Observation of the RNG

[Jaytee Bushwacker]

Well, it may be a horrible conclusion but its certainly intuitive that smaller combines are easier to skill up on. In addition they have the potential to fit more skill ups within a "streak".

Take Player1 making MHB and Player2 making Twice Brewed Constitutionals. Player1 cannot possibly place all 9 items for MHB in the barrel at the same pace as Player2 doing the 4 item TBC's. So Player2 is on average going to skill up faster than Player1 simply because it takes Player1 longer to do the same number of combines.

Now if there is any "streakyness" at all, Player2 will be in a better position to take advantage of a "good" streak simply because they are combining more within a short period of time.

[Fundin]

The Roots of Coincidence

All this talk of the RNG, and especially some of the deeper discussions of randomness overall reminded me of a book I read many years ago.

The Roots of Coincidence in which the author, Arthur Koestler muses on things such as how all the dogs in New York city know to stop biting people (or that some best keep biting) in order to satisfy the statistical norm for the annual number of dog bites in New York city. :roll:

All I know is that on guild raids, my partner Ayevee is 100% assured to win more rolls than I can ever hope to. Then why am I the one working on the coldain shawls?

[Soulfiend]

If for example you have a 25% chance to combine something successfully and you fail, you still have a 25% chance the second time, and third etc etc etc.

It's like a lottery, people still believe that 6 non related or random numbers will win more than picking 1,2,3,4,5 and 6.

[Jaytee Bushwacker]

On any one roll you have a 1 in whatever chance, but on a consecutive roll you are more likely to roll any other availible number.

Example:

Possible Numbers: 1, 2, 3, 4, 5, 6

Roll #1: you get a 4

On the next roll you have a 16.6% chance to roll any of the possible numbers. However, in order to attempt to predict what will come up next we use past history to influence our prediction.

Possible combinations of 2 consecutive rolls when the first roll results in a 4.

4 and 1
4 and 2
4 and 3
4 and 4
4 and 5
4 and 6

So you have a 83% chance to roll a number other than 4 on the second roll. Each consecutive 4 that turns up the chance for an additional 4 is reduced.

[Verdandi]

On any one roll you have a 1 in whatever chance, but on a consecutive roll you are more likely to roll any other availible number.

No. If you do a /rand between 1 and 6, the odds of getting a 4 are the same on every roll no matter how many times you roll. If you roll a 6-sided dice once, nothing about the dice changes that makes it more likely to turn up a different number on subsequent rolls. The odds of a given number on any particular single roll stay the same. When you deal with multiple rolls, what changes are the odds of getting two particular numbers in a row. Two 4's in a row are rarer than a single 4 on a single roll, because two rolls are involved and there are more possible combinations. But the odds of getting two 4's in a row are the same as the odds of getting any other combination of 2 rolls, like 4 then 1, or 2 then 6, or two 5s.

Edit: condensed to save space after reading Morani's post, but leaving the rest here so her post makes sense.

[Morani]

On any one roll you have a 1 in whatever chance, but on a consecutive roll you are more likely to roll any other availible number.

No. If you do a /rand between 1 and 6, the odds of getting a 4 are the same on every roll no matter how many times you roll. If you roll a 6-sided dice once, nothing about the dice changes that makes it more likely to turn up a different number on subsequent rolls. The odds of a given number on any particular single roll stay the same. This is one of those things I remember our teachers spending several days pounding into our heads in school.

Jaytee is correct but the wording is poor. Sorry Jaytee.
Essentially what is being said is that (with 6 sided dice) you have a 1 in 36 (or 2.777777%) chance at getting 2 sixes on 2 rolls and a 5 in 36 (or 13.888888%) chance at getting 1 six and 1 something-other-than-a-six on two rolls. AND that a 13.88888% is higher than a 2.7777777% chance. That's all. What is being left out is that there is a 30 in 36 (or 83.333333%) chance that something other than a 6 will come up on the first roll.

I reacted the same way when I read the first message and was going to correct the common misconception, but then I realized I was misunderstanding what was being said. Essentially, the two of you are saying the same thing but from two very different angles: Jaytee is loking at the 2-roll set picture and you're looking at them as seperate events. The numbers are the same. In the larger set, there are 36 possible outcomes with six of them being a 1 rolled first, six having a 2 rolled first... 6 out of 36 is the same thing as 1 out of 6; both are 16.666666%.

[PinyonTreedotter]

Its fun because Aid Gimil has an earring waiting for me =P Until then I avoided trades like the plague. And really until SoL there really wasnt much worth making imo.

This is how SOE is killing the tradeskills.

The RNG seems to have a fairly high autocorrelation. Like most games, I would guess that SOE didn't spend a tremendous amount of time on the RNG -- particularly since better RNGs require much more computing time.

