Everybody in this thread - "Math math math math math, I'm smart, math math math, you're wrong, math math"

The only answer that matters:

Whenever you have RNG, people "feel" like they come out on the wrong side of the equation. People being mad at your game all the time = bad. Implement a system where you have a specific # of attempts to achieve a schematic. You have to RE 5 greens to get a blue, and 10 blues to get a purple. I'll gladly trade the "joy" of getting a schem on my 3rd attempt, to rid myself of the frustration of not getting one by the 37th, and then we are still making the investment in creds/mats/time etc.

Apperently what i said has stricken close to home with sombody cause they felt hurt and reported me LOLZ.

You are thinking to much into this guys really its 20% for each individual piece you RE.

Now then, who didnt like my equation? Was the math wrong? 1x0= No chance

Reverse Reverse

Engineering Engineering

Attempts Successes

9 - 1

15 - 2

5 - 1

6 - 1

1 - 1 36 - 6 = 16.66%

5 - 0

5 - 1

5 - 0

10 - 1

19 - 5 80 - 13 = 16.25%

15 - 4

8 - 2

10 - 2

10 - 2

1 - 1 124 - 24 = 19.35% <----

24 - 7 148 - 31 = 20.94% |

10 - 0 158 - 31 = 19.62% E ===== Hmmm.... interesting....

7 - 2 165 - 33 = 20.00% |

5 - 0 170 - 33 = 19.41% |

9 - 2 169 - 35 = 20.07% <----

10 - 1

5 - 1

5 - 0

10 - 1

10 - 0 209 - 38 = 18.18%

10 - 1

10 - 1

10 - 0

4 - 1

15 - 0 258 - 41 = 15.89%

9 - 2

10 - 1

5 - 0

5 - 0

5 - 0 292 - 44 = 15.06% <--- after this point is where it fell into the outland

10 - 1 302 - 45 = 14.09%

5 - 1 307 - 46 = 14.98%

10 - 1

10 - 0

5 - 0 332 - 47 = 14.1%

2 - 1

5 - 0

10 - 0

7 - 1

13 - 2

8 - 1

15 - 2 392 - 54 = 13.77%

 

This gives a 13.4 percent mean. Using a standard confidence calculation with 99.7 percent boundary if 20 percent is the true mean as defined in the tool tip the average for the sample above should be between 14.3 to 25.7 percent. With the mean sample is out of the 99.7 boundary that mean that it is almost impossible that the 20 percent is the true rate of getting a new plan.

1) If your next 169 results played out like your first 169 you'd be looking at a 561 - 89 = 15.86%..... which is within your boundary.

2) If you were to add another sample this size that were a similar outlier but full of successes (say 392 - 98 = 25%)

that would give you.....784 - 152 = 19.38%

3) 0.03% is not anywhere close to "nearly impossible" for a sample this small out of the millions of RE checks that happen over the timeframe that you collected this sample.

4) This is a sample size problem. And that is illustrated by how rapidly your % successes is still changing over half way through the sample. 3 additional succeses would have brought your results within your boundary.

You do not need 10,000 data points. 400 is more than enough.

Whenever someone mentions random number, you'll get all these people with no understanding of probability coming out and say 'you need a bigger sample size'. You don't. The bigger issue however is that 20% is the mean, it doesn't tell you anything about the "distribution". Suppose the 'random number generator' goes like this : fail for the first 800 tries, and success for the next 200. The mean is still 20%, however it's not evenly distributed over time.

According to the data the OP has gotten, there is a 99.7% confidence interval only applies if it's a random even distribution. Most programs, when you use the default 'random' function, suffers from this. This is why when you run something in Python or Excel, you will gets a lot more 'streaks' compared to what the theory indicates.

True randomness is hard to do, most implementation uses 'psuedorandom'. For instance, you can start off with 1000 numbers in a box, scramble them up, and then pick one of them out until no number is left. What number you get will then depend on what number has already been picked out, and you're prone to more 'streaks' because of it.

You do not need 10,000 data points. 400 is more than enough.

 

..

400 data points is not enough.

I don't seem to be able to communicate my point, so I've just pulled out my dice roll modeling program to try and help (2 six-sided dice).

In the real world, the odds of rolling a seven are 16.6%. The odds of rolling a twelve are 2.77%. For this test, I ran 4 separate sets of 400 iterations each.

Set 1:

Hit seven 20% of the time

Hit twelve 1.5% of the time

Set 2:

Hit seven 16% of the time

Hit twelve 4.25% of the time

Set 3:

Hit seven 13% of the time

Hit twelve 2.5% of the time

Set 4:

Hit seven 20.5% of the time

Hit twelve 1% of the time

With only 400 data points per set, are you able to tell if my RNG program is written correctly or not?

400 data points is not enough.

 

I don't seem to be able to communicate my point, so I've just pulled out my dice roll modeling program to try and help (2 six-sided dice).

 

In the real world, the odds of rolling a seven are 16.6%. The odds of rolling a twelve are 2.77%. For this test, I ran 4 separate sets of 400 iterations each.

