Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

Broken (or incorrect % chance of success) Reverse Engineering?

STAR WARS: The Old Republic > English > Crew Skills
Broken (or incorrect % chance of success) Reverse Engineering?

psandak's Avatar


psandak
06.14.2012 , 09:02 AM | #21
I recently did a LOT of crafting to help a guildmate gear up and I would say that my attempts to get both blue and purple schematics was "typical"

there were times when I made 20 of one item and did not get a schematic off RE
there were times when I made one and got a schematic right away (not necessarily the one I was hoping for though) and canceled the rest of the crafting
And there were times when the percentages worked out and I got a schematic in 5 or 10 crafts

The point is that over a much longer haul then ANYONE can do alone, the numbers will work out - 20% = 1 in 5 and 10% = 1 in 10.

Just because you crafted 100 items and did not get a schematic off RE does not mean the system is broken; it means that somewhere someone else got all their schematics in one shots.

That being said, there is something else to be considered. I remember early on when i REed something I got a notice in the chat window that I discovered a schematic but I already had it. I do not know if they took that "feature" out or whether they just hid it from view. So there is the possibility that you are discovering schematics you are just discovering the same ones you already have. It sucks, but that could explain why some find it difficult to get that last schematic.

Sothicus's Avatar


Sothicus
06.14.2012 , 01:35 PM | #22
Quote: Originally Posted by psandak View Post
I recently did a LOT of crafting to help a guildmate gear up and I would say that my attempts to get both blue and purple schematics was "typical"

there were times when I made 20 of one item and did not get a schematic off RE
there were times when I made one and got a schematic right away (not necessarily the one I was hoping for though) and canceled the rest of the crafting
And there were times when the percentages worked out and I got a schematic in 5 or 10 crafts

The point is that over a much longer haul then ANYONE can do alone, the numbers will work out - 20% = 1 in 5 and 10% = 1 in 10.

Just because you crafted 100 items and did not get a schematic off RE does not mean the system is broken; it means that somewhere someone else got all their schematics in one shots.

That being said, there is something else to be considered. I remember early on when i REed something I got a notice in the chat window that I discovered a schematic but I already had it. I do not know if they took that "feature" out or whether they just hid it from view. So there is the possibility that you are discovering schematics you are just discovering the same ones you already have. It sucks, but that could explain why some find it difficult to get that last schematic.
It is possible that we have succeeded but the learned schematic is one already learned. I didn't put that in the total list of possibilities of which BW has not told us. However, the where RE'ing 100 items and not getting a schematic is considered okay, is not within the normal 3 standard deviations for 20%. Please read the previous posts for the math. What I have been saying is that Bioware either has a broken system or a hidden component to the learning % which is not represented in the tool tip. And because it is not in the tool tip, I think BW needs to spell out the exact % chances.

Arcturys's Avatar


Arcturys
06.24.2012 , 02:25 AM | #23
The odds are roughly the same (actually slightly better) than rolling any one particular number on a 6-sided die. There is a 1 in 6 chance of rolling any one number. Over a large enough sample size, you will average 1:6 for each of the 6 numbers on the die. Which also means you should be averaging one recipe every 5 attempts (20% = 1:5) over a sufficient sample size. However, after having reverse engineered hundreds of items (probably well near a thousand by this point across all toons), I would say that the coding is broken. It is NOT properly calculating the chance of success. My average is more likely around 1:15 or 1:20, and definitely not 1:5.

Simply put, the average should equal the ratio. Roll a 6-sided die 100 times, and you should see each number show up around 16-17 times approximately (although 100 isn't really a sufficient sample size). That's a 1:6 ratio with 1:6 results. Reverse engineering is supposed to be a 1:5 ratio, but we are not seeing 1:5 results when reverse engineering, and there are far too many people reporting the same thing. So we're talking a sample size of tens of thousands of attempts when everyone's input is considered. Bioware needs to check their coding.
The Covenant of the Phoenix
We are recruiting!

steave's Avatar


steave
06.24.2012 , 06:06 AM | #24
Quote: Originally Posted by Arcturys View Post
The odds are roughly the same (actually slightly better) than rolling any one particular number on a 6-sided die. There is a 1 in 6 chance of rolling any one number. Over a large enough sample size, you will average 1:6 for each of the 6 numbers on the die. Which also means you should be averaging one recipe every 5 attempts (20% = 1:5) over a sufficient sample size. However, after having reverse engineered hundreds of items (probably well near a thousand by this point across all toons), I would say that the coding is broken. It is NOT properly calculating the chance of success. My average is more likely around 1:15 or 1:20, and definitely not 1:5.

