I picked up slicing before 1.3. And before 1.3 like a previous poster I also had a roughly 1 in 5 ratio. However, now I am seeing much lower than that. Usually, I have about 1 stack of 99 blues and 3-5 purple. Or if I am unlucky 140 blues w/o any purple at all (as is right now). That is with all maxed companions and 1 (SCORPIO) has a +2% crit to slicing (20% and 22% respectively for success rate). Yeah, I'd definately say there's something weird about slicing at the moment.

Also, more anectdotal, I have the direct comparison to pretty much all mission skills except Investigation. Take Underworld Trading for example, also with maxed affection companions and 1 companion has +2% crit on UT.

I have no problem at all roughly approaching the 1 in 5 ratio with Mando Iron like I did before 1.3 with slicing. Yes sometimes I am a bit unlucky and don't get Iron for a couple missions but its usually restrained to a 1 digit number of tries.

The original post has been updated. Total count is below.

Can someone more knowledgable than me compute the expected variance over this sample?

Updated: 7/21/2012

Total missions: 752

Critical success rate: 17.2%

Failure rate: 3.2%

Well, according to Mathematica, the variance of a binomial distribution with n = 752 (number of trials) and p = 0.2 (probability 20%) is n*p*(1-p) which is 120.32.

On a side node, I was up to 210 blue parts before I got a crit again. Now if you stipulate that you get 4 blue parts per yellow mission (of which there are 2) and 2 blue parts per green mission (of which there is 1) you get 10 blue parts per cycle of sending out companions.

This means I ran this cycle 21 times for a total of 63 missions before I got a crit.

Or broken down by probabilty (I have 2 maxed affection companions w/o any crit bonus to slicing and 1 companion with a +2% crit bonus (SCORPIO)): 21 missions with a crit probabilty of 0.22 (22%) and 42 mission with a crit probabilty of 0.2 (20%).

On the other hand, when sending out my companions for UT missions in search of orange schematics, I get a steady supply of mando irons, promethium, etc every 1 digit number of missions.

If I were to guess (e.g. purely subjective w/o empirical data, totally ignoring the law of large numbers yada yada etc.), I would guess that the base chance of slicing missions is slightly lower, maybe by a few percent.

Oh my...

the updated numbers for my Slicer:

14 crits in 135 missions, that's roughly 10,4%! (btw I got Scorpio with her +2%)

I guess it's just bad luck - but in the meantime I accumulated 98 tier four purples with just 2 companions running them?!?

Tracked since Friday July 20:

Advanced Neural Augmentors

Devoted to Duty / Whispers on the Net: 71 run, 10 crit = 14%

Mysterious Funds: 71 run, 14 crit = 20%

Lockbox + Mission | Schematic

Azure Databanks / Data Race: 21 run, 6 crit = 29%

Taking Back Control: 47 run, 12 crit = 26%

Finding Our Way / Fly on the Wall: 34 run, 5 crit = 15%

All run with pets with 10k affection, some with +Crit to slicing.

I do think it's a good idea to track these sorts of success/fail rates. There are a few problems, though. First of all, if you want to demonstrate that something happened at a particular point -- say, publish 1.3 -- you would need to have uncontaminated sets of data from before and after that point in order to be able to make any independent comparisons of the rates with what is expected. Second, you'd need tens of thousands of measurements for any reasonable expectation that your rates were converging fairly tightly to what your chances predicted.

So, unless someone out there has a very large set of data pre-1.3 that does not have post-1.3 data mixed in, the point is sorta moot right now. Not that it's not important; it's just not something that can be examined.

It also obscures what may be the true problem here. Should we be arguing about whether this stuff is broken or wondering about whether the whole system is a good idea in the first place?

So, unless someone out there has a very large set of data pre-1.3 that does not have post-1.3 data mixed in, the point is sorta moot right now. Not that it's not important; it's just not something that can be examined.

All of what you say is true, ofc. I think though what I quoted is probably the reason many of us are frustrated at the moment or at the very least are starting to wonder. As some people claim it was working fine pre 1.3 I don't think most even had the need to track this stuff to try and prove a point. I certainly didn't.

So, in absence of empirical sets of data pre 1.3 to compare with, are there other things we can compare to draw some conclusion? Can we compare the success rates of different mission skills (say UT) with the same companion boni and mission difficulty types to the success rates of slicing missions on the same character.

If the system is the same and the boni are the same, shouldn't the distribution of the outcome be close to one another (within acceptable deviation limits ofc).

I dunno, it's been a long time since I had probability theory in school (if you couldn't tell ). It just seems odd to me that there's just one mission skill that is so far off than the others in success rate.

