Dec 1, 2011

Posted by in Zlatto's Bazaar | 26 Comments

Zlatto’s Bazaar: Reverse Engineering Theorycrafting

Zlatto’s Bazaar is the one-third Wall Street, one-third criminal empire and one-half gossip rag. You need to keep up on the Bazaar if for no other reason than to keep an eye on your competitors. Don’t be a Ho-tah and visit often.  Your comments are encouraged, especially if they make me credits…

Hello, all you GTN moguls-in-training! I’m glad you decided to stop by so we can talk about the exciting math of Reverse Engineering! With the NDA gone, I wanted to get into some of the exciting non-combat mechanics of the game. No spoilers to worry about here, it’s all about crafting and Reverse Engineering in mass quantities to enable us to look a bit behind the curtain of ‘random’ as defined by BioWare.

You can toggle Reverse Engineering on in your Inventory panel, which will allow you to select certain items to... Reverse Engineer and perhaps learn to make a better version of that item!

I am not a statistics guru, but because I am skilled in short-changing purchases in any monetary unit of the known galaxy, I figured it would be best for me to share my accumulated numbers with you, and then together we could decide what they mean. I expect there is a closeted math geek or two that might read the column, so lets hope they decide to chime in and give us some deeper analysis.

This article will assume that you are all up to speed on the process, risk, and rewards of Reverse Engineering. If you are not yet a guru on the topic, take a quick moment and read Momus’ informative article on the subject found over here in Blasters, Beggars and Credits. For now, I’ve decided not to dive into the itemization gain seen at each level of Premium (green), Prototype (blue) and Artifact (purple.) That discussion could barely be covered with a 12-part series and would require more time than I can spend at the moment.

To start, I spent some time running my level 25 trooper around the Migrant area on Coruscant to gather enough resources to sit back and grind a high volume of items to help figure out some patterns. Let’s just say that those leveling in the area were a bit disappointed to see a higher level toon running around stealing all their nodes. I can say with great certainty that there is no Star Wars curse word filter on the general chat. What was more hilarious was that many could not spell the curses correctly; it is Bantha Poodoo, not Bantha Poodu, a total of four o’s you Di’kuts.

I believe a column on Resource Grief-ing might be a topic I cover in the future, as BioWare is almost pushing players into it with the costs of leveling crew skills.

Enter The Matrix

RE to Proc Schematic Used Result of Proc
15 Skill Barrel 2 Skill Barrel 2 – Blue
8 Reflex Barrel 2 Reflex Barrel 2 – Blue
7 Skill Barrel 3 Skill Barrel 3 – Blue
12 Reflex Barrel 3 Reflex Barrel 3 – Blue
9 Skill Barrel 2 – Blue Advanced Skill Barrel 2 – Purple
2 Reflex Barrel 2 – Blue Advanced Reflex Barrel 2 – Purple
7 Skill Barrel 3 – Blue Advance Skill Barrel 3 – Purple
11 Reflex Barrel 3 – Blue Advanced Reflex Barrel 3 – Purple

Above is my first bit of statistical analysis is using the first two tiers of barrel mods. The amount of Silca and Desh used is mind boggling and both the GTN and my guild ran out of Carbo-plas quickly, as that material is needed to generate the second tier of the item and is not found in standard nodes anymore (it comes solely from Investigation missions).

The numbers above show that the success rate of Reverse Engineering is all over the place. If you just took the first 38 attempts that returned 4 new patterns you would see about a 10% chance of learning a blue schematic from the greens that were reversed, but when you evaluate the numbers for the next tier you see closer to a 14% chance.

The Reflex Barrel 2 to Advanced Reflex Barrel 2 returned in 2 attempts, so that seems to have skewed the consistency between the two tiers. The materials needed to create the blue items to reverse are significantly more expensive. BioWare possibly has decided to give us a break on the success rate to balance out the cost of the high end resources.

On my next set of tests, I decided to work up the dead sexy Combat Enforcer. I mean who isn’t going to want a level 13 artifact blaster rifle? I spent a long time looping the Black Sun and Justicar areas accumulating multiple stacks of Aluminium and Laminoid.

