Statistical anomaly in Reverse Engineering

[Grayseven]

Curious if any other Cybertech is noticing a large statistical difference between the reported 20% success chance of reverse engineering and their actual RE successes.

 

After crafting large amounts of low level mods and armor for the guild, I noticed that reverse engineering on my Cybertech seemed harder than on my Artificer when RE'ing enhancements.

 

I started tracking my reverse engineering successes versus the attempts and I've seen that the actual RE chance is falling just below 10% on 140 attempts.

 

I'm expanding this to include all reverse engineering done by my Cybertech and plan on taking an alt up through Cybertech to get better data, but I'd like to know if other players are experiencing this.

Edited August 30, 2013 by Grayseven

[kaelthy]

remember its 20% chance per time you do it, so basically the RNG god was not on your side, thats pretty much it

[Reinae]

Actually I agree that the RE rate appears to have reduced in recent patches - Having taken all crew skills to max on one server and starting again on another, it is taking far more attempts to get RE successes on Cybertech than other crew skills - and yes I appreciate that it is a percentage chance for each attempt, but the numerous times its taking up to 40-60 RE attempts to get a result in comparison to those very few taking 5 or less makes the statistics simply look out of kilter with what it should be

[morfius]

Green to Blue is 20%, but Blue to Purple is 10% if I remember correctly.

Have discovered in 1st try, have up to 15 with no luck.

RNG...

[Grayseven]

RNG is only important in small sample sizes. If I were to just RE 10 and get a less than 10% success rate, I could easily equate it to RNG.

 

But the larger the sample size, the closer to the purported success chance you should get and this just isn't the case. As I RE more and more low level mods and armoring, I am seeing the success rate hover around 10%.

 

As I said, I'll know more when I push another character up through Cybertech, but from what I'm seeing the tooltip success rate of 20% (regardless of premium or prototype, they all say 20%) isn't accurate.

[Deathstalker_]

Isn't the Rule of Large Numbers only applicable when you have a massive number of data points? As in, thousands or more? Because I somehow doubt that you're going to magically hit the average every time in a sample size of 140.

[psandak]

Isn't the Rule of Large Numbers only applicable when you have a massive number of data points? As in, thousands or more? Because I somehow doubt that you're going to magically hit the average every time in a sample size of 140.

 

I am of two minds on this:

 

#1 - 140 is too small a sample for the Rule of Large Numbers to apply

#2 - 140 is a larger sample than normally posted here so I am inclined to give what the OP states more credit than normal.

 

Assuming a success rate of 9.5% (OP only said "just below 10%") the average number of attempts per success is 13.3. The probability of getting 12.3 failures in a row is 6.43% (assuming even distribution; even though my motto is random =/= even ), very far from statistically impossible (more than 1 in 20 experience this type of streak).

 

My point is that with a 20% chance of success comes an 80% chance of failure and too many do not take that aspect into account.

[Grayseven]

The sample size is getting larger with only minor fluctuations in the success rate. It currently hovers around 9.6%.

 

My suspicions on Cybertech evolved only because I craft and RE on multiple characters and it is glaringly obvious that Cybertech is "behind the curve" when it comes to REing.

 

I expect streaks. I've seen them before on multiple toons. When I'm REing gear for a low level character and companions, Cybertech is always the last craft to complete the iterations to Artifact quality schematics. Cybertech eats the most resources even though it often has the lowest resource cost for the level of gear I am crafting. It was that anecdotal evidence that caused me to start tracking RE success for my Cybertech.

 

At one point, many of the Prototype Cybertech items had a 10% chance of success when every other crafting profession had a 20% chance across the board. The tooltips for Cybertech say 20% now, but more and more I'm seeing actual evidence that it is still 10%.

 

I urge others to begin tracking their Cybertech reverse engineering in order to see if it's just me, or if there really is something going on.

[magecutter]

just my two cents,

Last night I Re'ed the four 53 armors I wanted from blue to purple.

3 comps making 1 of each type (skill, resolve, agile, might) and each time 1 comp hit for a crit and on the might armor all 3 comps hit a crit.

 

also, on Re'ing each type I got the purple schematic within the 4 available tries, (within the 6 I had for might).

 

to get the blues from greens I must have Re'ed all four 30 or 40 times.

 

my point is, it's a total crap shoot, in that the RNG and 20% for each new attempt seems to be in effect.

I know almost nothing about statistics. but I do see runs of bad luck and sometimes runs of incredibly good luck (like last night) which makes me think that it all balances out in the end.

 

btw - just read your last post OP and the 10% vs 20% question is something I will pay closer attention to.

Edited August 30, 2013 by magecutter

[mike_is_valid]

well the RNG rules were there for long time and i think they were more or less true, but now i have the feeling aswell that RE is somehow a lot more difficult.