Then again, I did notice a slight surge of Biochem threads about Diplomacy crit success rates being borked, so maybe it is all one big coincidence after all...

Well, according to Mathematica, the variance of a binomial distribution with n = 752 (number of trials) and p = 0.2 (probability 20%) is n*p*(1-p) which is 120.32.

Thanks for the concise answer. Unfortunately the question may have been off since I don't know what 120.32 represents. Let me try another question

Given a number of experiments with a fixed probability of success, what is the likelihood for a given error? In this case (OP to be updated later), 837 experiments with a 20% expected success rate returned 17.1%. What would the probability be that we would be under by 2.9% ?

I do think it's a good idea to track these sorts of success/fail rates. There are a few problems, though. First of all, if you want to demonstrate that something happened at a particular point -- say, publish 1.3 -- you would need to have uncontaminated sets of data from before and after that point in order to be able to make any independent comparisons of the rates with what is expected. Second, you'd need tens of thousands of measurements for any reasonable expectation that your rates were converging fairly tightly to what your chances predicted.

I understand where you are coming from, but the reference chosen was not the 1.3 event, but the dev post stating our expected crit rate should be 20%. Given that, shouldn't we be able to prove a discrepancy with much less than tens of thousands of tests?

Good data, DarthTHC, please keep it up !

By the way, I'm not crafting for this post, I'm posting for the craft

Thanks for the concise answer. Unfortunately the question may have been off since I don't know what 120.32 represents. Let me try another question

 

Given a number of experiments with a fixed probability of success, what is the likelihood for a given error? In this case (OP to be updated later), 837 experiments with a 20% expected success rate returned 17.1%. What would the probability be that we would be under by 2.9% ?

Keep in mind that its been a while since I had this stuff in school. Having said that, it seems what you are looking for is the standard deviation of the distribution. This can be calculated from the variance very easily since it is just the square root of the variance.

First off, let "n" be the number of trials, n = 837.

Let "p" be the probabilty of success, p = 0.2.

Let "q" be the probabilty of failure, q = (1-p) = 0.8.

Let "mu" be the expected number of successes for n amount of trials (also called the mean of the distribution), mu = n*p => 167.4

Let "sigma" be the standard deviation for the distribution, sigma = sqrt(n*p*q) = 11.5724.

So for 837 trials of a binomial probability distribution (binomial meaning the outcome can only be "success" or "no success" while the probabilty stays the same for each trial and all trials are independent of each other) you should have roughly 167 successes +/- 12 successes standard deviation to be within the bounds of expected outcome.

I understand where you are coming from, but the reference chosen was not the 1.3 event, but the dev post stating our expected crit rate should be 20%. Given that, shouldn't we be able to prove a discrepancy with much less than tens of thousands of tests?

I picked 1.3 as an example. The event doesn't matter; what matters is keeping the pre- and post-event data sets independent. And sadly, no, it doesn't matter how recent or distant the event is, the law of large numbers is what it is. Theoretically, you expect convergence as the number of trials approaches infinity. Reasonably, you're looking at thousands of events, preferably tens of thousands of events. If there is any suspicion that something non-random is going on, then without proof if it's (1) which companion you use, (2) what their affection level is or (3) which mission they run -- to name a few possible sources of non-random variance -- then ideally all of those trials would be restricted to a the same companion at the same affection level running the same mission every single time. That sort of rigid approach is the only way to completely rule out everything other than it being "broken" at it's core ... should a systematic variance be shown.

Think, I'll need to start a second slicer character because this is getting unreal.

I get so many failures now.

Welcome to the 'new" slicing nerf. This is number six, and was nowhere to be found in the patch notes. Neural augmentors are almost impossible to come by now, since 1.4. I'm pretty sure the devs won't be happy until no one has it.

Here's the weird thing. I've been running slicing missions repeatedly, getting nothing after nothing after nothing.

Then last Saturday I managed to crit 9 times in a single day.

O_o

I ended up with about 25 Advanced Neural Augmentors as a result.

Also wanted to mention to the OP that you can use an online binomial calculator as another way to answer that question earlier. For example this one.

If you plug in number of tries for "n" (837) and probability for "p" (0.2 for 20%) you can then set the property of X you are interested in.

So, if you want to know, what's the likelihood that with 837 tries and a 20% probability of success you get at most 143 successes (your 17.1% success rate), you'll set the bottom field to 143 and switch the drop down to "at most". In this case, according to that calculator the likelihood of this happening is 0.018 or 1.8%.

