Math and Spawn Rates

[Zenlina]

I conclude u probably killing the wrong skeleton. That one is quick to spawn.

[Sirken]

btw, your poker analogy is good, but doesnt apply because you are removing cards from the deck, hence every round your odds of drawing to that flush change for better or worse depending on the previous card or cards put out (depending which game ur playing).

a closer analogy would be blackjack, but with a brand new deck for each hand.

each pop (be it ph or the rare mob) is a completely separate encounter, and what pops has absolutely 0% to do with what popped before that.

not trying to be a dick, just explaining [You must be logged in to view images. Log in or Register.]

[Splorf22]

Originally Posted by Tanthallas [You must be logged in to view images. Log in or Register.]

This is highly misleading. The conditions that have to be met in order to calculate the probability of events in terms expected value are extremely unrealistic and in most cases make the calculation itself moot.

Best way to think about it is like Sirken said. It doesnt give you a practical answer because there is no practical answer.

You must be an undergraduate. I'm not sure whether you just dismissed the entire discipline of statistics with an airy wave of your fraternal hand or whether you have never heard of machine learning. There are computer poker players that use precisely this concept to play poker. Obviously you cannot know for sure what your opponent will do when you checkraise him on the turn, which is why you can estimate it from training data using any number of techniques.

[Nlaar]

Originally Posted by August [You must be logged in to view images. Log in or Register.]

No need to be patronizing. A random number generator, based on implementation, can very well get 'stuck', or atleast provide very non-random outcomes with regularity

I'll give an example. I work at a pretty nice software firm that suffered a problem whenever running... things through other things.

See, the test in question always got hung and no one could figure out why. It was a series of logic steps that required a Random() call 10 times. It just so happened that if a certain number hit during this random jump, then the test would enter into an infinite loop. Shoddy programming, and the fix was easy. The strange part, however, was the frequency in which it happened.

See, the calculated frequency of this happening was below 1% if you went by pure odds. However, we noted that on 33% of our runs this would happen.

But, August! you say, what if you used the same SEED -- of course it would happen over and over! And so I investigated and found that seed was taken off of milliseconds of the current time -- 'random' to the 1000 chance, I suppose.

So, I made a little program that does the same thing as the test without the encumbrance, of, say, the test. And then I printed the results out with the different seeds, the random numbers generated, etc.

I found a couple of things upon parsing:

1) The same seed repeatedly got used when I ran the test -or rather, the same 'set' of seeds got cycled over and over. We were just taking milliseconds, but the same 3 digit seeds kept popping up. This could be a problem in numerous places, but maybe the system clock!

2) When we modulo'd by our bound there was the dreaded value in about 1/3 of these sets of numbers. Hence our near 33% hang rate. I then went and force-seeded all 1000 millieseconds and found that in around 23% of them, the first 10 rands on our modulo would produce a hang-result by random 10.


So, what can we learn?

1) Nothing is random
2) Implementation is everything

The OP may be unlucky, or there may be something wrong w/ the code behind it, or the OP may be doin something wrong -- I don't know.

But let's not call people tinfoil hatters when there could be a logical explanation. Programmers aren't perfect, and nobody that I'm aware of has ever created a truly 'random' implementation in software.

Thank you August (for both the info as well as calling S. out on his tone).

Well said - with my somewhat limited CS knowledge I would have butchered the point to be made.

[Lagaidh]

Originally Posted by Frieza_Prexus [You must be logged in to view images. Log in or Register.]

The RNG on live was considered to be notoriously streaky. My inner conspiracy theorist suspects that something has been done here in an attempt to recreate the frustratingly "off" nature of EQ's RNG.

I've been thinking the RNG is way streakier on p99 after the patch that brought high quality ore as drops in sol a. I've also not had the initiative to try and record any numbers =)

From one gut feeling to another, these are the assumptions of our lives.

[Shamen]

Originally Posted by Handull [You must be logged in to view images. Log in or Register.]