Simply put, the average should equal the ratio. Roll a 6-sided die 100 times, and you should see each number show up around 16-17 times approximately (although 100 isn't really a sufficient sample size). That's a 1:6 ratio with 1:6 results. Reverse engineering is supposed to be a 1:5 ratio, but we are not seeing 1:5 results when reverse engineering, and there are far too many people reporting the same thing. So we're talking a sample size of tens of thousands of attempts when everyone's input is considered. Bioware needs to check their coding.
The problem is that your sample size of a few thousand include only those that are complaining because they had bad luck. Everyone else that match or exceed the expected amount of successes just keep going without complaining.
If you travel around to schools and take down everyone's ages, you're gonna have a very large sample and your sample would show that the average age is quite low, and if you go to retirement homes instead, it would show a very high average age, but despite both studies having large sample sizes neither of them are accurate.
Tl;Dr: Large sample != good sample.
For example, you claim you've had a success roughly one in 15 tries. I've had roughly 1 in 3, or in other words 5 in 15. For the 2 of us, that means 6 in 30, or 20%.

Highcommander's Avatar


Highcommander
06.24.2012 , 02:31 PM | #25
Quote: Originally Posted by steave View Post
The problem is that your sample size of a few thousand include only those that are complaining because they had bad luck. Everyone else that match or exceed the expected amount of successes just keep going without complaining.
If you travel around to schools and take down everyone's ages, you're gonna have a very large sample and your sample would show that the average age is quite low, and if you go to retirement homes instead, it would show a very high average age, but despite both studies having large sample sizes neither of them are accurate.
Tl;Dr: Large sample != good sample.
For example, you claim you've had a success roughly one in 15 tries. I've had roughly 1 in 3, or in other words 5 in 15. For the 2 of us, that means 6 in 30, or 20%.
Still doesn't explain how some people can go 30, 50 or even 100 tries in a row without success (myself included). Individual results wont always match the greater number of results in a larger sample, however they shouldn't differ so greatly.

We are not looking for a 20% chance across the board, we are looking for a 20% chance for an individual. As a crafter what does it matter to me if i never hit an RE when my best friend hits every time (making us 50% chance total)? No, the only solution is to look at individual results and see what is broken
The Ebon Hawk (Light - RP-PvE) UnderWorld Alliance - UA
------------------------------------------------------Visit our website
-------- System Alerts
-------- Server Status

Arcturys's Avatar


Arcturys
06.24.2012 , 02:35 PM | #26
I don't know how you are getting that high of a result, unless you simply don't pay attention because wasting time and money farming mats doesn't bother you like it does some of us. I think some of us are just more sensitive (yes, perhaps overly sensitive) to the poor ratios. Positive people that don't mind things like this tend to think it's going better than it really is, because they just don't care and don't really take notice. Maybe there's a benefit to that.

At any rate, there's also an issue with combining ratio with mine without any further information. If you RE'd 5 items and got a ratio of 1:3, and I RE'd 1000 items and got a ratio of 1:15, it would still be close to 1:15 figuring in both of our contributions as the addition of your sample size to mine would have a negligible effect on my results. You cannot simply average ratios that are not composed of equal variables. It is far more complex than that.



Regardless, after leveling all of these crafters and RE'ing so many items I am already sick of crafting (tedium being something else that could be addressed in regard to crafting, but I won't do that here), I can say with absolute certainty I am not getting anywhere near the 1:5 average. RNG is not "luck based" so it's really rather silly that people are using luck to discount one theory and support another. A computer possesses neither metaphysical nor spiritual properties. It is neither lucky nor unlucky.

RNG is just that - it's random. Over large samples, random still has a pattern in the results (given enough probabilistic resources over a small enough result range to witness the pattern, and a fixed pool). Chances are that, as I said, over enough attempts, the average should approach the expected ratio.Theoretically, every time we make an attempt to learn a recipe, the system should basically be "rolling" a 5-sided die - /random 5, where 1 is a success and 2-5 is a failure.

Take for example a 6-sided die. If you were to roll it 30-40 times and never once get a 6, and you could do this multiple times, you would suspect (rightly) that there was something wrong with the die. It likely has a flaw, either by accident or intention. However we have a better chance of learning a recipe than rolling a 6 on this die, yet I have seen countless equivalents with the crafting system. Right now, for example, I just passed 35 attempts (EDIT: about to attempt #39 while writing this article) trying to learn a mid-level purple recipe by reverse engineering blue items. Probability suggests that is extremely unlikely if the system is balanced, and yet I've experienced this same thing repeatedly over the course of all my crafting.