A funny observation; the players questing in that area seemed to be just as disappointed in me stealing all the nodes as the players in the Migrant area. I do love consistency in my game play.

RE to Proc Schematic Used Result of Proc
7 Combat Enforcer Overkill Combat Enforcer – Blue
9 Combat Enforcer Redoubt Combat Enforcer – Blue
3 Combat Enforcer Critical Combat Enforcer – Blue
6 Redoubt Combat Enforcer Exactitude Combat Enforcer – Purple
8 Critical Combat Enforcer Expert Combat Enforcer – Purple

So the resulting schematics learned from reverse engineering a weapon is much different than that of a mod. Instead of getting a blue version of the original mod, a successful new schematic or ‘proc’ for a weapon can generate one of three possible blue formats.

From what I have seen a blaster or electrostaff can give – Redoubt (plus defense), Critical (plus crit), and Overkill (plus power). Looking at the numbers from green to blue, 19 items reverse engineered into 3 schematics gives us around 15%. The numbers from blue to purple seem to be 14 items into 2 schematics which gives us around 14%.

To finish off the math of the article I wanted to work up one more set of green to blues. An interesting observation is that some of the higher end schematics in staffs and swords are actually better than some of the average greens players will be acquiring during the course of leveling. Sure its not a lightsaber, but better damage is better damage.

RE to Proc Schematic Used Result of Proc
9 Battle Champion’s Electrostaff Critical Battle Champion’s Electrostaff
11 Battle Champion’s Electrostaff Redoubt Battle Champion’s Electrostaff
9 Battle Champion’s Electrostaff Overkill Battle Champion’s Electrostaff

The resulting math seems to circle back to a 10% schematic discovery rate. Could this be a base number where the other factors can adjust for each player? How does the affection rating impact this?

I avoided using any crew that had a +Crit to Armstech to keep the numbers as general as possible, but there are a lot of unknowns that can change the math here. It will take thousands of items created and RE’s to find all the possible dependencies.

Let us know your experiences if you did any theory-crewing in beta!


  1. I spent a couple hours waiting on a friend to run Black Talon. While waiting i used the Archaeology and Synthweaving Crew Skills to do a bit of grinding, before I had even left the space station. I got up to about the mid-70s in both, but i was down to around 800 credits (from about 3000 when i got to the station) after doing so.

    I did some reverse engineering and I was always fairly pleased with the results. For armor, at least, it felt like i was going from green to blue in about a 20-25% rate. I didn’t document numbers or anything, but I felt pretty good while I was doing it. Much like weapons, you can create “Overkill” and similar versions of the piece.

    I don’t recall all the stats I had, but the blue chest piece I made lasted me all the way through Dromund Kaas and well onto Balmora. Of course, the lack of a single “Heavy Armor” drop in either Black Talon or Hammer Station. Seriously, it was pure Light Armor the entire time.

  2. So what do you think about the reverse of Purple items? I have not been able to pull together enough items to build any volume to see if they open up to the next level.
    Should it even be available as an option?

    • Lady Republic says:

      What would be the next level of reverse from a purple? Legendary? If so, then I do think it should be possible but extremely difficult. They did originally say that they wanted to reward dedicated crafters and have some schematics be so rare, only one person on the server/shard might have access to that schematic.

      • Good question… is there another level of Reverse Engineering, or does it stop at Artifact? Would be interesting to know why, if it does.

        Side note, I disagree with the item color changes in the new build! :)

  3. I don’t spouse toot can throw in the numbers for what your crafting level was when each RE procced? Apparently the chance seems to depend on crafting level vs item level.

    • (s/spouse toot/suppose you/ – bloody Swype)

      • Lady Republic says:

        In my case, I did synthweaving when I was around level 10 for a while, and I’d say about 10% change was correct for that profession and level range. My skill was lower level, 1-100 or so. I only did greens to blues, not blues to purples, due to lack of resource.

