The crafting & gathering thread.

[Raineer]

Is there a certain data set which would help your research? I have crafts spread among all the levels from 20-50, can run some controls for you if it'd be helpful.

[Intervention]

Also is Innovation multiplicative or additive with Inner Quiet?

I believe its additive, its a + 1.5(BaseControl)

On an odd note, not sure if was mentioned, that if you have turns left in Innovation it keeps that the control that the Inner Quiet makes.

As well not sure if it was mentioned but you can "see" the Inner Quiet in the character screen.

[Grey Jorildyn]

Could use Ingenuity I and II to approximate certain thing since they effectively lower the synth by a level and three levels respectively. I don't know if that helps but it's an idea.

[Rehwyn]

Is there a certain data set which would help your research? I have crafts spread among all the levels from 20-50, can run some controls for you if it'd be helpful.

I appreciate the offer for help. I should be fine for all the testing of equal level synths (to find the base formulas) and synth below my level, but I'll most likely need some help with data for synths 1-4 levels above you.

Don't run anything quite yet though; I'm doing the equal-level stuff to derive the base formulas first. Most likely after I finish these I'll have a better idea of what data is missing to figure out the level difference penalty.

Could use Ingenuity I and II to approximate certain thing since they effectively lower the synth by a level and three levels respectively. I don't know if that helps but it's an idea.

Not quite; they actually change the effective level of a synth to your crafter level or three levels below you, respectively. I may end up doing that if there isn't an easy synth 3 levels below me with Ing2. Also, Ing1 technically raises a synth to your level if it's below you.. might do that to save shards for the rest of my research.

On a side note, I haven't completed my data set yet, but it actually appears that the Quality function is not truly linear, and the Progress function may very well be the same way. It seems to instead be a 2nd order polynomial that is almost, but not quite, linear. Didn't finish my testing tonight before the servers went down so don't want to post any half-finished regression formulas, but I should have some tomorrow after maint.

[Meresgi]

Is there a good source to be had to find the best leves for each crafting job? I usually just stuck with the ones in Limsa when doing BSM to 50, but some of the ones that are at other camps tend to be decent too. Talking about multi-turn in leves usually, but I did farm some single turn ins as well since the exp was good and the items were easier to make/less materials.

Just wondering if there is a best list for materials : exp gained.

[Jaybar]

Is there a good source to be had to find the best leves for each crafting job? I usually just stuck with the ones in Limsa when doing BSM to 50, but some of the ones that are at other camps tend to be decent too. Talking about multi-turn in leves usually, but I did farm some single turn ins as well since the exp was good and the items were easier to make/less materials.

Just wondering if there is a best list for materials : exp gained.

http://www.bluegartr.com/threads/118...Material-Tiers

[Andradi]

Regardless of everything else, courier leves are always the best XP for a single craft in that tier. If you favourite the town and the destination you can easily break even or better, too.

It's just a shame that from 35 all the way up they need a lot of travelling.

[Rehwyn]

So, still working on my new data but in the process I've figured out something rather strange. The Progress and Quality gains for a same-level synth are occasionally different compared to a different-level synth with Ing1 up. For example, on my 50 ALC, I'd get 18 Progress from a Basic Touch with 81 Craftsmanship while on a Mega-Ether synth. However, if I then do an Antidote and use Ing1, the Basic Touch did 19 Progress with the same craftsmanship, despite the fact that with Ing1 up the synth is temporarily a "level 50" synth and I'd expect the same Progress.

I suspect what is happening is that the game calculates your Progress and Quality result prior to accounting for Ingenuity using the actual Synth Level and applies any appropriate penalties/bonuses. Then, Ingenuity multiplies/divides the resulting Progress and Quality by some factor to "correct" the value as if you were making an even-level synth. Depending on how the game rounds values, this could lead to the fluctuations I've been seeing.

Also, with 338 control, Basic Touch adds 157 Quality to the lvl 50 Mega-Ether and a lvl 43 Mega-Potion, both with and without Ing1. However, it adds 158 Quality @ 338 Control to the Antidote synth, both with and without Ing1. I have no idea what the reasoning behind this would be.

What this means is that a 1/2-star synth with Ing1 up will not behave exactly the same as a true level 50 synth (and I suspect the same can be said for Ing2 compared to 47). I confirmed this on WVR crafting a Twinthread. For the same various Control values, I was getting slightly less Quality with Ing1 up on the Twinthread than compared to a true level 50 synth.

This could potentially complicate things greatly in terms of data collection and generating useful formulas for crafting optimization. I'll have to do some follow-up testing in just a bit to see how much more data would be needed.

In the meantime, I can say that a 2nd-order polynomial function is a significantly better match than a linear function to the values I've gathered so far for Mega-Ethers. I'd like to get a few more data points in the 300-700 control range, but I should be finished with at least the base quality function soon. Craftsmanship and Progress should come soon after.

[Deadgye]

I wouldn't rule out piecewise functions either. FFXI had gazillions of those after all. Perhaps instead of a function based upon quality we should try creating fuction based upon the difficulty listed for each synth. Maybe even a delta between the difficulty and our stats.

[Azninvazion]

I believe its additive, its a + 1.5(BaseControl)

On an odd note, not sure if was mentioned, that if you have turns left in Innovation it keeps that the control that the Inner Quiet makes.

As well not sure if it was mentioned but you can "see" the Inner Quiet in the character screen.

I've noticed oddly large quality gains following a Byregot's when I still have two ticks of Innovation, but I chalked it up to the Innovation bonus. This gives Innovation quite a bit more utility if it does preserve control for three steps.

