dex/crit relation, WS crits, WS gorgets discussion

[Nagamaki]

OK so on the topic of the Al'taieu gorgets. I've tested most of what I wanted to look at except a possible effect on pdif, I have some indication on that but since the WS I used for tests was Jin (multihit and can crit) getting a really accurate pdif from the WS dmg (and seeing if it matches the normal melee pdif) really isn't accurate enough. I'll just use single-wield and Blade: Ten at 100% for that.

The first thing I wanted to test on the gorgets was getting a true 100% answer on whether or not the fTP +0.1 bonus applied to every hit of a WS or only to the first. There's been some questions on how that could be tested in the past, but the answer became obvious because of the crit WS tests I had done. The problem with tests was how to rule out pdif variation in the tests. The key is that crits with fully capped pdif will have a pdif of 3 half the time. Using a 2hit crit WS with Sneak Attack, that means every other WS where crit kicks in on the second hit should be a double 3.0 pdif WS. Backhand Blow happens to have a stupidly high crit rate (about 70% when regular melee crit rate is at 30%) so getting a few double capped crits wouldn't take too long.

So anyway, damage calculations using Destroyers and dex gear said that BB with Breeze gorget (the .1 fTP) should hit for 392 at 3.0 pdif, and BB without gorget should hit for 360. I confirmed both of those using Lower delk giant bst bat pets (they are just as retarded weak as bunnies so pdif will cap the same way, but you can use the stuff around to build tp instead of all 1hit kill lv0 mobs. After confirming that, I tested SA BB with the same gear (no gorget) on the giants themselves (so they'd survive 1st hit) and quickly saw a roughly 1 in 3 pattern of the WS doing 720 dmg as expected. Then, put on Breeze and repeat the same process, and 8 of 20 WS hit for 753 dmg, none above. 753 = 392+360+1 (the 1 is from the added up decimals before rounding).

So that test clearly proves that on BB with Breeze Gorget, the first hit has a fTP of 1.1 and the second hit has a fTP of 1.

The only flaw I can see here is that one could think perhaps the second hit of BB being a offhand hit would be the reason, that theory isn't supported by the damage ranges I've seen in other tests though.

The result of that test made me pretty concerned about whether or not the acc applied to all hits as well, and just how much acc it was. There have been reports of very good results on acc by rangers but ranged WS are single hit, so those say absolutely nothing about the accuracy applying to multihit WS.

To try and test acc I picked Groundkeepers, because they have a reasonably small level range that includes lv75, so I could ensure no variation bu only fighting EM. I did two sets of test, one with acc well below the number needed to make them switch from neutral evasion to low evasion, and 1 set with just slightly less acc than needed to make them switch. Each specific test was done using 30 repetitions of Blade: Jin (7-8 keepers, thanks to horrible attack they each lasted around 4 ws).

/c with...
+13 acc = neutral
+14 acc = low

This is the gear used in the first test :

Mamushito +1*2
Ungur Boomerang
Walahra Turban
NECK SLOT
Reraise Earring
Loquacious Earring
Ninja Chainmail
Seiryu's Kote
Rajas Ring +3 (uneven base dex)
Behemoth Ring -11
Forager's Mantle
Swift Belt +3
Ninja Hakama
Fuma Kyahan

So this set is acc-5, 19 less than the amount to /c low eva.

Acc -5, no neck

Regular melee acc : 73.1%
WS acc : 80%

Acc -5, Thunder Gorget

Regular melee acc : 72.8%
WS acc : 80%

Acc -5, P.Charm (so +5)

Regular melee acc : 76.7%
WS acc : 77%

After those very inconclusive results boosted acc a bit to be near the /c changing point. I just put a Jaeger instead of the Behemoth Ring, so acc went from -5 to +10.

Acc +10, no neck

Regular melee acc : 80.73%
WS acc : 88%

Acc +10, Thunder Gorget

Regular melee acc : 78.5%
WS acc : 84%

Acc +10, P.Charm (so +20)

Regular melee acc : 83%
WS acc : 84%

I did one last test with some more stacked acc just to have an idea of how much was needed to get around 95% ws acc. I kept Jaeger and used both SH and charm.