  4. Ive only managed to play during last weeks open beta stress test. I went a scavenger and armormech. Following something I read on here I took my trooper and ran straight to the shuttle bay and went up to the station and got my scavenging. I was approx lvl 80-90 by time id completed all the quests and had a fair amount of silica and desh to begin crafting once I got my companion. I found it took approx 10-15 of the greens to learn the blue and took me about 7-8 of the blues to learn purple version (this was on the first belt you were able to craft) but i tried it on another item a little later and found it took a similar amount to upgrade the green to blue but couldnt get enough matts together to make sufficient amounts of the blues to learn the new purples. I found while doing the green to blue over 3 diff items it took between 10 and 15 reversals to learn on 2 and think the third it only took about 8 so dont know if skill level might count to help you learn faster?

  5. Any math geek will tell that your data is too small to give reliable results. Let’s look at your first data set – 4 new Prototype schematics from 42 attempts of REing Premium items – gives 9,5% chance with standard deviation of 3%. This means that proc chance can range from 6,5% to 12,5%.

    Affection does not impact this, because it’s player character who does REing, not companions.

    I have some information, but unfortunately from few builds behind so it could have changed. There are 5 chance rates of researching new schematics:
    Very Low – 4%
    Low – 8%
    Medium – 10%
    High – 16%
    Very high – 20%

    I got into this Beta Weekend so I will keep pushing my Crew Skill (200+ already) and see if I find out something useful.

    • Yep, it was simply a first approximation.

      The numbers that are going to give solid returns are going to be hard to come by for a single or small group of dedicated crafters in a weekend, because there are only so many Investigation missions you can run, and they can fail and give you nothing.

      This will be the major bottleneck on any crafting of the higher quality gear.

      At the time, the most common of those missions returned only 1 or 2 carbo-plas, and you need at least 2 to make one blue barrel. So 15 minutes of time to make one blue that you RE in 4 seconds (now “moderate” missions give 3-4 of the blue item, which is quite nice – but Rich don’t give purple materials each time, which makes me sad).

      As an approximation and general trend, I would be comfortable (from his experience, my experience, and your numbers) with claiming a 15-20% general schematic return rate. I would think the “very low” and “low” comes from RE’ing a pattern you just learned and have bought some from the AH.

      Side note: starting to get a decent amount of schematic rewards from Underworld Trading (~260 skill now, synth/arch are at 280) and “rich” level missions – I’ve gotten 3 or 4 now, and the schematic I got at 260 was for 300 synth (mod-able medium JK/SW robes).

      • Lady Republic says:

        Yes but the other up side is once we go live, presumably you’ll be able to find more rare components on the GTN too. At least, that’s my hope – I imagine a fair number of people will pick up just gather and/or mission skills rather than crafting too, purely for profit purposes.

        • Sure, when we start thinking about the aggregate of all Investigation missions to the amount of, say, carbo-plas, you would expect some to wind up on the GTN.

          One other note: it seems that there’s now a sliding scale for the mission costs. Rich “orange” missions cost a decent amount, but the jump up from bountiful “yellow” missions was somewhat severe. It seems that difficulty and your relative crew skill level dictate how much, and that goes up and down based on where you are.

          To which I say, I like it. I think it offers some nice options for sending someone back to get a high amount of older materials at a cheaper cost, which would be especially helpful to alleviate supply issues.

        • As far as I know, Prototype(blue) and Artifact(purple) materials comes only from missions, so mix of costly missions and people going for profit will give quite high prices for them.

  6. I dug up some of the rudimantary statistics I’m supposed to know. If we look at the first experiment, we can consider this a sample of the probability of proccing a mod. For sake of simplicity I assume a universal rate without respect to green to blue vs blue to purple. It is a sample of a larger population of proc samples.

