Subscriber research chance bug

[DarkJuliaArcher]

After some thorough testing, I have discovered that there is not a 20% chance for subscribers to learn a new schematic, which is one of the subscriber perks. I tried reverse engineering 1000 times now over the course of the last week or so, always using items I have a chance to gain research in, and here are my findings: out of 1000 attempts, I successfully researched 103 times.

 

As such, it's very conclusively clear that the research chance is still 10% for subscribers rather than 20%. Hope this gets fixed in the patch.

[dlcman]

That's not the way it works. You get a 20% chance per RE. That is not the same as 200 schematics per 1000 things RE'd.

[DarkJuliaArcher]

While you're partially right, it means exactly the same thing I posted. The experiment proves definitively that it's not 20%. That's statistical calculation, math. The odds of me only getting 10% off 20%, over 1000 tries, is infinitessimally small and not even worth considering (talking billionths of a percent here, my odds of winning the lottery are better). Had I only done 10 or 50 you might have a point, but over 1000, the odds should even out and, statistically speaking, the number of successful schematic learnings should fall near 20% if the chance is actually 20%. That means you technically should get at least near 200 schematics per 1000 attempts. The fact that it didn't even come remotely close to 200 is the proof. Edited February 9, 2015 by DarkJuliaArcher

[Maliceen]

Your sample size is too small to be meaningful.

[Nireos]

While you're partially right, it means exactly the same thing I posted. The experiment proves definitively that it's not 20%. That's statistical calculation, math. The odds of me only getting 10% off 20%, over 1000 tries, is infinitessimally small and not even worth considering (talking billionths of a percent here, my odds of winning the lottery are better). Had I only done 10 or 50 you might have a point, but over 1000, the odds should even out and, statistically speaking, the number of successful schematic learnings should fall near 20% if the chance is actually 20%. That means you technically should get at least near 200 schematics per 1000 attempts. The fact that it didn't even come remotely close to 200 is the proof.

 

But this isn't statistical math. This is probability math. Each time you RE, you have exactly the same chance (20%). It's like rolling the dice. This is why the odds are in favor of the house when you gamble. People think statistically that they should win if they keep play; but you don't improve your odds with each roll.

 

Same with buying a lottery ticket. Each ticket you buy does not increase your odds of winning. It is PROBABILITY not statistics. End of story.

[MeNaCe-NZ]

I don't see why people are arguing against his experiment without doing one of their own.

 

I would think if you see a 10% rate after 1000 tries then you have fairly solid grounding for saying that yes it may indeed be bugged and is indeed coming up at 10%. I get the whole RNG argument and people just love to argue semantics over this issue but unless they are going to provide a similar sample with a very different result then I personally believe you have made a fair point if you've gone through and done your sample correctly.

 

Out of curiosity what were you reverse engineering? Aren't there some items that are only 10% even for subs? Your sample wasn't skewed by RE'ing anything you had already learned a schematic for by mistake?

[Nireos]

I don't see why people are arguing against his experiment without doing one of their own.

 

I would think if you see a 10% rate after 1000 tries then you have fairly solid grounding for saying that yes it may indeed be bugged and is indeed coming up at 10%. I get the whole RNG argument and people just love to argue semantics over this issue but unless they are going to provide a similar sample with a very different result then I personally believe you have made a fair point if you've gone through and done your sample correctly.

 

Out of curiosity what were you reverse engineering? Aren't there some items that are only 10% even for subs? Your sample wasn't skewed by RE'ing anything you had already learned a schematic for by mistake?

 

Did you read what I posted? It is not statistics....it is probability.

 

Example:

Take a 6 sided die. Now let's make a few assumptions like the die is properly balanced, the rolling surface is smooth and level, ect. Roll that die and you have a 1 in 6 chance of getting any number. No mater how many times you roll that die, you have a 1 in 6 chance. Just because you may roll a 1, more times than any other number does not change the odds.

 

RE is exactly that. Every time you RE, you have a 20% chance. You cannot add the rolls together and say the percentage is off. Probabilities cannot be statistics; however statistics can give you probabilities.

Edited February 9, 2015 by Nireos

[Maliceen]

I don't see why people are arguing against his experiment without doing one of their own.

 

No single player could possibly generate enough data to prove anything. His results were very UNLIKELY, but not Impossible. The odds of any individual winning the lottery is incredibly low, but there are winners.

[MeNaCe-NZ]

Did you read what I posted? It is not statistics....it is probability.