I have done an statistical analysis and discovered that you are more likely to get skillups on success than failures. SOE tells us they did not program it this way. If we assume SOE is not lying (possibly a big assumption), then this effect means "streakiness." On the other hand, I have not yet figured out a way to exploit.... errr..... use this information, as you can tell by my tinkerer, who is averaging about 50 combines per skillup in the 210s.

[Verdandi]

Ohhhhhhhhhh, thanks Morani, I see where he was going with that now. Sorry I wasn't following you there Jaytee. ops:

[Itek]

Jaytee is correct that "once you have succeded you are slightly higher likelyhood to succed again" ... Marginally.

But, and I can't stress this enough.... One -cannot- predict -either- the beginning OR the end of a "streak" without cheating. Unless you rig the wheel, load the dice, stack the deck, or decompile the RNG it's just NOT possible.

Thank you to the person that pointed out a "server wide RNG" is not only silly, it's computer science theory a pain in the buttocks to try to pull off. (each computer has a RNG built in, why bother importing a RNG-result when you can produce one locally "cheaper" in FLOPS? Floating Point Operations Per Second)

They also pointed out why you don't get the same "hit/miss" or "damage" when you swing two equal weapons on a dual wield/double attack.

Because, and let's again go back and read my first reply, there are a few /ran requests between each of your swings. You can only get the same result from a RNG if you can /ran faster than the seed number changes.

Keep the lucky rabbits foot. My only -real- problem with the theory of "use faster combines in a streak" is this....

If you have a "faster combine" formula available (say Champagne instead of Mino Hero Brew) why aren't you using it all the time?

More combines per hour = more skill ups per hour. No matter WHAT the streakiness of the the RNG.

COST. It's much cheaper to do MHB vs DSC.

So Jaytee is saying "save those PRICY combines for use in the -player-friendly-streak- and do those cheap slow combines till you hit a streak" ...

Again and again and again.... If someone tells you they want to sell you a "betting strategy" or "proven winning method" to beat craps or roulette or a slot machine... They are

a) a con man

b) really sadly mistaken

"Las Vegas exists for a reason" (people are bad at math)

"There is no spoon" -The Matrix
"There is no streak, until it ends, and then it's too late." -Millions of Gambler's Anonymous

[Morani]

I'm going to offer a way to test the randomness of skill-ups that's relatively easy and certainly cheap: Languages. It is now possible to train one character in 1 point of a language and use that character to train another up to 100 from 0. (I've done it with 9 languages in the past week). There are 25 languages so doing 20 of them from 0 to 100 gives a data set of 2000 skill-ups per character that is a minimum of level 19. It's then a simple matter of plotting the numbers out to see the spread of time between them. If there is some sort of streakiness, it should show up as clustering in a simple X-Y graph of the intervals.

I can volunteer a character on Druzzil Ro that has mastered 22 languages and will shortly have the rest as a trainer.


Personally, I do my combines in a semi-slow and steady rate until I get a skill-up and then I go really fast for the next 3-4 combines because it just makes me feel good to do it that way. Either that or I don't even pay attention to them and run my set of combines (I only do the latter when I KNOW I'm not going to trivial out with that set).

[Jaytee Bushwacker]

Jaytee is correct but the wording is poor. Sorry Jaytee.

Mez wurdin aint da bestest but glad u getz it!

Huk'd on foniks werk'd fer me!

[Nyalria]

I'm human, I notice the steaks/patterns, and I believe they may be useful, or they might be imagined. I sure as hell notice that fizzles often come in twos. And I really notice when a EXTREMELY trivial item fails 2-3 times in a row. <shrug> Just thought of an idea, assuming they're not imagined.

For like tailoring, you can have multiple kits open right? Set up 2 kits with what you're doing, and then do combines normally in the 3rd. As soon as you get a skill up, quickly drop product and click/drop/click the other containers. =p

It might make you feel better, or maybe worse, and will probalby take up precious bag space. Maybe it'll be fun and interesting experiment. Maybe it'll be something to make the RSIs less bothersome. =P

[Itek]

I'm human, I notice the steaks/patterns, and I believe they may be useful, or they might be imagined. I sure as hell notice that fizzles often come in twos. And I really notice when a EXTREMELY trivial item fails 2-3 times in a row. <shrug> Just thought of an idea, assuming they're not imagined.

Yes, you are human. (unless you are an alien, but presumably they have access to better games *just kidding*)

And yes, you notice the patterns.

