Tweet
Every time I look at eqtraders, I read about how screwed up the RNG is. In point of fact, it is functioning perfectly well. Let me explain (for the mathematical purists here, I will admit to lots of rounding and other similar simplifications - so sue me):
The first thing you need to know about random systems is that they tend to cluster in small samples. That means that if you don't do the same thing LOTS of times, you will get a lot of repeats and a lot of results that never happen at all. As an experiment, take a 20-sided die and roll it 20 times. Here is what I got when I did that: 7 11 1 3 5 6 19 3 4 19 18 9 18 5 3 8 13 4 7 7. If I look at just 20 rolls of the die, I am going to think something is wrong - I got 3 7's and no 14, 15, 16, or 17 after all. I did it again and I got 9 13 3 14 16 17 9 14 20 20 20 3 4 15 12 10 15 13 1 19. Wow, 3 20's in a row! Two 14's, 15, 16, 17, and 19!
When most people think of random numbers, they expect to get a pretty even number of each result. The fact is, unless you do ALOT of trials, values will cluster around a few values. As you do more and more trials, the results tend to even out. Some of you might remember being asked to flip a coin and record the results in a math class. Do you know how the teachers figured out if you actually did the test? Well, the students that did the test had a record like this:
H T H T T H H T H H H H H H H T H H H T T T H H T T
If the student made it up, it more often looked like this:
H T H H T H H T T H T H H T T T H H T H T T H T H T
Real random systems are virtually GUARENTEED to have a "streak" of 5 heads or tails in a row in a small number of flips. By counting up the number of flips in a row, the teacher can tell pretty much at a glance if the student actually did the test or made it up. Counting on the second, there was ONE sequence of 3 tails in a row, and all the rest had a maximum of 2 in a row. It appears to be random to the student, but in fact is not random at all. Each time the student was deciding to put an H or T next, they were thinking about what came before it. In other words, the second string of Heads/Tails had a memory effect (the student) that was trying to even things out because that was the way they thought it should be. In the first string, the coin didn't care what came before, because it was a truly random effect. In a short number of trials, it looks screwed up, but if you do a LOT of trials, it would be virtually identical numbers of heads and tails.
Okay, so what does that have to do with skillups in tradeskilling?
Most people have found that the skillup rate over a large range (say from 190 to 250) is around 1 in 20. In that range, there are 60 skillups. Let's say that I just skilled up a point. On the NEXT try, I have a 1 in 20 shot of getting another skillup. Since I have 60 skillups, I can EXPECT to get about 3 "double" skillups from 190 to 250. About 1 in 6 tradeskillers will get a "triple" skillup, 1 in 120 will get a "quad" skillup, and 1 in 2400 will get a "quintuple" skillup between 190 and 250. People that get these lucky streaks claim the streaks as evidence that the RNG is broken. In point of fact, the streaks are the biggest proof that the RNG is working as it is supposed to.
On the other side of the coin are the people with really long streaks without a skillup. Unfortunately, they too are sign that the RNG is working. With a 1 in 20 chance of getting a skillup, each sequence of 100 combines has a half-percent chance of giving you NO skillup. (19/20)^20. It doesn't happen to many of us, but it is almost guaranteed to happen to someone. My personal worst was from 199 to 200 in jewel crafting: 320-ish combines. For the record, the odds of that result are astronomically low, but I am ever the lucky guy.
Part of the problem with the RNG is that we remember the streaks - 3 fizzles in a row, or 3 skillups in a row get our attention, as do 320 combines without a skillup. If you average the results over all the tradeskillers in EQ, the RNG works and gives even results across all attempts. Your personal mileage may vary.
In other words, the game design may suck, but the RNG works just fine.
The first thing you need to know about random systems is that they tend to cluster in small samples. That means that if you don't do the same thing LOTS of times, you will get a lot of repeats and a lot of results that never happen at all. As an experiment, take a 20-sided die and roll it 20 times. Here is what I got when I did that: 7 11 1 3 5 6 19 3 4 19 18 9 18 5 3 8 13 4 7 7. If I look at just 20 rolls of the die, I am going to think something is wrong - I got 3 7's and no 14, 15, 16, or 17 after all. I did it again and I got 9 13 3 14 16 17 9 14 20 20 20 3 4 15 12 10 15 13 1 19. Wow, 3 20's in a row! Two 14's, 15, 16, 17, and 19!
When most people think of random numbers, they expect to get a pretty even number of each result. The fact is, unless you do ALOT of trials, values will cluster around a few values. As you do more and more trials, the results tend to even out. Some of you might remember being asked to flip a coin and record the results in a math class. Do you know how the teachers figured out if you actually did the test? Well, the students that did the test had a record like this:
H T H T T H H T H H H H H H H T H H H T T T H H T T
If the student made it up, it more often looked like this:
H T H H T H H T T H T H H T T T H H T H T T H T H T
Real random systems are virtually GUARENTEED to have a "streak" of 5 heads or tails in a row in a small number of flips. By counting up the number of flips in a row, the teacher can tell pretty much at a glance if the student actually did the test or made it up. Counting on the second, there was ONE sequence of 3 tails in a row, and all the rest had a maximum of 2 in a row. It appears to be random to the student, but in fact is not random at all. Each time the student was deciding to put an H or T next, they were thinking about what came before it. In other words, the second string of Heads/Tails had a memory effect (the student) that was trying to even things out because that was the way they thought it should be. In the first string, the coin didn't care what came before, because it was a truly random effect. In a short number of trials, it looks screwed up, but if you do a LOT of trials, it would be virtually identical numbers of heads and tails.
Okay, so what does that have to do with skillups in tradeskilling?
Most people have found that the skillup rate over a large range (say from 190 to 250) is around 1 in 20. In that range, there are 60 skillups. Let's say that I just skilled up a point. On the NEXT try, I have a 1 in 20 shot of getting another skillup. Since I have 60 skillups, I can EXPECT to get about 3 "double" skillups from 190 to 250. About 1 in 6 tradeskillers will get a "triple" skillup, 1 in 120 will get a "quad" skillup, and 1 in 2400 will get a "quintuple" skillup between 190 and 250. People that get these lucky streaks claim the streaks as evidence that the RNG is broken. In point of fact, the streaks are the biggest proof that the RNG is working as it is supposed to.
On the other side of the coin are the people with really long streaks without a skillup. Unfortunately, they too are sign that the RNG is working. With a 1 in 20 chance of getting a skillup, each sequence of 100 combines has a half-percent chance of giving you NO skillup. (19/20)^20. It doesn't happen to many of us, but it is almost guaranteed to happen to someone. My personal worst was from 199 to 200 in jewel crafting: 320-ish combines. For the record, the odds of that result are astronomically low, but I am ever the lucky guy.
Part of the problem with the RNG is that we remember the streaks - 3 fizzles in a row, or 3 skillups in a row get our attention, as do 320 combines without a skillup. If you average the results over all the tradeskillers in EQ, the RNG works and gives even results across all attempts. Your personal mileage may vary.
In other words, the game design may suck, but the RNG works just fine.
Comment