For a while now I’ve seen several crafters touting that Magic Craftsmanship is responsible for orb change, and others who subsist that the ratio of orb change is weighted heavily towards any specific orb. While I’m aware that crafting is due for a change and to have much greater clarity (if Yoshida follows through with what he has said) I wanted to find out through actual testing rather than “eyeballing it”. I went into this unsure of what I would find, mostly because as any human would tell you, I see patterns where there are none. I see prismatic less often because I use it for Magnum Opus, or I see red all the time when I want white.
What I’ve found is that there is no correlation between Magic Craftsmanship and orb change; furthermore I did tests on each and every stat, Control, Craftsmanship and even low stats across the board. I also did a wide variety of different synths on different classes all with a mix of stats for my final test. None of them altered the base percentage out of what could be expected given the margin of error for each test was ~3.1%. Below I’ll list the individual tests I did for anybody who cares to see the numbers.
Method: I used primarily carpenter because they have a cheap and easy 50 synth (Dated Walnut Mask) to provide dLVL0 so that I could rule out dLVL as an influence. Additionally to rule out any sort of issues due to action failures I equipped 100% success and progress abilities for the first five trials, the sixth randomized trial only had 100% progress abilities. Using these I felt I had isolated the synthing mechanics enough that I could get an accurate representation of what I was after. I needed ~1,000 samples to hit an agreeable margin of error (~3%) and thus make sure my results were statistically significant.
Abilities: Unerring Hand, Grand Design, Tricks of the Trade, Greater Calling* (*Only the first five trials).
Trial #1: Low Magic Craftsmanship
Craftsmanship: 112
Control: 114
Magic Craftsmanship: 76
Samples: 1,004
White: 264 (26.294%)Note: As you can see they’re all pretty even, and with a margin of error of 3%, they could easily swing (assuming 25% is the true base) from 22%-28% as we’ll see in later tests.
Yellow: 255 (25.398%)
Red: 246 (24.501%)
Prismatic: 239 (23.804%)
Trial #2: High Magic Craftsmanship
Craftsmanship: 99
Control: 111
Magic Craftsmanship: 143
Samples: 1,000
White: 273 (27.3%)
Yellow: 244 (24.4%)
Red: 245 (24.5%)
Prismatic: 238 (23.8%)
Trial #3: Low Control
Craftsmanship: 116
Control: 67
Magic Craftsmanship: 105
Samples: 1,115
White: 302 (27.085%)
Yellow: 262 (23.497%)
Red: 254 (22.78%)
Prismatic: 297 (26.636%)
Trial #4: High Control
Craftsmanship: 95
Control: 166
Magic Craftsmanship: 81
Samples: 1,221
White: 287 (23.505%)
Yellow: 321 (26.289%)
Red: 299 (24.488%)
Prismatic: 314 (25.716%)
Trial #5: Low Stats
Craftsmanship: 47
Control: 37
Magic Craftsmanship: 17
Samples: 1,149
White: 282 (24.543%)
Yellow: 279 (24.281%)
Red: 294 (25.587%)
Prismatic: 294 (25.587%)
Trial #5: Randomized Stats & Synths & Disciplines
Samples: 2,560
White: 661 (25.82%)
Yellow: 625 (24.414%)
Red: 636 (24.843%)
Prismatic: 638 (24.921%)
Conclusion: As you can clearly see, the orbs are on a set %, more than likely a completely equivalent 25% for each orb. Particularly the last trial is most telling because I switched to WVR, LTW, BSM, ARM and ALC for a wide variety of synths from 36 to 54. With a larger sample size the breakdown of the orbs was less varied than in most of the other trials, with the largest variance being only 0.82%.
So, I hope all those people who keep advocating to “lower your Magic Craftsmanship because it will increase your orb change rate” will understand that they have been perpetuating a rumor that has no factual basis. It simply isn’t true, so go and stack Magic Craftsmanship as much as you like without the worry that your prismatic orbs are going to vanish.
TL;DR Magic Craftsmanship doesn't affect the color variance of the orb.
Magic Craftsmanship and Orb Change
Printable View
-
08-25-2012, 02:52 AMKestielMagic Craftsmanship and Orb Change
-
08-25-2012, 03:09 AMREDace0
Finally! Actual hard data on crafting.
Thanks for being all scientific and debunking the rumors. Can't give you enough likes for all your work. -
08-25-2012, 03:36 AMtymora
Interesting. 25% at DLvl 0. Thanks.
