Damage Formula Thread

[Nyaanyaa]

Fair points, Kenji. At lower levels, though, it is fairly simple to find the min/max for a given data set, as the range is fairly low. Has anyone done any testing at all on how/if the caster's Lv, target's Lv, ΔLv, target's def/mdef factors into the equation?

[Kenji1134]

I believe Valk did some testing with that and came to the conclusion that relative levels do not matter.
He also came to the apparent conclusion that either every mob at every level has the same defense... or that they all have 0 defense and just have more/less HP to make them "tankier" or "squishier".

I have done parsing on lvl 1 dummies, and lvl 50 dummies, same results.
It also seems to hold for mobs as well. Relative levels mean nothing beyond requiring accuracy, and everything has a base defense of 0.

[Nyaanyaa]

Thanks, Kenji. So ΔLv affects relative accuracy? I thought I saw an increase in hits as I leveled, but wasn't sure if it was from increased acc or ΔLv. I only just the other day found out that acc=m.acc as well in this game lol, still somewhat new.

[Kenji1134]

You gain accuracy every time you gain a level, so in effect yes, ΔLv affects accuracy simply because you would have more if you were of that mob's level.
In Coil you end up fighting mobs who's levels are much higher than your's, so you need enough accuracy on your gear to bring it up to the level it "would" be if you were the mob's level... plus a bit more to overcome the mob's base evasion rate.

[Nyaanyaa]

So it's a simple acc vs eva check, and a Lv50 DRG with 300acc would have the same relative accuracy on a given mob as a Lv30 DRG with 300acc.. theoretically anyhow. Thanks again.

[Nyaanyaa]

Hypothesis

The Cure Formula follows one of these models:

• HP=floor(3%±RHP),
• HP=floor(3%±floor(RHP)),
• HP=floor(3%±round(RHP)),
• HP=round(3%±RHP),
• HP=round(3%±floor(RHP)), or
• HP=round(3%±round(RHP));

where:

• HP = HP Cured
• RHP = Raw Healing Power (the "average" cure before deviation)

Data and Immediate Results

All data was collected on Thaumaturge.


Figure 1

The red encircled areas in figure 1 show how the closest integers for all these values are below the minimum RHP needed and above the maximum RHP allowed. We can therefore evidently rule out 4 of the 6 supposed constructions, leaving us with:

• HP=floor(3%±RHP),
• HP=round(3%±RHP),
  
• HP=floor(3%±floor(RHP)),
• HP=floor(3%±round(RHP)),
• HP=round(3%±floor(RHP)), or
• HP=round(3%±round(RHP)).

Graphing the Data


Figure 2

Figure 2 shows the expected RHP range for the supposed model HP=floor(3%±RHP) for the data from figure 1. The x-axis shows the MND values while the y-axis shows the RHP. I believe it's evident that no function without some sort of rounding/flooring can be found for this range (unless I'm missing something). We can therefore conclude that the model HP=floor(3%±RHP) is false.


Figure 3

Figure 3 shows the expected RHP range for the suposed model HP=round(3%±RHP) for the data from figure 1. Similar to the above case, we can conclude that the supposed model is false.

Conclusions and Closing Thoughts
All of the suggested models have been proven false. Evidently, some flooring/rounding must be done to reach integers as final values. My current belief is that the real model is somewhat like this:

HP=floor(3%±RHP), or
HP=round(3%±RHP);

where:
RHP=round(x)/2 (or 3, 4, 5, ...), or
RHP=floor(x)/2 (or 3, 4, 5, ...).

Of note is that the possible RHP range for MND34 in the given data is the narrowest and therefore the most accurate.

Edit: Crossposted on http://www.bluegartr.com/threads/119...24#post5994524

[Celaeris]

I'm going to point out that classes and jobs have different formulas for healing. I did some healing tests on WHM/CNJ and noticed that with the same MND and Det, you'd get different healing ranges. Be careful with testing stuff with classes, since it won't necessarily apply to jobs. For my test WHM/CNJ mind's can be matched with the DL pants to compensate for the loss of the job crystal, and this was done in 2.0.

[Nyaanyaa]

Thank you, Celaeris. Clearly more testing needs to be done and on different jobs/classes. I could see the denominator/multiplier that applies after rounding/flooring for RHP changing depending on job/class e.g., for THM it could be RHP=floor(x)/2 whereas for WHM it would be RHP=floor(x)/1; however, at the current point that's merely hypothesizing. I believe it at the current point more important, regardless, to find the similiarities rather than the differences in the formula for different jobs/classes.

[Deroth]

I haven't done any in depth checking in FFXIV, but I do remember from FFXI it was a player stats vs target stats calculation instead of the bracket approach many people seem to believe is in use for FFXIV.

I remember for a long time in FFXI many believed it was a bracket system based purely on the players' stats until some people had enough gear variance and alts to prove otherwise. As most FFs use the player stats vs target stats method it would not be surprising if they use the same system for FFXIV.

However, getting accurate formulas with the player stats vs target stats method is more difficult as we don't know the targets' stats nor if there is a variance on them.

[Kevee]

-Note above about classes/jobs different scaling.

-Also note, different attacks scale differently, especially with DTR(can't confirm if for STR/DEX/VIT). You need to test every attack. It's possible it's based off of potency, but I'm not sure.

See: https://docs.google.com/spreadsheet/...3SjEzeUE#gid=0

From chocobro, which I noticed you posted in.

Also, just ran the numbers for Fester.

There is probably a massive difference in the formula for magic users.

DTR: 262
INT: 494
WD: 69 Magic Damage
Formula puts Fester(300 potency) at ~577 damage.
Mine hits for ~735-780

Maim and Mend II?

Sigh.

I'm stupid.

Completely forgot about our passive boost.

That puts it in the general realm(750).


As another data point: Highest Fester I've seen is 804, lowest is 728.
Still currently on the dummy, but that's it for now.





ALSO, is the formula supposed to represent the average, or the base from which +/- 5% is supposed to come from?