Magic Attack bonus vs Int gear

[delayedlaw]

Heyo! Delayedaru here, from Whasf Test server. rev 1612, Job 75blm/rdm. Ive been doing some testing with my blm on Burrower worms in Zeruhn mines, they are lvl 75 to 80. So I gear up with in gear, Mahatma stuff for the most part, full int build, earring, rings, and in all other spots. Shoot off some nukes Unresisted tier 4 elemental spells, hit for 1k-1.2k. So I switch to my MAB gear, Moldy earring, badass tattoo armor, and swap all the othe MAB stuff I can into my lil taru equip slots. Shoot off some nukes and they hit for 800-950 un resisted. The resist rate seems about the same. But I wasnt keeping track of that for this test. Looking at this blog, with retail testing http://tuufless.blogspot.com/2009/08/ca ... d-mab.html[/color], this quote stands out-

"Today, a fairly common rule of thumb is that “2INT ≈ 1MAB,” which has been largely shown to be true from general observation.

However, I have not yet come across any mathematical support for this statement."

Then he goes into the testing he did with complicated math formulas that make my head swim.

Though he adds the disclaimer, that is the standard formula for stat, to what it effects, 2str=1attk, 2dex=1acc, and so forth.

That being true when I gear up in my MAB gear my tier 4 nukes should be hitting for 1.1k-1.4k on unresisted, non crit nukes, which is a pretty substantial differance.

With the upgrades being made to mobs, and the "nerfs" (all of which I have fully have your backs on) to pots and what not, I think its time to make sure the job abilities/traits and gear bonuses are functioning properly to allow us to make up the differance in gear and macros. For example as far as I can tell Elemental Seal is not working, Nor do any of the Bard abilities, and 2 hour work (and a few spells what would be appreciated Elegy *cough* *cough*, but thats another post in another thread).

I thank you all very much for your efforts, and enjoy every min I spend in game. Great crew of people here, with way too much on their plates as far as projects, and with that its easy to get lost and over look important details. Keep up the awesome works.

D

[delayedlaw]

On a side note, probably off topic, the Skill+ gear does not seem to work, I have not noticed any difference in resistance rates when I have Elemental Skill+ gear build on either, nor a difference with Mag Acc gear. These are just as important as the MAB gear for endgame nuking activities. Particularly the Enfeebling Magic+ gear bonuses.

[PrBlahBlahtson]

Almost all of that is handled in /scripts/magic.lua if you're feeling bored.

LUA is pretty easy to read and write, too.

[codiddle]

This isn't a bug however, it's more of a complaint. Which is why you got the answer you should've been expecting when posting a complaint. The best place to take this complaint is to somebody who scripts, and they may just tell you the same exact thing. I had to learn that the hard way. But there are many little problems like this. The "nerfs" you're refferring to are actually "fixes", depending on which side of the glass you're looking through. The core has to work for progress to be made. Was this a good answer? I'm just getting sick of hearing so many complaints on how things are broken when they're just getting fixed. Anyway, I will take a look at the file tonight when I have time, as I do have a small amount of programming experience.

[PrBlahBlahtson]

Before you get too into it, codiddle, this was actually reported as a bug, and the dev team may have made some progress on it. Bug #85. If you are going to put some major progress on it, it would definitely be a good idea to touch base with kjLotus and see where it stands.

The perpetuation cost also (stealthily) added Elemental Affinity to status.lua and the item_mods table from what I saw, so that may be of use as well.

I've considered chipping at this myself, I just haven't worked up the ambition. Everyone insists that skill isn't being calculated, but when you look in magic.lua, there's a function that does pull and total mod_elem (or whichever skill is in use.) The function gets called for the calculations, so... I dunno ¯\(°_o)/¯

[kabitokaii]

just looks like the gear with +skill isnt getting into the mix to make your skill good enough to not get resisted at HNM levels.

[PrBlahBlahtson]

just looks like the gear with +skill isnt getting into the mix to make your skill good enough to not get resisted at HNM levels.

I'd really, really recommend looking in magic.lua under applyResistance. Magic Skill is the least of your worries when it comes to landing a nuke right now. I'm not sure which formulas were used to calculate magic hit rate, but they don't match either wiki when you break them down.

They do not calculate your stat vs your target's stat.

They also penalize the player by always calculating your accuracy with a 1.0 multiplier, instead of favoring you with a 0.5 if the target has a higher MEva than your MAcc.

They do slant pretty heavily against you if you're targeting a T+ or anything over L83. (B skill MDef instead of C.)

I may take a stab at rewriting applyResistance on my own server, just to see if I can figure out why things are the way they are. It'll be Dwarf Fortress-style fun.

Notes:
Black = Int, White = Mnd, etc
M1 = Caster Stat - Target Stat if M1 <= 10, else M1 = 10 + (Caster Stat - Target Stat)/2
M2 = MAcc from gear/merits
MAcc = Skill + M1 + M2
Hit rate is floored at 5, capped at 95.
MA = Caster MAcc
ME = Target MEva
If MA <= ME-90, Hit rate = 5%
If MA <= ME, Hit rate = 50 - 0.5*(ME-MA)
If ME <= MA, Hit rate = 50+(MA-ME)
If ME+45 <= MA, Hit rate = 95

Hit rate = p
1/2 = p* (1-p/100)
1/4 = p*91-p/100)^2
1/8 = 100*(1-p/100)^3

Issues:
MEva is unknown on Darkstar. Yay! The C/B precedent carries forward, cause I'm lazy!
It's speculated that there's a level adjustment in the formula. Which is unknown. This favors players, until monsters get nukes. Then it probably still favors players. Must spite players somehow.