Welcome to the 'new" slicing nerf. This is number six, and was nowhere to be found in the patch notes. Neural augmentors are almost impossible to come by now, since 1.4. I'm pretty sure the devs won't be happy until no one has it.

That is interesting and I will have to take a look at the GTN the next time that I am on. I dumped a ton of augmentors when 1.3 hit and made a killing and then ran slicing missions (wanted to keep a number of stacks in my Cargo Hold for my own use) to obtain more for sale. I stopped running the top level Slicing missions when the augmentor price dropped below 24K on the GTN and there was a lot of price cutting around that point. The last time I looked, the augmentors were going for around 17K (below the price pre 1.3) and there were a ton of listings. That was not to long ago.

If slicing was indeed nerfed, I would expect the number of sales listings to drop pretty quickly (not sure if many people have stockpiled for sale at this time) and the prices go up. Nerfed Slicing cannot maintain the current state for long.

Although I said that I stopped running Slicing missions, my OCD made me run some last weekend so that I could top off my last stack of augmentors in my Cargo Hold. As I remember I ran about ten missions and had two crits for six augmentors. Not a bad return even at current pricing. That proves absolutely noting (rng wonderland), but because of the results I did not suspect a phantom nerf.

Also wanted to mention to the OP that you can use an online binomial calculator as another way to answer that question earlier. For example this one.

 

If you plug in number of tries for "n" (837) and probability for "p" (0.2 for 20%) you can then set the property of X you are interested in.

 

So, if you want to know, what's the likelihood that with 837 tries and a 20% probability of success you get at most 143 successes (your 17.1% success rate), you'll set the bottom field to 143 and switch the drop down to "at most". In this case, according to that calculator the likelihood of this happening is 0.018 or 1.8%.

Using this calculator should not be confused with thinking that your chance for success on any individual trial is broken (and I'm not saying that the poster is claiming it; rather I'd like to stop anyone from pointing at this and thinking it's proof).

That calculator is basically stating that if 1000 people attempt 837 REs with a 20% chance of success, then you can reasonably assume that 18 of those 1000 would have at most 143 successes. Some of those 18 people will do even worse. That's not the whole truth of it, but it's close enough for explanation.

It does not mean that your chance for success on any one 20% RE has mysteriously dropped to 1.8%. It is, I believe, a good demonstration of just how screwed up the system is, relying completely on chance as it does.

It does not mean that your chance for success on any one 20% RE has mysteriously dropped to 1.8%. It is, I believe, a good demonstration of just how screwed up the system is, relying completely on chance as it does.

I think it goes without saying that not everybody will have the same results in success rates of course.

Also, if one looks at the SWTOR population as a whole, 1.8% probability of the OP's success rate is low but certainly not impossible if thousands of people keep trying over and over.

It would still be nice if a dev could just claim they checked it again and it's working fine (not saying it isn't just confirmation would be nice. Because dealing with proof of RNG system can only reasonably be given by the entity who controls the deterministic system governing said RNG).

Of course I fully understand dev time is limited so only as long as their time permits.

The original post has been updated, results are...

Total missions: 975

Critical success rate: 16.8%

Failure rate: 2.9%

As the sample set increases, I am more convinced that the chance of obtaining a critical success is not what was stated in Patrick Malott (Systems Designer)'s post. The pertinent text follows..

Patrick Malott (Systems Designer): The baseline crit chance for crafting is dependent on the difficulty of the crafting action relative to your current crafting skill level. The difficulty is color coded in the crafting GUI.

Orange Difficulty: 10%

Yellow, Green, and Gray Difficulty: 15%

Companion Affection scales Crew Skill chance up to +5% at maximum affection.

A Companion Trait critical chance bonus of +1 or +5 is a percentage. Example: The Imperial Agent companion character Kaliyo grants a +2 bonus to Underworld Trading Critical. This is +2% bonus.

Using Kaliyo at maximum affection as an example, if she ran a green difficulty Underworld Trading mission for you, she would have a 22% chance to score a mission critical success.

The original post has been updated, results are...

 

Total missions: 975

Critical success rate: 16.8%

Failure rate: 2.9%

 

As the sample set increases, I am more convinced that the chance of obtaining a critical success is not what was stated in Patrick Malott (Systems Designer)'s post. The pertinent text follows..

 

Patrick Malott (Systems Designer): The baseline crit chance for crafting is dependent on the difficulty of the crafting action relative to your current crafting skill level. The difficulty is color coded in the crafting GUI.

 

Orange Difficulty: 10%

 

Yellow, Green, and Gray Difficulty: 15%

 

Companion Affection scales Crew Skill chance up to +5% at maximum affection.