 

Set 1:

Hit seven 20% of the time

Hit twelve 1.5% of the time

 

Set 2:

Hit seven 16% of the time

Hit twelve 4.25% of the time

 

Set 3:

Hit seven 13% of the time

Hit twelve 2.5% of the time

 

Set 4:

Hit seven 20.5% of the time

Hit twelve 1% of the time

 

With only 400 data points per set, are you able to tell if my RNG program is written correctly or not?

^^^^^^^

400 isn't enough. You need +2k to get an accurate result.

You do not need 10,000 data points. 400 is more than enough.

 

Whenever someone mentions random number, you'll get all these people with no understanding of probability coming out and say 'you need a bigger sample size'. You don't. The bigger issue however is that 20% is the mean, it doesn't tell you anything about the "distribution". Suppose the 'random number generator' goes like this : fail for the first 800 tries, and success for the next 200. The mean is still 20%, however it's not evenly distributed over time.

 

According to the data the OP has gotten, there is a 99.7% confidence interval only applies if it's a random even distribution. Most programs, when you use the default 'random' function, suffers from this. This is why when you run something in Python or Excel, you will gets a lot more 'streaks' compared to what the theory indicates.

 

True randomness is hard to do, most implementation uses 'psuedorandom'. For instance, you can start off with 1000 numbers in a box, scramble them up, and then pick one of them out until no number is left. What number you get will then depend on what number has already been picked out, and you're prone to more 'streaks' because of it.

Your last paragraph is not relevant to the argument at hand. RE is more like putting numbers 1 through 100 into a hat, shaking it up, drawing a number, annotating that result, putting the number back in the hat, shaking it again, drawing another number, annotating that result and repeating again and again and again for X number of attempts. No single attempt affects the result of any other attempt. Every time you draw a number the chance of drawing a given range of numbers (say 1-20 ) remains constant.

That being said, you are correct that no random number generator is ever 100% truly random, but that is only because computers cannot truly randomize numbers. Computerized RNGs are biased but the only way to truly find the bias is to run millions upon millions of attempts...400 attempts does not even come close.

I love how every time someone starts discussing the statistics, people trot out the same arguments about the invalidity of observation.

Personally, I have nothing to say about the statistics (I'm terrible at math), but I love the logic that any reasoning made based off of observable data cannot be valid, because that data is just a subset of a larger pool of unobservable data. The beauty of it is that it's typically accompanied by the argument that 20% means that each instance is at 20%, and multiple attempts cannot change the outcome because the effect of the attempts are not cumulative... except apparently when they are, across whatever the maximum number of attempts is.

It's a hilarious logical fallacy that doesn't really involve math at all, but comes up every single time as if it were some secret counter-intuitive knowledge of statistics proudly trotted out for the occasion. It's a scientifically framed argument that is utterly anti-science, since it requires faith in the presented prediction because it argues that the data can only be understood based on unknowable quantities whose existence is only framed by the presented prediction, and therefore must be presumed to support it.

@Lodril, Quite the philosophical analysis of RE threads.

Dare I ask, do you hold a viewpoint of:

1. The RE tooltip percentage (20%) is probably accurate.

2. The RE tooltip percentage (20%) is probably incorrect.

3. Don't care poking fun at forum posters.

400 data points is not enough.

 

I don't seem to be able to communicate my point, so I've just pulled out my dice roll modeling program to try and help (2 six-sided dice).

 

In the real world, the odds of rolling a seven are 16.6%. The odds of rolling a twelve are 2.77%. For this test, I ran 4 separate sets of 400 iterations each.

 

Set 1:

Hit seven 20% of the time

Hit twelve 1.5% of the time

 

Set 2:

Hit seven 16% of the time

Hit twelve 4.25% of the time

 

Set 3:

Hit seven 13% of the time

Hit twelve 2.5% of the time

 

Set 4:

Hit seven 20.5% of the time

Hit twelve 1% of the time

 

With only 400 data points per set, are you able to tell if my RNG program is written correctly or not?

Once again i will state that your dealing with a joint distribution where each distribution in very different from a binomial one we are talking about?

I generated ten random series of a thousand samples each with a simple Python script, and here's the result:

 

http://snag.gy/JPFI4.jpg

 

Apologies for the horrible colors. Notice how one of the lines is slightly below the 99.7% confidence interval for a while, before eventually climbing back up? Exactly like you are seeing in your experiment.

 

While 99.7% may seem like an impressive number, it only covers 997 of every 1000 cases. The remaining three fall outside it. And there are many thousands of players on each server; maybe even tens of thousands. So it's far from impossible for someone to have such a streak of bad luck.

I am working on coming up with a more stringent test but the statistics expert at work doesn't think there would ever be a need for it, but in theory it is possible.

We know exactly what standard deviation is. In fact, we know everything about this distribution (it's what is called a normal distribution) because this is the classic high school binomial distribution problem. But in fact, even if it was not so simple, a lot of other scenarios (almost all that anyone cares about) converge to the same thing. Read up on the central limit theorem if you want to know why.

 

By the way, here are my results so far: 74 tried, 15 successes.

thank you I will add your data to what i have and post an update

Regardless of how the math is done - I think the RE numbers need a boost say 30% for green to blue and 15% from blue to purple. It's bad enough that time, money and ,material is wasted on green to blue for a presumably negligible market.