The last thing to consider is that the people who get the bad luck of not getting a spawn over X hours will likely be more vocal about their failure than the person who just found the mob up randomly, or got a spawn in only a short period of time. So just by browsing the forums it would sound like tons of people have this bad luck, but there are just as many who have good luck and don't post about it.

True, I started doing frenzy camp like 2 weeks ago and I got 3x fbss in first 7 cycles! Been doing it a lot since that (est. 60 cycles) and not a single fbss drop... its all about luck.

[Estu]

Yeah, you guys talking about faulty RNGs have a point, but so does Sirken. Someone early in the thread talked about how people don't think random numbers are as streaky as they actually are. To illustrate this, let me tell a short story.

I once took a probability course and the professor divided the class into two halves - he told one half to flip a coin and write down a string of 100 ones and zeroes for the heads and tails flips, and he told the other half to try and think of 100 random ones and zeroes in a row. He told us he'd leave the room, and to have one half of the class write their string of ones and zeroes on one side of the board, and the other half of the class on the other side. When he came back to the room, he'd have to guess which half was 'truly' random (the literal coin-flippers) and which half was trying to come up with random numbers.

He left the room, we wrote up the numbers, and we called him back in. He looked at the board for a couple of seconds and correctly pointed out the truly random numbers. How did he know? They had longer streaks. The coin-flippers had streaks of 7 or 8 zeroes or ones in a row, while the people who tried to think up random numbers never had a streak longer than, say, 5 zeroes or ones.

OK, so what's the point of the story? The point is we humans have a poor intuitive understanding of random numbers. We play EverQuest or whatever game and we see a long streak and we think, "Shit, this game's out to get me. It's not truly random, there's a bug in the code or it was done intentionally to screw with people." But we're INCLINED to think that because we intuitively believe that random numbers are less streaky than they can be mathematically shown to be.

The point is we can't rely on intuition to make these judgments. We need to take raw data and make statistical analyses, like was done in this thread. Yeah, if the skeleton has a 5% spawn rate then the probability of a guy going 270 spawns without a pop is one in a million. My personal conclusion to that story is that the wiki has something wrong in the spawn rate or the placeholder information (which is EXTREMELY common since the wiki has various sources, whether it be newer code from the game (e.g. maybe the spawn rate was increased in a later patch to make it less of a pain, which is the info the wiki is using, but in P99 it's the low original rate), hearsay (someone read something on an old allakhazam post or heard something from a random guildie), or straight-up speculation), so the wiki is giving the wrong probability or the guy trying to get the spawn is doing it the wrong way.

But if we're looking at anecdotal things without actual probabilities and data, NO WAY are we qualified to make judgments about the RNG and its streakiness based on our intuition. Which is what some people are doing later in this thread, and where I agree with Sirken and his tin foil hat image.

[Lagaidh]

Originally Posted by Estu [You must be logged in to view images. Log in or Register.]

Phil Spector's Wall of Text

I agree with you whole heartedly. Like I said of myself, I didn't have the will to truly analyze what I was perceiving.

Even as I type now, I know I've put in enough time at a camp that I begin to let "instinct" tell me something is different before and after a patch. Too bad instinct is a tool to keep us alive and feeling justified in decision making.

It's sort of amazing... on one hand, I'm a human that understands the staggering improbability of Joe DiMaggio's 56-game hitting streak, but on the other hand, I "kinda" listen when my gut is telling me that the RNG is a bit "too" streaky.

We really are still an infant species aren't we? I know better, but my biochemical programming likes to interject lies...

Man. I'd better start smoking pot again if I'm going to get all metaphysical over a RNG discussion.