 

Example:

Take a 6 sided die. Now let's make a few assumptions like the die is properly balanced, the rolling surface is smooth and level, ect. Roll that die and you have a 1 in 6 chance of getting any number. No mater how many times you roll that die, you have a 1 in 6 chance. Just because you may roll a 1, more times than any other number does not change the odds.

 

RE is exactly that. Every time you RE, you have a 20% chance. You cannot add the rolls together and say the percentage is off. Probabilities cannot be statistics; however statistics can give you probabilities.

 

But you can apply statistics to probability as is the case in the test sample.

 

He ran a set of tests in which the result came in at roughly 10% ( the statistics of his testing the probability ) , the same rate that is set for non subscribers. The game states a 20% chance to RE successfully for subscribers.

 

Whilst I agree RNG dictates his results may not reflect his success rate is definitively correct due to probability/RNG, the statistics of his testing indicate there may very well be an issue and it warrants looking into.

 

Even in your dice example you would expect over 1000 throws to roll around 160 1's and whilst nothing promises this return if you were rolling a dice and rolled say only 60 1's in 1000 throws I would then question the dice etc. as that is half of the probability that is expected from 1000 dice throws - not slightly off, half, quite a big difference.

[Nireos]

But you can apply statistics to probability as is the case in the test sample.

 

He ran a set of tests in which the result came in at roughly 10% ( the statistics of his testing the probability ) , the same rate that is set for non subscribers. The game states a 20% chance to RE successfully for subscribers.

 

Whilst I agree RNG dictates his results may not reflect his success rate is definitively correct due to probability/RNG, the statistics of his testing indicate there may very well be an issue and it warrants looking into.

 

Even in your dice example you would expect over 1000 throws to roll around 160 1's and whilst nothing promises this return if you were rolling a dice and rolled say only 60 1's in 1000 throws I would then question the dice etc. as that is half of the probability that is expected from 1000 dice throws - not slightly off, half, quite a big difference.

 

You CANNOT apply Statistics to Probability. you can apply Probability to Stats. As in his example, he could say that 10% of the time he is likely to hit the 20% probability.

 

Again, this is why in Gambling the house wins. Because you people think that statistics should apply to the rules of probability and they do not. Consult any mathematics teacher/professor and they will tell you exactly what I am telling you. THe two things are mutually exclusive of the other.

[MeNaCe-NZ]

You CANNOT apply Statistics to Probability. you can apply Probability to Stats. As in his example, he could say that 10% of the time he is likely to hit the 20% probability.

 

Again, this is why in Gambling the house wins. Because you people think that statistics should apply to the rules of probability and they do not. Consult any mathematics teacher/professor and they will tell you exactly what I am telling you. THe two things are mutually exclusive of the other.

 

I'm not sure if you're just explaining things poorly or what or maybe it's semantics over definitions of terms?

 

He ran a sample test. He had set results and those are his statistics. From those statistics we can deduce probability.

 

103 RE's form 1000 attempts are statistics as I see it. There is no probability involved in that what so ever as that are HIS RESULTS.

 

If you then wanted to deduce probability based on his results you could then say the probability ( after rounding his result ) is 100 in 10000 or 1 in 10 or 10%.

 

The 20% is a number a text box and communicated by bioware. His analysis is simply stating that Bioware may be wrong and it may be bugged yet everyone is jumping down his throat over RNG etc. without running THEIR OWN subset of data or compare with and simply assuming the number in the text box is fine, it may not be. Just like you weren't supposed to be able to pick up loot from a boss without beat it at the end of an operation.

[DarkJuliaArcher]

You CANNOT apply Statistics to Probability. you can apply Probability to Stats. As in his example, he could say that 10% of the time he is likely to hit the 20% probability.

 

I think perhaps you don't understand how this works. Suffice it to say, you are incorrect. 1000 is a meaningful enough sample size (I decided to go with 1000 instead of 100 specifically to avoid smart-alecks who like to throw that around, but alas), and if there is a 20% probability of something happening, you should absolutely see that result (within a reasonable margin of error, +/-3-7%, but certainly not 50%) statistically when testing. The fact that it didn't even come close is the big issue here.

 

Out of curiosity what were you reverse engineering? Aren't there some items that are only 10% even for subs? Your sample wasn't skewed by RE'ing anything you had already learned a schematic for by mistake?

