[Armando]

First off these are NOT my formulas. My official stance on magic accuracy testing has always been that it's waaaaaay too much trouble for me to test, especially since I'm a non-caster. Still, I stumbled upon this on FFXIClopedia, and they seem reasonable, and apparently well backed-up by a lot of data some JP player collected. So here's the formulas in a nutshell:

(Total) Magic Accuracy = Spell's Accuracy + Magic Skill + [Stat Difference Contribution] + Gear/Merits

Stat Difference Contribution =
Your INT - Enemy INT if Difference <= 10
10 + (Your INT - Enemy INT)/2 if Difference > 10

Pretty straightforward. Basically, if the difference between your INT and the mob's INT is over 10 (which is pretty much always the case for any dedicated casting job in EXP, except possibly for Elvaan non-BLMs) every 2 INT is +1 Magic Accuracy. Just replace INT with MND in the case of the elusive WHM nuker.

NQ Staves = +20 Magic Acc
HQ Staves = +30 Magic Acc

Magic Accuracy -> Magic Hit Rate relationship

Magic Hit Rate will be defined as the probability of landing your spell unresisted. The odds of partial resists are based on that, and it's also pretty analogous to successfully landing a melee hit without it being stopped by Evasion, Parry, Shield, Block, etc.

If your Magic Hit Rate < 50%, +1 Magic Accuracy = +0.5% Hit Rate
If your Magic Hit Rate >= 50%, +1 Magic Accuracy = +1% Hit Rate

E.g. NQ Staves give +10% chance to land spells unresisted if your Magic Hit Rate is less than 50% (assuming that 10% doesn't bring you over 50%) and +20% chance to land the spell if your Hit Rate is over 50%. This certainly explains how I sleep DCs on PLD/RDM with Dark Staff.

Magic Hit Rate Formulas:
Hit Rate = 50 + (Magic Acc - Magic Evasion)/2 when Magic Acc < Magic Eva, until the 5% floor.
Hit Rate = 50 + (Magic Acc - Magic Evasion) when Magic Acc > Magic Eva, util the 95% ceiling.

Elemental Resistance is basically Magic Evasion, just like <Some Magic Skill> translates directly into Magic Accuracy. Additionally, each spell has a different amount of base Magic Accuracy (or the enemy has a different Evasion to different spells; numerically there's no difference, but it's easier to think of it as spells having different accuracies.) We know this intuitively from the fact that different enfeebles have different accuracies even against elementally neutral mobs. Apparently lower tier nukes are less accurate than higher tier nukes too. It's also known that your Magic Accuracy takes a hit (or the mob's Magic Evasion takes a buff) for each level the mob is higher than you; likewise, the mob gets a Magic Acc boost for each level it's higher than you, which is why we need Barspells, as strong as they are, to see any meaningful sort of resist.

Discussion/Applications?

If the formulas are valid, this puts a lot of stuff into perspective. For starters we have confirmation that 2 INT does indeed contribute 1 Magic Acc. Elemental Ninjutsu and Threnody are insanely potent: -30 Resistance from a Ninjutsu equates to +15% Magic Hit Rate when hit rate is bad, +30% when it's good, which means I was on to something when I was suggesting WAR/NINs and THF/NINs should use Hyoton before Distortion SCs. Death Blossom's -10 Magic Evasion provides the equivalent of 1/2 an NQ staff or 1/3 of an HQ staff to the whole party. Same thing for Enspell II's against their particular element.

So, who wants to check if the formulas are valid? Would be fairly straightforward - nuke something with Stone I enough times to get a reasonable number for your Magic Hit Rate (record the number of unresisted nukes), then vary your Magic Accuracy in any way and repeat against the same mob (if the mob won't last, pick another of the same kind at the same level.) Naturally RDMs are best suited to this because of their soloability. Obviously don't do this on Earthsday or Earth Weather. And obviously if you're going to use something like an EP mob don't come in with HQ Staff and +1million Elemental Magic Skill or you risk capping your hit rate.

Conclusions: Laevateinn still sucks without Aftermath. Liberator provides DRK with the equivalent of all 8 NQ staves simultaneously.

[Lunaryn]

I will note that Kamiel has been helping me test resist rates vs IT on Dark Magic (Specifically Drain II, as there's not a lot of available choices and the base damage on Drain varies too much to conclusively determine resist tiers), I have some preliminary data I hope to share soon, but I've been stalled a bit as I've reached a point where organizing in a spreadsheet is too much work, so I stopped to write a database frontend that'd work well for my computer situation (which turned out to include learning ncurses, since readline has the wrong sort of interface for what I need, libedit is too complex for my needs, and libkinput is just broken). Once I have things sorted out I will try to compare my results with the suggested formula.