Edit:
Hey. Question. Why do all of these formulas assume capped skill? Not cool ಠ_ಠ

Edit edit:
Taking donations of a BLM spellcast XML for testing, 'cause... well, yeah. Lazy, and all the ones on Windower wiki are 80+.

[PrBlahBlahtson]

Okay, so drafted up an applyResistance that uses only the available formulas on FFXIclopedia (hereafter lolwiki, because nobody updates it) and wiki.bluegartr.com (BGWiki). It's kind of messy, but it was fun.

The current magic.lua will double macc if Elemental Seal is present. lolwiki offers no info on elemental seal. BGwiki has no hard data, but speculates that Elemental Seal adds approx 256 acc. Which is effectively the same, once you cap everything.

There is a speculated level correction, but neither wiki takes a stab at it. If it follows melee accuracy, you're looking at the equivalent of leveldiff * 4 to the acc vs eva check, or leveldiff * 2 to the hitrate percentage, since 2 acc = 1% hitrate for melees. I left it alone for now, due to lack of data. Old formula would give you a 5%/level increase in acc, or a 10%/level decrease. Anyway, yadda yadda, back and forth, here's some output.

In a damage set put together by a total moron (that's-a me!), hit rate against Fafnir was essentially capped by a L75 tarutaru BLM/RDM. The INT difference contributes 12 acc, the staff gives 30, all vs 286 MEva. Raw hitrate is 96.36%. 4.50% chance of a half resist, 0.23% of a quarter, 0.000125% chance of an eighth. tl;dr: You probably wouldn't see a resist in some fights.

Set is HQ Staff, Bugard Strap +1, Demon's Helm, Prudence Torque, Loq earring, Moldy, Oracle's robe, zenith +1 hands, Diamond ring, demon's ring, no back, Hierarch belt, Nashira legs, Herald's feet. Told you it was bad.

Naked taru BLM, capped skill, is 278 Acc vs 286 MEva for a hitrate of 44.93%. 11.1% chance of half resist, 13.5% chance of quarter, 0.16% chance of eighth resist.

Sorry, but the raw formulas are waaaaaay too soft on players. Needs some level correction.

Edit:
More crappy tests. L75 Mithra RDM/BLM, capped skill, +22 Elemental skill, +15 INT, HQ staff, no other actual MAcc. 299 Acc vs 286 Eva. 63.36% hit rate, 15% chance of half resist, 8.5% chance of quarter, 0.05% chance of eighth.

Formulas act a bit odd, even though they check against the wiki. Going to run them through some spreadsheet checks, see if I can pin down what's going insane.

[delayedlaw]

Thank you for the updates to this post. I am sorry if I came off as a complainer, that was not my intent. I had data behind a bug, or problem however you want to look at it and posted accordingly. I can be a bit long winded, and maybe add comments more for entertainment value, than for actual productivity. As I said, Im all for the fixes, "nerf" was the wrong word I suppose. Ill be happy to do any nuke tests or gear tests needed on the BLM side of things while I work on my SMN and RDM. Thank you all again for all the work behind the scenes of this project.

D

[PrBlahBlahtson]

Nah, I'm having fun with it now that I understand the formulas.

[used to be some preaching that current formula is wrong. After actually looking at the output, I stand corrected. My formula was wrong though! I can fix it.]

You just made me find a bug. This is good :p Unfortunately, my formulas occasionally go crazy and produce resist rates literally over 9000%, so... I have some investigating to do.

Edit:
Decided to compare the old formula and new formula using a "naked" stat set and geared stat set. I also added in the level correction used in the old LUA, and a "rough" level correction that penalizes 4 MAcc for every level below the target. I want to stress that I have no idea if that's the correct formula. The correction also assumed L88 Fafnir on both sides, which is the best case scenario for players.

I only noticed now that I neglected to multiply the old formulas by 100 to convert a decimal to a percentage, which makes the numbers look a bit uglier. Please do so in your head.

"Real Formulas" present the individual odds of that resist rate on the left, and the total odds on the right.

I also apologize for using "Real Formulas." "Adjusted Formulas" would have been more... neutral.

Naked: 81 INT vs Fafnir's 82, elemental skill of 256, no MAcc gear.

Naked comparison.

Capture.PNG (6.55 KiB) Viewed 8172 times

Geared: 96 INT, elemental skill 286 (no elem skill gear), 30 MAcc from staff.

Geared comparison.

Capture2.PNG (6.7 KiB) Viewed 8172 times

Not pictured: Geared, with 338 elemental skill.
Adjusted formulas: Half 4.75, quarter 0.2375, eighth 0.0125, 95% chance of unresisted.
Old formulas: Half 95%, quarter 90.25%, eighth 85.73%, sixteenth 81.45%, 5% chance of unresisted.

Not included is a lot of cells that show the numbers progressing. The 338 elemental skill is still floored at a 5% hit rate. So in summary, it's not that magic skills are calculated incorrectly; it's that the current level correction completely whips your ass in a very fast way. -10% accuracy * 10 levels = -100% accuracy. You're floored to 5%, you can't win.

I've never danced with Fafnir on retail personally, nor have I played a magic-heavy job. DRG, BLU, DNC, and SMN for me. I have no idea if my formula even approached realistic resist rates. I'll keep the code around if anyone wants to just tweak it and implement.

Edit edit:
Reconsidered my previous "something is wrong." That chart is wrong, I'm okay. Level adjustment remains an unknown.