 

A Companion Trait critical chance bonus of +1 or +5 is a percentage. Example: The Imperial Agent companion character Kaliyo grants a +2 bonus to Underworld Trading Critical. This is +2% bonus.

 

Using Kaliyo at maximum affection as an example, if she ran a green difficulty Underworld Trading mission for you, she would have a 22% chance to score a mission critical success.

What "color" are the missions you are running?

Do any of your companions have a crit for the missions you are running?

Also, do not expect a small (i.e., 500 more, even 1000 more...) additional set of trials to improve your numbers. It's all random, so your next 500 trials might all be crits. Odds are, though, if it is a "yellow, green or gray difficulty" mission with a companion at 10k affection, you have an 80% chance to fail to crit the next time you run a mission. That means your success rate has a 4 in 5 chance of getting worse, not better. The same is true of the trial after that. And the trial after that. And so on. Things don't "average out" or converge to the expected rate that quickly.

That's the genius of SWTOR's system for ya ... odds are that if you start off "in the hole", in the theoretical short-run, which can be an actual long set of days and weeks or more, you will not only stay in the hole, but get further into it.

I reset my tracking tool a week ago so my numbers are probably statistically irrelevant, but here goes.

Devoted to Duty / Whispers on the Net

Run: 156

Crit: 20

Crit %: 13%

(Crit = Advanced Neural Augmentor return)

Mysterious Funds

Run: 157

Crit: 31

Crit %: 20%

(Crit = Advanced Neural Augmentor return)

The Azure Databanks / Data Race

Run: 106

Crit: 23

Crit%: 22%

(Crit = Mission or Schematic return)

Taking Back Control

Run: 84

Crit: 16

Crit %: 19%

Finding Our Way / The Fly on the Wall

Run: 86

Crit: 15

Crit %: 17%

I only track missions for companions with 10k affection. I tend to optimize my crew skills to maximize +Crit so my results may be skewed very slightly higher than the norm. Then again, it's only ~100 missions each...

I also watch crit% as I send pets out on missions and choose the higher crit% when available. I don't know if that affects the numbers beyond meaning that I tend to run The Azure Databanks and Data Race much more often than the other 2 lockbox missions.

That's the genius of SWTOR's system for ya ... odds are that if you start off "in the hole", in the theoretical short-run, which can be an actual long set of days and weeks or more, you will not only stay in the hole, but get further into it.

Eh, maybe not so much. Over time (number of instances), the returns should more or less normalize to their designed percentages. I believe if we could track 10 million of any given mission, we should know what the designed crit % should be within a few 100ths of a percentage point.

Anecdotal evidence, at one point Mysterious Funds was all the way down to 11% crit for me.

I reset my tracking tool a week ago so my numbers are probably statistically irrelevant...

Yep, that they are. But they are all over the place, so a good example of why you should not trust small sample sizes.

I also watch crit% as I send pets out on missions and choose the higher crit% when available. I don't know if that affects the numbers beyond meaning that I tend to run The Azure Databanks and Data Race much more often than the other 2 lockbox missions.

Yes, it does. But, for slicing, the best you can do is a +2 crit with T7 or Andronikos or Scorpio. That's 2%, which probably gets lost in all the "noise" of a small random sample. Even then, a companion with 10k affection gets a 5% bonus, so those "expert" crew members have their bonuses blown away by a crew member who "really really likes you." So much for "expertise"....

Yep, that they are. But they are all over the place, so a good example of why you should not trust small sample sizes.

Right. But if enough of us put our small sample sizes together, we could at some point have a large enough sample size to ferret it out.

The problem with that approach is that EA could change the percentages with any patch so we'd be limited to only the number of samples we can gather between patches.

Obviously EA isn't going to tell us when they adjust things - they are piss poor communicators as it is and given the backlash last nerf they're probably timid now too. The best we can do is all we can do, right? It's better than not tracking anything, because at least we know something.

Right. But if enough of us put our small sample sizes together, we could at some point have a large enough sample size to ferret it out.

Only if we get a random sample of people playing to report. Just because people would post to such an item doesn't mean they wouldn't have an ax to grind. I'm guessing that if we were to start a post to track success rates that a good number of people would be honest about it and report their numbers as is. Another good number of people would be reporting their lower-than-expected rates, completely honestly, but doing so because they suspected something was wrong and/or wanted to biyatch and so came to the forums and found the thread. I'd also expect that for a good number of people for whom things were going fine or better than expected, they'd never even think of looking here or knowing to report.

We would never know, but that sort of bias in reporting -- nothing intentional, but just a consequence of who would most be likely to be reading this -- would toss out the validity of anything we'd do.