(I've only been testing dark magic skill and nothing else changed (aside from game day, which I am tracking), but once I've got a satisfying amount of data on that I should try throwing in some tests for adjusting INT and maybe a few with Dark Staff for comparison.)

(Also, not thinking things through fully sucks. Drain II has a terribly long recast, but I'm kind of stuck using DRK/WHM since I started on that and /RDM gives MAB at my level. Oh well.)

[Yellow Mage]

Originally Posted by Armando

which means I was on to something when I was suggesting WAR/NINs and THF/NINs should use Hyoton before Distortion SCs

You mean Huton?

The perspective would be incredibly nice, though, if this turns out to be valid.

[Armando]

(Specifically Drain II, as there's not a lot of available choices and the base damage on Drain varies too much to conclusively determine resist tiers)

Is it possible to get consistent numbers on Drain II or do you just mean it's less random than Drain I?

You mean Huton?

Yes. Goddamned NIN spells are a bitch to memorize.

I've been digging around BG and this info was already discussed around December '08. Kaeko himself made the thread and he appeared to be quite convinced: "These tests were extremely well done with formal standard statistical measures. Trials exceeded 8000+. Thanks to both the translator and the original provider. It's really well done. For a statistics and lab rat nerd like myself this was a good read." You can find the thread here. I'm not quite good enough at statistics yet to appreciate the level of certainty in the tests, but if Kaeko says so I take his word for it. Doesn't take a genius to figure out that 8000+ trials = pretty damned accurate anyway.

Other random magic-related tidbits I found while browsing BG:
- It seems Spell Interruption Rate down doesn't have a cap - you can hit 100% and get guaranteed casts on any spell (assuming no movement or stun.) Apparently it takes around 102% due to the inaccuracy of the % values presented in equipment. Also, confirmed that Aquaveil is a static -25% interruption rate.
- You can equip elemental obis of the current day/weather to penalize the enemy's spells. E.g. Wearing the water obi during Waterday when taking Fire nukes.
- It seems that the +15% damage bonus that Circle abilities now give also applies to magic.

[Lunaryn]

Originally Posted by Armando

Is it possible to get consistent numbers on Drain II or do you just mean it's less random than Drain I?

It is FAR less random than Drain. When I went over my Drain damage, I attempted to normalize the tier estimates by taking an initial average of all the values that were at least 75% of the maximum damage I did, then made preliminary estimates of which resist tier each value was in by which side of each half-way mark they were on (i.e. 75%, 37.5%, 19.25%, 9.375% of the estimated average baseline). Then I estimated the baseline damage for each based on the estimated tier. Then I re-averaged based on the baseline estimates. I had to repeat this for many times as each re-averaging knocked some more of the results into the next resist tier. On one mob's results, this continued until the estimated average baseline was suspiciously higher than the highest actual damage result. On another, I failed to stabilize the results at all; the estimated average got high enough that I think it put the highest damage result as a single resist and suggested that there was a fifth resist tier (3.125% damage), which seem like very dubious notions on both counts. I basically figured I had to throw out Drain results as useless since I honestly couldn't figure out conclusively what any given damage number represented resist-wise.

Drain II baseline damage varies by a much smaller margin. While it does change somewhat with target level and randomly, it's consistent enough that most third-tier resists do from 31-35 damage on my targets, falling on 33 very commonly, and with lesser-tier resists being similar predictable ranges; it is possible to eyeball the resist tier without falling back on calculations and estimations due to this level of consistency. I've also noticed that Drain II seems to be incapable of hitting a 4th-tier resist; all damage seems to be 100%, 50%, 25%, or 12.5% of the baseline.

For reference, my targets thus far have all been Qiqirn Poulterers in Alzadaal Undersea Ruins, ranging from level 76 to 78 (wiki was wrong about this; I have since corrected it). We originally started testing on Magmatic Eruca at Mt. Zhayolm, but they proved too difficult to solo in our setup, so those results have been abandoned for now. I am verifying level via xp gain (Kamiel is partied with me on RDM75; all mobs con as Tough to him). Kumei herself is DRK66/WHM33, so all mobs con as IT to her. She has an INT of 73 which has remained unchanged during all testing. She has tested with both capped, unaugmented Dark Magic skill (227) and augmented with Chaos Burgeonet (232). I have recently acquired a Dark Torque, so I will most likely add some results with skill 239 for additional comparison in further testing.

At the moment I'm transitioning from a purely text log plus haphazard spreadsheets for resist rate calculations to using a database to track and sort data. Once I've finished setting that up, I may have some preliminary data to post.

Incidentally, do we have the info necessary to get an exact HP calculation on a beastman mob? I have to admit to being a bit underinformed on some of Studio Gobli's stuff, and I've been adding a bit to the complexity of my info in the hopes of getting accurate max HP figures; if I can get this without adding it up myself I can cut down on the complexity of my database a fair bit.