[Raldo]

W...why are you still using quality actions after a Byregots?

Also, I don't think you lose the Inner Quiet control stacks once you use Byregot's (they simply don't continue to increase), so that's completely expected.

Apparently not, so good to know, I guess.

Now people are arguing both ways. I'm at work, so I can't test.

[Niiro]

Also why would it be peculiar for the control from Innovation to last 3 steps?

Have you read the description?

[Ksandra]

W...why are you still using quality actions after a Byregots?

Not everyone is mathy? I tend to accidentally do byregots too early and get to like 98% quality. So sometimes have to throw in another synth to cap it out.

[Spider-Dan]

Also, I don't think you lose the Inner Quiet control stacks once you use Byregot's (they simply don't continue to increase), so that's completely expected.

You lose IQ bonus when you use Byregot's (i.e. your control goes back to base).

[Raldo]

Not everyone is mathy? I tend to accidentally do byregots too early and get to like 98% quality. So sometimes have to throw in another synth to cap it out.

Sure, sure, I do that same thing too on occasion when the thing I'm synthing isn't important, but your particular example is a bad one, since you don't get to see the full quality numbers if you bump up against the 100% line.

You lose IQ bonus when you use Byregot's (i.e. your control goes back to base).

Ah, good to know. I had heard both ways, but never cared enough to check 'cause Byregot's is (almost) always the last thing to do with quality.

[Ksandra]

but your particular example is a bad one, since you don't get to see the full quality numbers if you bump up against the 100% line.

You didn't ask that though you just asked why people would do a synth after byregot's

I don't think anyone is suggesting this means you should do byregot's whenever, it's just a nice bit of info to know.

[Intervention]

Also why would it be peculiar for the control from Innovation to last 3 steps?

Have you read the description?

It maintains the IQ stacks even after Byregots. My initial thought would be that you lose the bonus after it.

[Rehwyn]

So, I've got some concrete results for base Quality for even-level synths. Still need some more samples to narrow down Craftsmanship, but collecting those takes a bit more time due to the lack of something like Inner Quiet.

My original Control sample weeks ago was around 40 samples between 30 and 150 Control. My current one is 216 Samples between 0 and 1058 Control, so the overall sample quality is much higher.

I used both Excel and this website (http://www.arachnoid.com/polysolve/) to compute my regressions. Also, I mistakenly only looked at r-squared last time instead of standard error as well, which is actually more useful in this situation.

When attempting to fit a linear function, I get the following results:

f(x) =  3.1624326328095464e+001 * x^0
     +  3.7764581988053514e-001 * x^1

Correlation coef (r^2) = 0.9994581328952155
Standard error = 2.2218606977257687

Even though the r^2 value looks nice, that standard error is pretty huge considering the fact that the game is giving us exact figures rounded to the nearest whole number. Also, I got 34 Quality at 0 Control, so I'm very confident that the intercept would be around 34, whereas this function would return 31 Quality at 0 Control.

Based on these results, a simple linear function is a poor fit.

I decided to test a polynomial function next. The following is the result of a 2nd-order polynomial regression:

f(x) =  3.4193540129360166e+001 * x^0
     +  3.5022953538382334e-001 * x^1
     +  3.4893511935074471e-005 * x^2

Correlation coef (r^2) = 0.9999900379936876
Standard error = 0.30126164256577614

Here's a nice plot for the visually inclined:

Spoiler: show

The correlation coef is even higher for this, but what is really noteworthy is the standard error of 0.301 instead of 2.22, indicating a greatly reduced predictive error for this formula. Testing higher order polynomials did not significantly improve the standard error.

And some detailed info on each coefficient for the statistically inclined:

Coeff	Lower 95%	Upper 95%	Lower 99.0%	Upper 99.0%
c0	34.11831679	34.26876347	34.09435327	34.29272699
c1	0.349697811	0.35076126	0.349528422	0.350930649
c2	3.42486E-05	3.55385E-05	3.40431E-05	3.57439E-05

Because the function is only moderately curved and I can't get values for negative control, the confidence intervals for c0 and c2 are a bit bigger than c1.

Subjectively speaking, the fact that both c1 and c2 are essentially 3.5 times a power of 10 seems like something simple a programmer might choose as well, though honestly I have no idea why they didn't just keep it linear, as he effect is relatively minor (yet observable). Also, you gain 35 Quality at 1 Control, which would indicate the intercept is definitely is greater than 34.14, otherwise it doesn't round right.

Personally, I'll be using the fully-expanded values for my own spreadsheets, but if I were to recommend a rounded function to others that want to use the formula in a program, something like this seems to fit my data best:

 Base.Quality = round(34.23 + (.3502 * Control) + (0.0000349 * (Control ^2)))

That particular function matches the data I've found on all except 1 value out of 217. The intercept is higher because rounding the other values affects what intercept is a best fit.

Anyways... craftsmanship should be coming later tonight or tomorrow. Then I'll see what I can do about the level modifier, but I may need some help from others for that. Thankfully, figuring out the modifiers should required a lot less data points, so should go faster.

[Niiro]

You lose IQ bonus when you use Byregot's (i.e. your control goes back to base).

No, you don't. You do lose the IQ buff itself, but the accumulated Control bonus is still there.

It maintains the IQ stacks even after Byregots. My initial thought would be that you lose the bonus after it.

Now I see what you meant, but that has nothing to do with Innovation.

[Deadgye]

If you fail byregot's, you lose the control bonus. If it succeeds you still have the control bonus. That's what I remember happening off the top of my head at least.