 

I made sure to check that. I used the three core items my toons can craft, Hilts, Armoring, and Barrels (as well as a few dozen Mods). These all state when you try to DE that the chance is 20% from green to blue and 20% from blue to purple. I marked down each time I DE'd an item, and then marked each success, stopping any further production of the item upon successful research and then moving to the next.

 

Even if the listed chance for blues is wrong on the tooltip, the data shouldn't come in at 10%, as I would always go by making and DE'ing greens and then their respective blues. If even half of the DE's were green, and only green is 20%, the test should have shown near 15%.

[dlcman]

The way it works is an rng on each re. There is a chance that you could get 0 schematics from your 1000 re's. It's not a big chance but it does exist that it could happen. But you will believe what you want and this will never be answered by a dev or even a cs.

[Nireos]

What you guys are failing to understand is that statistics and probability are two different things and you cannot determine that a probability (or chance) is wrong based on a statistic. Let me try to simplify this even further for you....

 

Take a coin (2 side) and flip it. You have a 50% chance of heads and a 50% chance of tails.

 

Flip this coin 100 times. If it lands on heads all 100 times, you cannot say the coin is broken. If you flip it the 101st time, it still has a 50% chance of tails. All you can say is that statistically, you landed on heads 100% of the time you flipped it.

 

So out of the 1000 tests of RE, he succeeded a % of the time. But each RE still has a 20% chance of success and you cannot equate all 1000 tries together to determine that % of chance of success is broken.

 

I cannot make it any simpler than that.

[eldefail]

OP, I think it's just you. I haven't sat down and calculated my own research rate from REing things, but it seems to me that it happens all the time. Quite the reverse of your own observation.

 

Still, probability aside, 10% out of 1000 is interesting. Maybe instead of arguing we should all go buy some cheap crap off the GTN to RE and see if we all have similar numbers. One player simply isn't a large enough sample to determine if there's a bug, it's something many players need to experience and provide feedback. And here's mine: it seems to be working fine. /shrug

[Nireos]

OP, I think it's just you. I haven't sat down and calculated my own research rate from REing things, but it seems to me that it happens all the time. Quite the reverse of your own observation.

 

Still, probability aside, 10% out of 1000 is interesting. Maybe instead of arguing we should all go buy some cheap crap off the GTN to RE and see if we all have similar numbers. One player simply isn't a large enough sample to determine if there's a bug, it's something many players need to experience and provide feedback. And here's mine: it seems to be working fine. /shrug

 

This again proves nothing.

 

Probability : a ratio expressing the chances that a certain event will occur (ie a 20% chance to RE an item)

 

Statistics : collection, calculation, description, manipulation, and interpretation of the mathematical attributes of large sets or populations (ie 10% success on 1000 tries to RE an item)

[Maliceen]

1000 is a meaningful enough sample size (I decided to go with 1000 instead of 100 specifically to avoid smart-alecks who like to throw that around, but alas), and if there is a 20% probability of something happening, you should absolutely see that result (within a reasonable margin of error, +/-3-7%, but certainly not 50%) statistically when testing. The fact that it didn't even come close is the big issue here.

 

While your results are extremely unlikely, they are NOT impossible. You might be right that the 20% stated rate is incorrect, or you might have just won the anti-lottery. Your test results are good enough that Bioware should double-check their programming, but it doesn't PROVE anything.

[MeNaCe-NZ]

Agreed it doesn't prove anything as there is always the chance for less than expected results ( however slim they may be ) and it does indeed warrant investigation.

 

I am happy to take the 1000 results at the OPs word and not test it myself though it would be an interesting thing to try and test had I the time or patience.

 

As to the coin example ... probabiltiy of 1 coin flip coming up heads or tails is 50% but what are the odds of every coin flip for 1000 flips coming up heads? It's certainly not 50%. Whilst it's not IMPOSSIBLE, it's VERY IMPROBABLE in fact I would be curious if anyone in the history of humanity has managed to flip a perfectly accurate, unaltered legitimate coin 100 times and got the same result 100 times in a row let alone 1000.

 

Saying "it could happen" really detracts from the point of this thread.

[AshlaBoga]

No single player could possibly generate enough data to prove anything. His results were very UNLIKELY, but not Impossible. The odds of any individual winning the lottery is incredibly low, but there are winners.

 

You put that very well Maliceen.

 

For example, I just RE'd 10 different items and it took me 43 tries to learn 10 recipes.

 

Does that mean that my Research Chance is 23%? Heck no, although I wish that were the case:p

Edited February 11, 2015 by AshlaBoga