquote>I dunno for everyone claiming random is random ... Well this random was made by someone. Secondly this random is now confirmed played with when it does the class check. This random has been mathematically proven to be well within the limits of what can be considered random. I do not see why it is at all unacceptable to call this out. How can we be sure that the class check isn't indeed screwing the random over. This isn't probability, this isn't the laws governing mathematics and the universe, this is the laws governing EQ2s random number generator. Which has been mathematically proven to be well within the limits of what can be considered random. As has been pointed out there are very obvious and serious issues with programming RNGs. They are simulated random. None of the people that pointed out these "very obvious and serious issues" has any clue what they are talking about. They have provided no proof, unlike those that have mathematically proven that the RNG in this game is well within the limits of what can be considered random. Etc., etc., etc. more nonsense. See above. Just something to consider in my opinion. </blockquote>You are welcome to your opinion. My opinion that the moon is made of green cheese -- despite overwhelming evidence to the contrary, despite proof after proof after proof, despite people that know better explaining over and over again that it's not -- is just as valid as yours that the RNG is broken. </blockquote> Yes, please show me the mathematically proven documentation for class specific armor drops ... furthermore have you seen the code? Skepticism is a perfectly healthy alternative. I haven't said the RNG is broken. So please, lay off. The RNG seems to function like most other RNGs made. That doesn't mean it doesn't have flaws. Just like a roulette wheel spinning, you can't argue that a mechanical device will wear and possibly "prefer" a set location. I am merely questioning something and providing (well what I consider decent) theories to the result. And yes the people that have showed it do know what they are talking about. It is impossible. We can get very very very close to random but it just isn't the case in modern day.

thebunny

06-17-2007, 11:34 AM

Jrral@Unrest wrote: <blockquote><cite>thebunny wrote:</cite><blockquote>Keep in mind that, for a finite sample size, randomness does NOT imply an equal distribution of results, but rather an equal probability for each result to occur on any given roll. Perceptions aside, unless you can scientifically prove that is not the case, there is no way you can say with any certainty that the RNG is broken.</blockquote>Yes, but if each result has an equal probability on each roll then for a large sample size the distribution will approximate closely an equal distribution. Note that there is a statistical test for whether or not the RNG is broken. Take two characters that appear anecdotally to be getting different results, get fairly large and roughly equal samples from each, then test the null hypothesis that those two samples represent the same underlying population. </blockquote>True, the problem is we're hardly talking about large sample sizes here. If you could generate a large enough sample size (say 1 million data points) that show a non-equal distribution of results, then you may have something. But simply doing /ran 100 a couple hundred times is statistically irrelevent, and hardly proves there is anything wrong with the RNG.

Jrral

06-17-2007, 02:22 PM

<cite>thebunny wrote:</cite><blockquote>True, the problem is we're hardly talking about large sample sizes here. If you could generate a large enough sample size (say 1 million data points) that show a non-equal distribution of results, then you may have something. But simply doing /ran 100 a couple hundred times is statistically irrelevent, and hardly proves there is anything wrong with the RNG.</blockquote>I doubt you'd need a million sample points. If, as is claimed, the RNG is good and it's results aren't being skewed by other things in the game like other players' activity, then I should be able to get past the 95% confidence level with fewer than a hundred data points. I can hit that many harvest nodes in an hour, so I should be looking at two samples of a thousand or so nodes each. That's in the neighborhood of the sample sizes used to predict national elections, bear in mind. And /random isn't the correct thing to test, because it's as much the use of the random numbers as the actual RNG that's involved. That's also why I have to compare two characters: since I don't know the underlying distribution of for example the harvest node drop tables I can't test for whether a single character's results match the expected values, but I can test whether two characters' results come from a population with the same expected values. Now, for /random 100 I'd need more than a couple hundred rolls. I need each result to have an expected frequency of at least 10 or so for the tests to be usable, so I want a sample size of at least a thousand. But when I'm looking at harvest results, I'm going to be lumping things together into common raws, imbues and true rares. I'll be ignoring node and harvest type, other than the requirement that both characters harvest roughly the same number of each type of node, and looking at quantities of those three categories. I'll be recording data at the most detailed level, mind, so I'll be able to get more detailed eventually, but first pass is simply whether their total haul matches up with the idea that the system isn't favoring one with better harvest results than the other.

Leemeg

06-17-2007, 03:29 PM

We as player has no possibility to actual prove that the random generator is correct or not. Why? 1. No way to gather enough data spread all across the server to say that one item drops more than another one. If you raid a zone once a week, there is maybe n other guilds that does the same. You will need the data from all the guilds that is raiding that specific zone, which I hardly believe anyone has. The same is valid for group instances. If enough data is available, it is probably still not possible to make out a pattern (see next point). 2. The random generator may be random, but is the chance of a specific item drop compared to another equal? We don't know. SOE can probably adjust the ratio of the different loot in the loot table. If there is twice as many fighters than mages on the server, is it that wrong thinking that they will adjust the ratio so that the total number of fighting gear drops is actual higher than mage gear drops? This is a design decisions from SOE side, and we will never know the fact around adjustments like this. We can only guess.