Acc +30 (charm)

Regular melee acc : 88%
WS acc : 94%

So conclusion on the acc effect of gorget, not absolutely fully certain because 30 WS (120 potential hits) is a pretty small sample, but at the very least it is clear that if there is any effect, it is extremely minimal.

[VZX]

/littlehijack

I sent Aurik a PM...
It seems Gorget pDIF bonus is more than 0.1, assuming Ruke's test shows all the lowest possible number as well as Studio Gobli's monster VIT/def data and pDIF min function are correct.

[Zirk]

I rarely post (heck I've made up 4 accounts over the past couple of years because I can't remember my passwords in between posts) but I have read this thread from start to finish and am just wondering why nobody has tried this in ballista.

Granted I don't do ballista, so I don't know the 'rules' it plays by. However, it would be easy to manipulate the targets AGI to see the role DEX vs. AGI plays (or DEX vs. whatever). With a couple of mages doing some outside curing and a low damage weapon you could autoattack and simply sit back and collect the results I would think. Throw in a target 5 levels lower and adjust equipment to give him the same stats as the original target and you could even see if level difference plays a role.

Anyway, suggesting something I have no experiance in may not be wise, but it seems that if all other things are equal then it would be the perfect testing ground for something like this.

Damage calculations are totally different in ballista, there could be hidden crit rate increase or decrease, or they just use a different formula for critical hit occuring, or there could be a much lower crit cap.

I'm pretty confident mob AGI has an effect, which I suppose ballista might be able to help confirm, but carefully evaluating crit rate increase/decrease as each point of AGI/DEX is subtracted/added could be meaningless for normal game play.

And this is very unfortunate that it's turning out WS gorgets have hardly any accuracy bonus for multi-hit WS

[Thitaubian]

...And this is very unfortunate that it's turning out WS gorgets have hardly any accuracy bonus for multi-hit WS

I don't think we should jump to conclusions on this.

If you look at Nagamakis tests Acc+10 no neck parsed higher than Acc+10+P.Charm
This can only mean that the sample size was so small that the variation is beyond indicating a trend.

More testing would be necassary to know anything.

[Nagamaki]

About the sample size of the acc tests, I do agree that it is small, however the overall trend in results does show them behaving mostly as one would expect them to (exluding tests with gorgets, I was expecting/hoping very hard for a significant boost).

The acc+30 (rajas swift jaeger charm harness) test was done to make sure a moderate amount of extra acc would take WS acc to at least, and the result with it seem to achieve that. I will do more repeats on some of these tests to get a bigger sample, getting a truly big enough sample would be extremely time consuming though. The test is pretty easy, but while EM Groundkeepers can't hit the wide side of a barn, they can still blow up, and fighting them with crappy acc/att on purpose gets pretty damn tiresome, not to mention having to zone sections of them when you can't find any EMs anymore.

Anyway, here's the Ten test about pdif :

Gear

Mamushito +1 (single weild)
Ungur Boomerang
Walahra Turban
NECK SLOT
Reraise Earring
Loquacious Earring
Scorpion Harness
Seiryu's Kote
Rajas Ring
Jaeger Ring
Forager's Mantle
Swift Belt
Byakko's Haidate
Fuma Kyahan

STR88
DEX87

Attack 351

/c with these as my attack
316 High Def
317 Neutral
397 Neutral
398 Low Def

So according to the accepted /c info, that means they have 317 Def (since high def is as soon as your att drops below their def).

351/317 = 1.107

So the pdif assuming the equation is direct (EM so no level correction applied) should be around 1.107.

Regular melee avg throughout all my tests (and I made sure to never have any attack or str variation) was 45.4, over thousands of swings so that number is most likely quite exact, so I used it to figure out fSTR.

45.4/1.107 = 41.01

Also for reference as it may help ppl check the pdif, melee hi/lo was 59/34, and crit hi/lo 103/71

Mamu+1 have a base D of 38, so that means fSTR sitting at 41-38 = 3.

WSC with 88 STR and 87 DEX for Ten is 43, so D+fSTR+WSC is at 84.