Quote: Originally Posted by Highcommander View Post
We are not looking for a 20% chance across the board, we are looking for a 20% chance for an individual.
I couldn't agree more, which is why I am using my own results solely as the substance of the opinion I just posted.
The Covenant of the Phoenix
We are recruiting!

SelinaH's Avatar


SelinaH
06.24.2012 , 02:57 PM | #27
Quote: Originally Posted by Highcommander View Post
We are not looking for a 20% chance across the board, we are looking for a 20% chance for an individual. As a crafter what does it matter to me if i never hit an RE when my best friend hits every time (making us 50% chance total)? No, the only solution is to look at individual results and see what is broken
It's not broken. It's working the way a random system works. That is to say: A few people are going to get lucky, and another bunch is going to get really screwed. That's how the distribution shakes out.

You might want to put your support behind some of the suggestions to add a streakbreaker mechanism, or to replace the current RNG system with one that tracks crafting points and guarantees a success after a certain amount of progress.
"Homogeneity breeds weakness: theoretical blind spots, stale paradigms, an echo-chamber mentality, and cults of personality. A diverse dramatis personae is a powerful tonic against these ailments." -V.S. Ramachandran

steave's Avatar


steave
06.24.2012 , 03:32 PM | #28
Quote: Originally Posted by Highcommander View Post
Still doesn't explain how some people can go 30, 50 or even 100 tries in a row without success (myself included). Individual results wont always match the greater number of results in a larger sample, however they shouldn't differ so greatly.

We are not looking for a 20% chance across the board, we are looking for a 20% chance for an individual. As a crafter what does it matter to me if i never hit an RE when my best friend hits every time (making us 50% chance total)? No, the only solution is to look at individual results and see what is broken
The chance for an individual is 20%, but the results for each individual will vary but equal out. Random systems do not provide fair results, but if you have a better idea for how to do it (AKA a system that doesn't rely on chance at all, or you'll always have cases like these), head over to the suggestion box.

Quote:
I don't know how you are getting that high of a result, unless you simply don't pay attention because wasting time and money farming mats doesn't bother you like it does some of us. I think some of us are just more sensitive (yes, perhaps overly sensitive) to the poor ratios. Positive people that don't mind things like this tend to think it's going better than it really is, because they just don't care and don't really take notice. Maybe there's a benefit to that.
The reason is quite simple, I had good luck while you had bad luck, and it's quite easy to measure. I level by producing 10 of the hardest to make items that I have access to at once (so, for example, a few each of different hilt, barrel, mod or enhancement types) before switching to a different char again and coming back a bit later to RE them all and queue an another 10 of the next level. That means there's been quite a few times when I wasted a few REs due to getting all available schematics already, and yet I have 2-3 blue items in nearly every level bracket. There is a very small amount of tiers where I have only 0-1, but those are more then made up for by the amount of tiers with 3-4...and again, that's despite the wasted REs.

Quote:
If you were to roll it 30-40 times and never once get a 6, and you could do this multiple times, you would suspect (rightly) that there was something wrong with the die.
If I was doing that, yeah, I'd think there was something wrong. However, if I knew there were thousands of other people getting the expected result with that same die, then I'd think no more of it.


You are exactly right about having us compare ratios say very little, but that also applies to you combining your result with the other people that are complaining here.

Aurojiin's Avatar


Aurojiin
06.24.2012 , 03:46 PM | #29
Quote: Originally Posted by Highcommander View Post
Still doesn't explain how some people can go 30, 50 or even 100 tries in a row without success (myself included). Individual results wont always match the greater number of results in a larger sample, however they shouldn't differ so greatly.
Actually, it explains it entirely, and your assertions are completely erroneous. This is a simple random roll system. It's entirely possible for you to experience 100 failures in a row. It's not very likely, statistically speaking, but it's absolutely possible, and entirely within the conceivable outcomes of the system.

Quote:
We are not looking for a 20% chance across the board, we are looking for a 20% chance for an individual. As a crafter what does it matter to me if i never hit an RE when my best friend hits every time (making us 50% chance total)? No, the only solution is to look at individual results and see what is broken
Except we are looking for a 20% chance across the board, because that would indicate that the system is functioning correctly. What you're asking for is not a 20% chance to RE, but a guaranteed success for every 5 RE attempts, which is an entirely different concept.

Karl-Just-Karl's Avatar


Karl-Just-Karl
06.24.2012 , 08:30 PM | #30
If the "You already know that schematic" result is simply suppressed you can make a iron clad argument that the 20% or 10% tooltips are indeed erroneous. I'd love to know whether it moves you on to a random alternate prefix or just dies there.