Orb Ratio however varies with DLvl.
Much lower samples than OP's but should still show a nice relationship.
Code:Orb Ratio DLVL White Yellow Red Rainbow
-40 0.262712 0.152542 0.186440678 0.398305
-20 0.25876 0.177898 0.169811321 0.393531
-9 0.352941 0.154412 0.139705882 0.352941
-7 0.277108 0.210843 0.162650602 0.349398
0 0.296167 0.202091 0.216027875 0.285714
-
08-25-2012, 04:24 AMGreen_and_a_Half
i have changed to the conclusion that Magic Craft and Craft should be double the synth lvl and have abandoned the lower magic craft = rainbow theory.
I would love to see that test if you feel like it as i am all 50 and dont really craft that much now.
also i have been telling people stat = 0.1% quality
meaning +10 stat = 1% quality boost
this one is all personal experience and also needs to be tested
so for me, its 120 crafting 120 magic crafting 147 control all day.
great job doing the tests, at 50 synths an hour, you did 140 hours of testing? -
08-25-2012, 04:25 AMKestielWhat's your sample size? You need ~1,067 samples (1,000 is good enough for a ~3.1% margin of error) otherwise you run the risk of getting skewed results that don't show the actual rates. I did synths varying from -15dLVL to +5dLVL and those were mainly in the last 2,000 or so synths, and they show the same rates of orbs.Quote:
Originally Posted by tymora
-
08-25-2012, 04:26 AMKestiel;_; Please don't remind me how long I spent...I try so hard not to think about it. Plus the hours of working out a way to get Notepad++ to cancel out the additional orb info from Greater Calling on the first five trials.Quote:
Originally Posted by Green_and_a_Half
Edit: 50 synths an hour is a bit generous. I clocked around 65-75 seconds per synth if I was paying extra-special attention, which is not possible all the time. So some synths took 2-3 minutes or longer if I got up and did something else. -
08-25-2012, 06:12 AMtymoraOnly a couple hundred.Quote:
Originally Posted by Kestiel
Yes, I know the values are inaccurate. Unfortunately no time to do more tests.
Clean slate is a good skill to do testing with. Effectively doubles the sample size from 1 synth. -
08-25-2012, 07:26 AMKestiel
I've not used it since the update, if it doesn't refresh abilities too it incurs the risk of synth action failure having an effect on orb color (which can erroneously skew results) and disregarding that, would make the synth take longer on average than using the 100% progress abilities. Of course if failed synth actions don't attribute to color change then that'd definitely be a preferred way to go. After I saw the new updates (recently returned ~1mo ago) to clean slate I pretty much disregarded it and haven't looked back on it.
-
08-25-2012, 07:30 PMKumaAkuma
It would have been kind of ridiculous if magic craftmanship gave more white orbs. You want rainbows when trying to HQ.
-
08-25-2012, 10:21 PMpompey_dan
this is what i find when crafting at least it appears to be true for all the new recipes i synth, it may sound stupid but please try it out before ruling it out :p
magic craftmanship - higher stat gives a higher quality gain with a successful synth.
craftmanship - higher stat gives more progress gain with a successful synth.
control - higher stat gives better chance of a successful synth.
I still dont believe the whole one stat for one synth sort of deal, i know people who stack purely craftmanship for hq nuggets whilst others stack mag craftmanship n both believe they have the best set up lol. -
08-26-2012, 12:31 AMGriss
And the quest to find out what the fuck magic craftsmanship is for marches on.
-
08-26-2012, 02:22 AMRaldo
I love how people doubt what's been officially posted.
-
08-26-2012, 03:48 AMpompey_dan
Then where is the hard evidence of people saying for example darksteel nuggets are standard synthesis, felt is magic craftmanship, most people still seem undecided or have different veiws on it,
I thought the dev post said that some synthesis will favor magic and others favour normal, then they could have meant a synth which you are trying to hq will be magic whilst a synth you are trying to just finish way above your level is standard. Thats just what i think, how about u actually try it out before blindly believing a vague dev post. -
08-26-2012, 05:20 AMKestiel
I don't advise anybody actually do any testing on what the stats do further until 2.0 comes out. We've been told the stats will change and be obvious as to their use. We've got probably a few months left of the game and any comprehensive testing would take a large chunk of that time.