84*2.5 = WD of 210 without gorget
84*2.6 = WD of 218 with soil

59/41 = 1.439 high melee pdif
34/41 = 0.829 low melee pdif


Without gorget calcs using melee pdif:
High 302
Low 174
Avg 232

With gorget calcs using melee pdif:
High 313
Low 180
Avg 241

Results over 25 WS, no neck

215
248
212
195
222
212
218
276
295
213

275
232
202
213
214
212
215
212
212
183

212
182
212
281
283

High 295
Low 182
Avg 225.84

Very high frequency of 212

Results over 25 WS, Soil

221
284
222
220
221
222
203
265
277
191

225
203
221
282
221
258
236
190
218
221

300
221
228
222
267

High 300
Low 190
Avg 233.56

Very high frequency of 221

Comparison

No neck

Calc
High 302
Low 174
Avg 232

Result
High 295
Low 182
Avg 225.84

Avg 7 points lower, highs and lows within the boundaries

Soil

Calc
High 313
Low 180
Avg 241

Result
High 300
Low 190
Avg 233.56

Avg 8 points lower, highs and lows within the boundaries


So my conclusion here, pdif seems unaffected, at least on EM mobs. If there is any pdif effect, it would have to be a modification to the level correction that applies when the target is above your level.

[Terrance]

S-E seems to be fond of "stat vs. stat" calculations- that there'd be some sort of "My DEX higher than your AGI" crit % calculation doesn't strike me as odd.

I had thought I'd read it was VIT. AGI would make thieves hard to hit and hard to damage and paladins easy to hit and easy to damage. Weird.

[VZX]

Regular melee avg throughout all my tests (and I made sure to never have any attack or str variation) was 45.4, over thousands of swings so that number is most likely quite exact, so I used it to figure out fSTR.

Is that the most frequent number popped too?

[Kyreth]

S-E seems to be fond of "stat vs. stat" calculations- that there'd be some sort of "My DEX higher than your AGI" crit % calculation doesn't strike me as odd.

I think its been generally accepted for a while that your critical hit-rate is determined by your DEX versus the mob's AGI. I've never seen actual mathematical equations and the like to support it, but there is some evidence available that does in a sense. IE, high/capped critical hit-rates on weak mobs and extremely low critical hit-rates on HNMs.

Right. I'm thinking that much like debuffing MND off a mob to enhance slow, pulling AGI off it might mean a lower level of DEX is needed for the "crit tier" some folks have been theorizing here.

Something like DRK ABS-AGI, etc. would increase everyone's critical rate, while ABS-DEX would only boost the DRK's (and critical hit % down the monster). Likewise, the BLM elemental debuffs would have subtle but useful effects (crit % on monster up, crit % of monsters down, etc.) that might mean it's worth blowing a few MP to keep elemental DoT's active on a monster.

[Nagamaki]

The mode (the most frequent value in the data set, had to look that up, not used to detailed statistical english terminology) was 42.

42/41 means the mode pdif was 1.024.

Applying this mode pdif to the WD values for 210 and 218 means the modes on the WS dmg should be roughly 215 (was 212) and 223 (was 221). Not an exact match, but there's room for rounding error since 42/41 is rather close to 1 in truncated calculations.

[Valyana]

I'm pretty sure that mode pdif is always 1.0 if you're between 0.5 and 1.5x mob DEF. Using that to determine fSTR should give more accurate results.

[Nagamaki]

If that's true, and it appears accurate, then the avg I used to calculate fSTR gave the wrong result because it was dragged down a bit by of the mode.

Using fSTR of 4 (42-38) to redo the calculations, we get...

D + fSTR + WSC
38 + 4 + 43 = 84

85 * 2.5 = 212

Mode 212
High 305
Avg 234
Low 175

85 * 2.6 = 221

Mode 221
High 318
Avg 244
Low 183

So this confirms the modes, it increases the gap between the result avg and the straight pdif avg, but it also explains that gap because of the forced 1.0 mode. Highs and lows are still within the calculated boundaries.

[Kawfee]

On the topic of just DEX affecting critical hits during regular DDing (as in, not during WS), I just wanted to point out something that keeps bouncing in the back of my head.

Now, maybe (probably) this already been disproven, but I haven't seen any evidence to that effect in this thread (I don't think), so, with risk of sounding like an idiot:

Is it possible that 1 DEX simply equals .1% crit rate? So 10 DEX = +1% crit, or 120 DEX = +12% crit.

Reason I'm asking is I've seen several elements of evidence for this:



Example 1:

NIN in Thickets at Mamool Ja Staging used Kikoku + Unsho and had 77+53 dex eating Sole Sushi ( 130 dex total ).