Oh well - it's time they did a major patch revision focusing on crafting/RE'ing.

Regardless of how the math is done - I think the RE numbers need a boost say 30% for green to blue and 15% from blue to purple. It's bad enough that time, money and ,material is wasted on green to blue for a presumably negligible market.

 

Oh well - it's time they did a major patch revision focusing on crafting/RE'ing.

That will create a huge influx of new items to the market decreasing the price of purple items. I see people complain all the time about crafting skills not making a lot of money, all you're doing with this idea is making the profit margin a lot smaller.

That will create a huge influx of new items to the market decreasing the price of purple items. I see people complain all the time about crafting skills not making a lot of money, all you're doing with this idea is making the profit margin a lot smaller.

Point taken. If it's just profit on purple then perhaps leaving it where it is for that. I still think the green to blue rate needs tweaking even if, unfortunately, profits on those items would drop.

Once again i will state that your dealing with a joint distribution where each distribution in very different from a binomial one we are talking about?

Absolutely. I wasn't trying to indicate that my rng test is apples to apples as the RE situation in crafting. It was more to indicate that my software could either be well written or poorly written, you just can't tell from only 400 data points.

I imagine that by now you've collected more data. Has there been any shift?

400 data points is not enough.

 

I don't seem to be able to communicate my point [...]

Actually you have made your point pretty well.

it is just that unfortunately many people have no interest in a rational explanaition (in this case based on actual math) and prefer to rather listen to their vague "feelings", how wrong they may actually be.

@Lodril and the OP for that matter.

I have been tracking my own RE attempts for the last couple of months. I track the overall number of attempts and the overall number of schematics gained. Across three characters, in three different crafting skills I have done 213 20% REs and 178 10% REs (ironically or not that is about the same number of total REs the OP has done). My results are as follows: 48 20% RE schematics gained (that's 22.54% success rate) and 18 10% RE schematics gained (that's 10.11% success rate). My data shows that the system is working better than expected.

Which one of us is right? the OP or me? I'll tell you...neither. it is far more likely that the true nature of the system falls somewhere in between. Right where it is supposed to be :eek:

FACT: Sample size is an important aspect of statistics. The larger the sample the more accurate the data. 400 REs from one player on one character is an insignificant sample size when compared to the hundreds of thousands of players performing millions, if not tens of millions of REs across all crafting crew skills done over the last year in this game.

This is why we who rail against the complainers use the small sample size argument so often.

It's 20% per RE.

When you RE Item 1-1, there's a 20% you will learn schematic Adv Item 1.

When you RE Item 1-2, there's a 20% you will learn schematic Adv Item 1.

When you RE Item 1-3, there's a 20% you will learn schematic Adv Item 1.

etc.

etc.

etc.

Not

When you RE Item 1-1, there's a 20% you will learn schematic Adv Item 1.

When you RE Item 1-2, there's a 40% you will learn schematic Adv Item 1.

When you RE Item 1-3, there's a 60% you will learn schematic Adv Item 1.

When you RE Item 1-4, there's a 80% you will learn schematic Adv Item 1.

When you RE Item 1-5, there's a 100% you will learn schematic Adv Item 1.

It's 20% per RE.

 

When you RE Item 1-1, there's a 20% you will learn schematic Adv Item 1.

When you RE Item 1-2, there's a 20% you will learn schematic Adv Item 1.

When you RE Item 1-3, there's a 20% you will learn schematic Adv Item 1.

etc.

etc.

etc.

 

Not

 

When you RE Item 1-1, there's a 20% you will learn schematic Adv Item 1.

When you RE Item 1-2, there's a 40% you will learn schematic Adv Item 1.

When you RE Item 1-3, there's a 60% you will learn schematic Adv Item 1.

When you RE Item 1-4, there's a 80% you will learn schematic Adv Item 1.

When you RE Item 1-5, there's a 100% you will learn schematic Adv Item 1.

But think of it like rolling a 10-sided dice, where 2 certain numbers will give you a good result, the more you roll, the higher the probability you come across one of the two numbers you want.

But think of it like rolling a 10-sided dice, where 2 certain numbers will give you a good result, the more you roll, the higher the probability you come across one of the two numbers you want.

Yes, you are correct that the more times you try, the less likely you are to continue to fail. However, that probability has no direct impact on success or failure of any given attempt; it is based on theoretical statistical analysis.

Chance of success for every attempt is 20%

Chance of eight failures in a row is 16.78% (80%^8)

So theoretically you have a better chance to succeed once than to fail eight times in a row.

Does that impact your chance of success after seven consecutive failures? No. Your chance of success is still 20%.

But think of it like rolling a 10-sided dice, where 2 certain numbers will give you a good result, the more you roll, the higher the probability you come across one of the two numbers you want.

This argument is bordering on the "Gambler's Fallacy" Here is a halfway decent explanation of that concept: http://en.wikipedia.org/wiki/Gambler's_fallacy

Note that people with the Gambler's Fallacy make casinos very happy.