The fact that people at one point were rampantly going about the crafting community saying mag. craft = orb color just goes to show that people say whatever they feel they see (which is often wrong). Our brains are hard-wired to see patterns, it helped our ancestors stay alive, stay well fed and rise to the top of the foodchain. So while one person sees Darksteel nuggets as needing high control (i've seen this suggested multiple times) another says craftsmanship (also seen this). I'd advocate balanced stats as high as you can go until 2.0 as the limited tests I've seen, and done don't show a massive swing in success rate/progress gain either way.
Please keep in mind you need ~1,067 samples to have relevant data, a few hundred is as good as worthless and eyeballing/gut/instinct is equally useless. Let's advocate proper testing rather than "he said, they said, dev said" please. -
08-26-2012, 06:38 AMtymoraYep, can't be bothered at the moment.Quote:
I don't advise anybody actually do any testing
Not really worthless if you are only going for a general pattern and not a good value. Or it can be used as a starting point for further testing.Quote:
a few hundred is as good as worthless
As above, too close to 2.0 to worth doing any detailed testing. -
08-26-2012, 07:59 AMKestiel
Worthless in the sense it doesn't give you an accurate representation and often will show you a trend that doesn't exist. It's why the usual accepted margin of error is 3% (which requires 1,067 samples) for most polling results to be considered "statistically significant". Giving a launching point though, I agree completely.
Edit: I felt I should elaborate a little. If you only do 100 samples your margin of error is a massive 9.7% which pretty much makes your work wasted since your white orb may be seen as ~16% while your prismatic could be around 34% drastically skewing your results and feeding you false information. Of course the percentage drops by around ~2% more when you get to 200 samples, and then falls to around 5% error at 300 samples. Which still could show you White at 20% and Prismatic at 30% leading you to believe something is significantly different than the truth. Of course had I stopped at one trial I could have been inclined to believe that the orbs were slightly in descending order from White->Prismatic since it seemed to line up that way. Or at least I would have had I not kept in mind that there was a 3% margin of error.
Point being, I don't think general patterns are helpful when there's little you can extrapolate in most things. If you're going through the trouble of testing, you might as well test properly. If you're okay with skewed results I'd have to question why you'd bother testing to begin with (Not "you" specifically to anybody, just the general use).
For this, knowing the orb rates is only mildly helpful. The true purpose was to see the effect of stats on orbs, and in that case a swing of even 5% could have led one to believe the orbs were influenced by one stat or another, more than a 3%. In an ideal world I would have done 9,000~ tests to get a 1% margin of error, but I'm not that masochistic. -
08-26-2012, 10:16 AMGreen_and_a_Half
when I craft class weapons, or any HQ items, I don't do 100 synths to get it.
I do 5 for the weapons/armor
and I do 1-2 for the HQ other stuff
So setting gear to what I think works Is what most people do.
Ive never had mag craft over 70 when hqing any item
until people started saying double craft lvl is best, now its at 120
same with crafting, never been over 100 until then, now its 120
and fro control, always ran with 150 but now its 130.
And still able to HQ the same amount of items.
I'm glad there are testers but ful 50 with "Crafting" gear on WILL make HQ items
Happy crafting
and not knocking any of the testers, thanks for all the work. -
08-26-2012, 01:33 PMtymora
Haha, can't say I don't agree with Green on this.
TBH, Orb ratio isn't really an area of interest I am concerned because it isn't anything I can control and it only affects one skill...Magnum Opus.
I am more interested in:
- Action Success/Great Success Rates on each Color (or failure rate if you take the inverse) and how they vary by craftsmanship and DLvl.
-> You'll need way more than 1067 samples for those. As it is, my couple hundred samples do confirm my own observations on their variations. While I can't give you a formula on them or decide whether craftsmanship trumps stats or the reverse, I am able to do my crafts comfortably with the partial knowledge.
My best records so far are +1000 quality on Natron (47 lvls above) and + ~830 quality on Cobalt Ingots (I still do get ~400 quality or so once in a while, but I have learned to accept those "destined" to screw up synths) and I am pretty happy with that.
---------------------
My observations tallies with Green's: 120/120 craftsmanships should be enough to cover the requirements for most if not all current recipes in game.
---------------------
Not bashing on the need for a detailed analysis of stuff but most people just don't have the time to do it.
Maybe if I can finish this crafting parser I am working on...lol