Mamool Ja Blusters: 633 swings, 94.06% accuracy, 20.22% crit rate. The crit rate was pretty consistently around 20% on every mob killed.

Mamool Ja Philosophers : 295 swings, 92.19% accuracy, 13.56% crit rate. On one philospoher the crit rate was 20% but the others all hovered around 11-12%.

Sea Puks : 367 swings, 95.57% accuracy, 17.71% crit rate. On the 2 highest level Puks, based on xp gained, the crit rate was actually in the 11-12% range again but was 18-20% against the others.

Here, just looking at it quickly, it would seem that on average the crit rate overall was around 17%-ish. Since she wasn't using Senj/Fudo, the only modifiers to crit would be DEX, and merits.

130 DEX = 13% + 4 merits (4%) = 17%...?



Example 2:
Also, I've seen something like this first hand. Unfortunately I didn't save the entire parse :/ but I know that the crit rate was 25.xx%. Basically, the NIN was wearing a full dex build, + senj/fudo, with capped crit merits.
So, say ~120 DEX (to be conservative).
That's 12% from DEX + 6% Senj + 3% Fudo + 4% merits = 25%.



Example 3:

Bleh, this parse is going to be worthless. Need a lot more DEX to get this to work at the Bhaflau Thickets (mamool staging point exit) camp, 121 is too small. I'm a little frustrated too because there's only three PTs in thickets and for some reason we went to the mamool staging point part, where the other 3 PTs are. Colibri is empty, but for some reason we're still here.

I'll upload the parse later, still in the PT. The setup is BRD, MNK/NIN, WAR/NIN, RDM, BLM and myself. A little over an hour into the PT and critical hit-rate is fluctuating around 14-15%; definitely not a tier here. Works out to about a 3-5% improvement in critical hit-rate in adding 30 DEX.

121 DEX = 12% + 4% merits = ~16%. His parse is pretty close to this, given allowable sample size errors.
Also to note 3-5% +crit rate from +30 DEX.



Some evidence against this would obviously be the people parsing around 29%, but the only evidence I've seen of this firsthand (not saying it doesn't exist just saying I don't have access to look at it) is the OP's post, but I'm a noob when it comes to MNKs :X




Also, a note to Nagamaki's earlier question on crit rate vs. Ranged attack. While your tests about the effects of gear show that that particular equipment more than likely has no effect, one thing I'm almost positive does affect R.Crit is Rogue's Roll from COR.
I'm currently static'd at ~lvl 60 with a RNG and a COR (and THF), and me as NIN. And while afaik the RNG has always outparsed me on Crit Rate, I believe that when we use Rogue's Roll, both of our crit rates go up (and yes, I'm referring to his ranged crits, not just melee). I've never really paid close attention to it though, so I could be wrong, and this doesn't really add much to the discussion... just pointing it out for possible future use. Next time we static I will try and see if it makes a difference, but my party probably won't want to go for long without rogue's roll



And, if anything I said is just plain stupid, feel free to disregard me as another nooblurker.

edits: formatting problems

[RKenshin]

That theory also ties in with what my results were in my test with/without Byakko's Haidate... As mentioned earlier in this thread I recieved a 1-2% critical hit-rate increase from Byakko's Haidate, and using your theory the 15 DEX turns into 1.5% critical hit-rate.

But, I think there's still an AGI factor in it somehow... In all the exp PT and HNM fights I have parsed my critical hit-rate is almost always lower on HNMs. I'm not the only one with those results, right? I suppose it could also just be me, since my sample size of HNM parses vs exp PT parses is rather small.

Then, it also leaves questioning for the sudden jumps in critical hit-rate (the theory about the tier). I think we still need a lot more info still.

[Nagamaki]

69 DEX (Elv MNK/THF base)
9.3%

89 DEX (Byakko's, Rajas)
21.2%

105 DEX (Byakko's, Rajas, Osode, Squid)
23.9%

100 DEX (Byakko's, Rajas, Thunder, Squid)
23.7%

This is against the TW Lufaise mobs I used for early cap tests.

It's pretty clear that crit rate isn't DEX/10+crit bonus items/merit, or something like those rates would never happen.