    By calculating the mean and squaring the difference from each sample to it, then taking the squareroot of (squared differences)/(samplesize-1) we get the sample standard deviation (SD) of 4,2 from the mean 7,3

    I don’t have time to calculate the proper boundaries of a confidence interval for an 8 sample size, but for ten samples in the 95% CI it is roughly 0,7*SD to 1,8*SD. In other words, the SD of the total number of potential sample sets will in 95% of the cases be in the interval 2,9 to 7,7. All that this tells us, is that our sample size is probably too small.

    But for the hell of it, let’s assume we got it right. In that case, the SD can hint us what samples are outliers from the distribution. Anything outside the interval mean-SD to mean+SD is worth noting. That interval is 3,1 – 11,5. Three samples (15, 12, 2) fall outside this. On the assumption that there is some fixed percentage chance of a proc, these results are anomalies and should be discarded from calculations (they scew the result). Looking at the remaining samples, we need 42 attempts to get 5 procs, or a proc rate of about 8,4%

    So the proc rate may be somewhat lower than Zlatto said. However, and this is a big however, we eliminated those results based on an SD that may be 1,8 times higher (as our confidence interval told us), in which case the eliminated datapoints should not have been.

    Right now the only conclusion we can draw is that the procrate is probably around 8,4%, but we need a lot more data (in the order of at least 100 I would say) to be confident in our result.

    • Bah, I call BS on one of my own sentences: “In other words, the SD of the total number of potential sample sets will in 95% of the cases be in the interval 2,9 to 7,7”

      This is me getting ahead of myself. Replace from “in the interval” with “up to 1,8 times bigger or 0,7 times smaller”

      • Also, I suck at percentages and units apparently. It doesn’t give a proc rate of 8,4%. It gives us that on average we have a proc every 8,4 items we RE.

  7. Ratkinantics says:

    Thanks peeps. I understood about 45% of that, but luckily I’m best mate’s with a Math Prof, so I’ll be dragging this page towards him with the offering of a pint or two and will see what he says. (I’m assuming he’ll say “More Beer please” but he might actually offer me some translation).
    I did however find the article very interesting as this is probably the first MMO I will be actually looking forward to the crafting side of things, after running my companion to death on scavenging/crafting missions whilst I was running Black Talon repeatedly last weekend.

  8. Chris Deke says:

    Just incase you didn’t know, you can proc the purple version from RE’ing the green version. I did this with the barrels and the combat enforcer. It took a lot longer, but it’s certainly possible, and saves those valuable Carbo-plas’.

    • That is interesting if in fact it wasn’t a bug… hmm.

    • I am excited to see this happen, in all of my reverse efforts I have not seen this. BUT, I did stop reversing as soon as I got the 3 blue options per gun or the blue mod version. If I would have continued RE’ing I wonder what I would see?

    • That must be a really low chance because I have constructed and RE’d 200 or so items and haven’t seen it yet.

      • Chris Deke says:

        It can’t be that low, I don’t think I got anywhere near 200. I maybe made 20 or 30 of the Combat Enforcer’s after getting the blue, and that got me the purple?

        It only took me 8 or 9 to get the blue version.

  9. I’ve just posted a guide to crew skills at http://sithwarrior.com/forums/Thread-Crew-Skills-Compendium – check it out.

  10. DurdensWrath says:

    Through Documentation it appears that for every green there are 3 blue possibles.

    And there seems to be at least 3 purples per blue.

    Confirmed is Accuracy and Leadership.

  11. One thing I have noticed and I am quite convinced about now, is that a character that is rested and has not been played for some time, has a better then 80% chance to successfully learn a schematic from an RE on the first time he RE’s something after logging in after a long break. I have successfully used this tactic about seven times now on multiple characters on the same server. It’s not an account wide thing either, It’s per character. I know its speculative to say it and I have no ‘hard’ data, but I gotta tell ya I almost guarantee a schematic if youve been logged out for more then two to three days on any given character. Possibly a little longer.

Trackbacks/Pingbacks

  1. Crew Skills Compendium | SWTOR Notes - [...] Suggested here: There are 5 chance rates of researching new schematics: Very Low (4%), Low (8%), Medium (10%), High …

Leave a Reply

Your email address will not be published. Required fields are marked *