[rahmani]

Originally Posted by Sirken [You must be logged in to view images. Log in or Register.]

btw, your poker analogy is good, but doesnt apply because you are removing cards from the deck, hence every round your odds of drawing to that flush change for better or worse depending on the previous card or cards put out (depending which game ur playing).

a closer analogy would be blackjack, but with a brand new deck for each hand.

each pop (be it ph or the rare mob) is a completely separate encounter, and what pops has absolutely 0% to do with what popped before that.

not trying to be a dick, just explaining [You must be logged in to view images. Log in or Register.]

Just a side note on the Gambler's Fallacy.

Gambler's Fallacy is only the wrong assumption that any individual random event is influenced by the previous random events.

However, it is true that over long periods of time and large values of n the binomial distribution will approach the probability of the possible outcomes. This is in fact how probabilities are deduced.

It is correct to say that it is increasingly improbable for the set of n observations to deviate significantly from the distribution as n increases.

So, if there is a probability of 5% (p=0.05) of an event producing a 1, and all other times (p=0.95) will produce a 0, then at n=1,000,000 you will have a distribution that is very close to the probability. In other words, if you don't have very close to 50,000 1s, something is seriously broken.

* Binomial Distribution
* Law of Large Numbers
* Gambler's Fallacy

[rahmani]

Also, no computers generate truly random numbers, which is a real problem. Computers that generate the most random-acting numbers sense environmental changes caused by fluctuations in the atmosphere.

However, EverQuest like all games, uses a pseudo-random number generating algorithm, and must, by definition, fail at the serial randomness test, which compares all identical lengths of digits and weighs them against other sequences of the same length.

Go to Random.ORG, a site dedicated to presenting with the most seemingly random numbers. The numbers are not determined by a set algorithm, instead they are sensed.

In the following sequence, for any string length, L there should be an identical number of occurrences as the total number of digits approaches infinity. And as the digit length increases by 1, the digits should be ten times less frequent.

Sample of single digits:
0:205x, 1:207x , 2:214x, 7:199x, 9:189x

Sample of double digits:
00:19x, 01:20x, 69:26x, 99:17x

The sample size of triple digits is small, although they do exhibit some of the same random properties.

Sample of triple digits:
100:4x, 999:1x, 434:2x, 901:1x



60801468158158128265500664544694727122124230258075 89571133284564798639346318666510097025111188884005 21923075241160765601985943448956025269830536604941 99504230101058822671207931152613178877543677697509 02706771661102759463379639530443615657163698062759 06337782166957101973173212845041749226483640829555 86423040123454106792574561123269430351901285071917 86620163674835453275555190813444328411810157776825 13234587722475732237420540620556244139615521226905 81598019524566660334579294592358173180858402456420 60342278560260571770605545673269139285436949040963 30631986605922117878478030732571918682208101686355 29768144509889511043749779223728743822621155920732 36368577581258703118950979428887523049210023168408 37427100135119107190534513416545895979251093257271 79042017501900644993469774830551072737674466300623 59739756943917079386634900082442716979766978266522 45471379075300444617560397168250547024130633105838 51880321514936357619795116325315994695612797571243 98436734812197672492480992432227236861653412859863 21364291004588255654211046475385828878867224992808 14151174625825430174357532675458656989750344694063 15091271709802352516436199242258530307756137327440 15705758761765938234998973921542774665875715631334 76368242047502643509555881716221668548280049122183 81667194126560994767831904541927309933743864620394 51851359628452986890550026423125255750838304060402 77216336873695117866982789903107180513800797868540 43313265047972927614346809614367102894172556879749 52825882627033025084021055514790494200502341827015 10966860314288573746489647976518028880180564128814 81050576929140396109697527053447366369392520319779 99590609037699619266954394131024058087880493199574 80381552872508846866412026033903449269278810722898 17264966881768923809416025761438974757512589601200 62056044130156627607563524672328072411751402038849 28602825848721207866700913852009939022369127335074 95818145156261571036972631389596368739085531394331 28376550816382510244987659629574966734292129941784 23792086413272992712103998674454880545579600500516

Try the same test on ANY pseudorandom algorithm, it'll fail.