Hehe, true story. While playing D+D one time my fighter missed 9 attack rolls in a row. With a "to hit" of 4. I rolled 3 or less 9 times in a row. The odds are about 26 million to 1 against doing that. I was outraged. But we also thought nothing of getting 2 20's in a row. I met a guy who once claimed to roll all 18's. *shrug* I know a friend once spent a weekend trying to roll all 18's. Hey, the odds are 101 trillion to one, but there's a chance right? (at that level there is some "skill" in throwing the dice which begins to skew the data *grin* practice makes perfect, except when slicing a cake in EQ)

Unfortuneately, and a previous poster pointed this out, humans are really, really BAD at noticing patterns correctly. They miss them when they ARE there, and notice / invent them when they are not. (Conspiracy theories for example.)

"Bad things come in threes." yes, because you group them in threes.

"Third time is a charm." talk to the guys that have failed tradeskill trophies 6+ times, sure they would love to hear this.

"I'm on a hot streak" said at a million blackjack tables by people losing the rent...

Yeah, I noticed the other day when I fizzled a lot more than normal. I got 3 in a row, I fizzled a spell in combat on 3 pulls in a row...

On the other hand I didn't really notice how much I had come to rely on NEVER getting interrupted till I got interrupted twice in one day.

Here's one.... "six degrees of seperation" .. The theory, supported by math and example, that if you know 100 people and they all know 100 people, all of whom know 100 people etc etc ... by the time you reach the "sixth degree" you know someone thru someone who knows ALL the people on earth. The math bears this out.

Except you know Bill from work. And you know Bob from work. That's 2 of the 100 people you know.

Unfortuneately Bill also knows Bob and you.
And Bob also knows Bill and you.

So, this is the part NO ONE will understand.....

YOUR 100 people have to be TOTALLY UNIQUE from all the 100 people you know's 100 people THEY know for it to work out.

People are BAD at math. Inherently. They have perceptions of how math and probability work that are just wrong.

Las Vegas exists for a reason. You CANNOT predict a streak, nor know when you are in one, NOR know when it will end.

Yes, you could keep 2 combines "in waiting" for the streak for things with portable containers.

But you are no more likely to skill up on them than if you just kept plowing along.

A poster said earlier "I got lots of skill ups all at once, even had 3 skill ups in one chat window. so streaks must exist and doing fast combines must be valid."

Um, gee. Hate to break it to them, but I got 3 skill ups in the high 200's with Mino Hero Brew all in a chat window. Not exactly a "quick" recipe. And the same thing happened to my wife. But there were other times, a LOT of other times, when we would get a skill up... and then nothing for 3 stacks. Then get 2 in a row... then nothing for 10 combines and a point... then 30 combines and a point. Then 3 combines and a point.

You remember the extremely GOOD and BAD runs. They are memorable.

If you have better success "doing the funky chicken dance" before a tradeskill run... happy dancing. I will admit I cross my fingers before runs, and on tough combines. I admit I tried the "fizzle a spell before attempting a trophy" trick... It can't hurt. It might help. Or at least put you in the mood to do the combines.

I only say "no, not true" because I don't want someone to come here and say "I did it Jaytee's way and got less than average success, you all !$%@#$%"

Know the truth, and HOPE that you can beat the odds. Except in Vegas. There just figure it's like paying to be IN the movie rather than watching it. *grin*

[Jaytee Bushwacker]

When I vacationed in St. Louis in 2001 I played blackjack each day. I gave myself a limit of 50 dollars I could lose, in fact thats all i entered the casino with.

Day 1 : won 45 dollars
Day 2 : won 100 dollars
Day 3 : won 150 dollars
Day 4 : won 250 dollars
Day 5 : won 503.50 dollars

I had the same cabby the whole time and every day he asked me how much I lost...I just told him the inverse of what I won (i.e. I lost negative 45 dollars the first day)

Basically I had a nice week in St. Louis on the casinos =P

Has nothing to do with streaks but hell I enjoyed it!

[sulyen]

Random number generators can be uniform or non-uniform. Uniform random number generators are used to simulate dice.

A uniform distribution is one for which the probability of occurrence is the same for all values of X.

There are many algorithms and methods for generating random numbers.

The majority of random number generators assume a single source drawing numbers. IE, for a single source drawing numbers the random number algorithm will function as designed.

There are several tests which can be run on number generators to tell us how well they work. Entropy, Chi square distribution, and arithmatic mean, are some examples. One single test can be misleading so many tests in combination are run.

The human eye/brain is the best pattern recognition system. If a high percentage of people find a pattern the result should not be dismissed. However, humans are surpassed by machines when tabulating said results.

Did the person designing the tradeskill system consider whether they wanted a high or low entropy random number generator?

Does the end result EQ random number generation system function mathematically correct? Do we/they care?

Is it a good tradeskill system?

Does the tradeskill system work?

Will people quit playing EQ because the tradeskill system doesn't satisfy them?

I don't know about you, but as I focus on individual parts of EQ I see extremely flawed and horrid designs mixed with a few gems. But EQ is a success and is a shining example of how the whole is much better than the sum of the parts.