Neither would the rate hikes observed on around 120 DEX.

What needs to be defined is the exact DEX vs mob stat relation, but it is 100% clear that crit rate is not some flat value.


About ranged crits, I don't doubt that rogue's roll, and most likely crit merits affect ranged crit rate, its just clear that Thunder Staff doesn't. I only tested that to see if if Sniper Shot would be a good test, but Backhand Blow turned out to be a good test subject for WS crits.

Finally, I decided to do practical Gorget vs Torque tests with AF, using Enkidus (cause they're stupid easy, reasonable solo xp and drop Diorite). I want to get a really good size data sample so I won't be posting results for awhile more, but what I got so far is rather extremely underwhelming when it comes to the gorgets.

[suji]

What needs to be defined is the exact DEX vs mob stat relation, but it is 100% clear that crit rate is not some flat value.

I've been pretty reluctant to mention this here due to the standard knee-jerk lolalla sentiment, but I've been helping Kwontos with some testing to try to derive this relation:
http://ffxi.allakhazam.com/db/jobs.html ... =42;page=1

But it's pretty much trying to answer that exact question. The gist of the study is to figure out the curve for Robber Crabs in Kuftal, and given that DEX vs AGI determines the crit curve, we should be able to find some crit% data points from merit mobs to determine how to shift the curve so that it can apply to merit mobs.

Then folks would hopefully have a really good idea of how much DEX they need to add to get a desired crit percent. So far the test has just under 30,000 samples at various values for DEX. The only real flaw with the methodology is the level variance in the crabs, but this still gives one a pretty good idea of how the curve is starting to look.

[Nagamaki]

Did a quick test just in case the exact stat involved in crits on the target's side was still in question..

MNK/DRK, DEX67+10 for the whole test, using Faiths.

Test A

300 swings, crit rate 12.1%

Test B

Absorb-AGI (target AGI-18) to pull each mob, 300 swings, crit rate 24.2%

Tiny sample, but still pretty obvious.

[Zirk]

Did a quick test just in case the exact stat involved in crits on the target's side was still in question..

MNK/DRK, DEX67+10 for the whole test, using Faiths.

Test A

300 swings, crit rate 12.1%

Test B

Absorb-AGI (target AGI-18) to pull each mob, 300 swings, crit rate 24.2%

Tiny sample, but still pretty obvious.

Please tell me you did this test on the Lufaise Meadows sheep/bees.

[Nagamaki]

Yes, bees/sheep were the targets.

If the DEX AGI check is a substraction, abs AGI might have a similar effect on high level mobs assuming your DEX is right on the edge of their crit tier (which is obviously higher than lufaise crap) before the -18 is applied.

[Zirk]

Bah, I should have worked out my numbers first. I was comparing your results from your absorb-AGI test to this test. The goal was to investigate which seemed the more probable relationship in the crit chance calculation: playerDEX - mobAGI, or as Ruke sorta kinda proposed, playerDEX - 2x mobAGI.

Adding 18 dex to your 77 dex from Test B would put you at 95 dex, which according to the test I linked would put you at between 21.2% and 23.7% crit rate.

Adding 36 to 77 would put you at 113 dex, which should put you between 23.9% and 24.9% crit rate. The actual crit rate from Test B being 24.2%, supporting a DEX - 2xAGI calculation being involved.

However, after making some other comparisons (in particular, comparing your crit rate in Test A being 12.1% with 77 dex, and your crit rate being 9.5% with 79 dex from the earlier test) it is apparant that the experimental error involved renders these findings meaningless.

Unless for some reason subbing thf has a hidden bonus to crit rate or you happened to be using a different weapon in both tests.

-edited for clarity.

[Kyreth]

Yes, bees/sheep were the targets.

If the DEX AGI check is a substraction, abs AGI might have a similar effect on high level mobs assuming your DEX is right on the edge of their crit tier (which is obviously higher than lufaise crap) before the -18 is applied.

Gives your DRK's something to do besides Stunning or Krakening, too. Keep tapping AGI off the mob and see if it doesn't up everyone's crit rates. (Obviously, BLM debuff for good measure- that's a pretty huge swing while it's active). I'm a n00b for DRK- how do ABS-xxx spells stack?

If doable, ABS DEX + ABS AGI might well = critfest for DRK.