The Ultimate Black Mage Resources with Damage Calculator

[EasymodeX]

The only time you should really be concerned about Acc is Turn 5. And in Turn 5 you are also almost always should be concerned about Vit. Which means you should be concerned about food anyways.

You will submit to your Culinarian overlords and you will like it.

[Cherub]

Sigh

/10char

[leteton]

So anyway, has anyone found out more information on the BLM ancient buckler? Where is it!?

[Exstal]

It seems food is absolutely necessary to reach 100% accuracy for Set B. Please tell me I am mistaken since I'm not the big food fan.

Replace Allagan Gloves with Sorceror's Gloves. Lose 8 DET, gain 15 accuracy. However, there is no way you'll make up 14 Vitality without dropping stats somewhere else.

[Cherub]

Replace Allagan Gloves with Sorceror's Gloves. Lose 8 DET, gain 15 accuracy. However, there is no way you'll make up 14 Vitality without dropping stats somewhere else.

I guess I'll go with the food then. Will have to pester my own personal culinarian for dat but fine.

[Kenji1134]

Hey Puro, Im taking another shot at the ol potency equation, seeing if I can get something that kinda works for every class... lol, started discussing linear and matrix algebra in my math class and the muse just kinda hit me.

Right now I need more data points. Specifically I am looking for the following setup and methodology:
*Note, this is best done using a parser such as ACT or something that lets see you exactly how much of your damage was from crits...

So the testing methodology is very simple.
1. Pick your gear-set.
2. Write down your Weapon Damage, Intelligence, and Determination. I do not need Crit or Speed for this portion of the formula.
3. Find a lvl 1 target dummy and just spam Blizzard 3 on it for ~3 minutes. This is roughly 50 casts of B3, NONE of which should ever be done under AF3. UI3 does not affect B3 damage.
4. Using that parser, determine the average NON-CRIT damage for your B3 spam using your recorded stats.
This is important, since Crit is not in this analysis, every single critical hit must be removed from the calculation to maximize accuracy. This is where a parser like ACT comes in handy, so you can see how many times you crit, how much of your damage was non-crit, how many casts, and you can figure out how many non-crit casts you did to get your total non-crit damage.
5. Take your average non-crit B3 and divide by 220.

In the end you will have a dataset consisting of: [ Weapon Damage , Intelligence , Determination , Potency (B3/220) ]

At this time I only have 6 datasets, due to time constraints, farming for gear, and upcoming finals. These datasets are using the lvl 1 wand which has 8 magic damage, and going from naked - stacking ONLY Int with NO Det - then adding Det back in over 2-3 sets.
So, here is what my current little database looks like:

8 272 202 0.502409091
8 305 202 0.5515
8 325 202 0.587318182
8 351 202 0.624240909
8 389 226 0.688059091
8 416 246 0.737877273

When I have a little more time I will try to get some damage using a ~30 damage weapon, and later the 69 damage +1.
But for now more datasets using this VERY controlled method would be appreciated. With some luck, I can eek out a good general form for the raw damage portion of the equation, after that the effects of Crit and Speed are multiplicative with the damage... Throw in some partial differentiation... and we should have a system where you can put in your current stats, and the equations will spit out how much an increase of '1' in a given stat will increase your effective potency... crits and spell speed and all.
Then a lil tweaking... and with some more luck, this may (should) work for all classes... with slightly different coefficients.

[Sleigh]

The only time you should really be concerned about Acc is Turn 5. And in Turn 5 you are also almost always should be concerned about Vit. Which means you should be concerned about food anyways.

You will submit to your Culinarian overlords and you will like it.

You never realize exactly how important ACC is to BLM until you miss your Blizzard III, use Thunder II, then realize your Blizzard III missed and you have no MP left due to no Umbral Ice 3 mode.

[EasymodeX]

Specifically I am looking for the following setup and methodology:

4. Using that parser, determine the average NON-CRIT damage for your B3 spam using your recorded stats.

I recommend recording your minimum and your maximum non-crit damage (sidenote: pretty sure there's no crit variance in this game, so you could take crit damage and /1.5, but w/e, not a big deal). Then, take the average of your min and max non-crit damage (also report your min and your max damage).

Using a population average leads to unnecessary RNG and variance with the result. For testing something in an artificial system where we know there is a real specific formula (FF14 damage formula), there's no reason to average all the results.

I can eek out a good general form for the raw damage portion of the equation,

Don't forget the multiplier trait for BLMs when you begin comparing to other classes. You should probably include this up front in your BLM assessment.

Throw in some partial differentiation... and we should have a system where you can put in your current stats, and the equations will spit out how much an increase of '1' in a given stat will increase your effective potency...

Not gonna happen with intervals that small. No one's identified a fully accurate damage formula yet. So it goes.

If you're interested you should speak with Eien at chocobro. He's on something of a crusade to solve this.

Edit:

In terms of raw problem-solving, you will want the following:

1. 0 DTR, range of Int * range of WD
2. mid DTR, range of Int * range of WD
3. high DTR, range of Int * range of WD
4. Couple random data points for DTR at random intervals.

... well that should be everything tbh.

One of the problems is that it's resource-intensive to adjust your DTR fluidly with gear. You're talking using weird combinations of crafted gear, etc.

[Kenji1134]

There is crit variance. If you have a WHM, try spamming Cure 2 until you get an Overcure proc then use Cure 3, Cure 3 will always crit, but not for the same value.

I'll take a look at the (max-min)/2 average, though that may require a LOT more than 50 casts to get a solid answer.

As far as equations and math... All damage mods appear to be upfront multiplication, so they should have the same effect on all of the coefficients if the model is constructed in a linear way... And some non-linearity can be built in by messing with value structures vs inputs... ie: A1*WD+A2*Int+A3*Det is linear, but (A1*WD+A2*Int/IntBase)*(1+A3*Det/DetBase) which isnt "quite" linear -> A1*WD + A2*(Int/BaseInt) + A1*A3*(WD*Det/DetBase) + A2*A3*(Int*Det/BaseInt/BaseDet) is linear, IF we claim that A1*A3 is A4 and A2*A3 is A5... Its a lil ugly, but it should work.

[EasymodeX]

There is crit variance. If you have a WHM, try spamming Cure 2 until you get an Overcure proc then use Cure 3, Cure 3 will always crit, but not for the same value.

Not good enough*: there is a +/- 5% damage variance (I'm not 100% positive it's +/- 5%, but everyone says it is). As a result, it's expected that a flat *1.5x multiplier on crits will also reflect the +/- 5% variance. A variance on crit damage itself would result in more than a +/- 5% variance. This hasn't been demonstrated to my knowledge.

* Unless you're saying Cure 3 / heals in general have no variance.

Either way it's easy enough to ignore crits, regardless.

All damage mods appear to be upfront multiplication,

There is multiple partial pieces of evidence that DTR scales faster with high primary stat / WD. Linear scaling of everything also fails to match actual results for most peoples' datasets.

I'll take a look at the (max-min)/2 average, though that may require a LOT more than 50 casts to get a solid answer.

Max and mins take a hell of a lot fewer casts than trying to wash out RNG using an average of the population. You're talking 30 casts versus 200. Especially if you assume a specific variance (+/- 5% for example), and stop when you make a range wide enough to fit that.

FWIW I've brute-forced some random formulas across various datasets. The only conclusion I've seen is that DTR probably has a -202 adjustment.