Spell interruption down cap and Aquaveil Tests

[Spekkio]

I don't think pchan disagrees here. Or else you and I disagree and I wasn't clear. What I'm saying (and what I think you're saying and pchan is saying) is that it seems most likely to me that the way it works is (chance to interrupt based on magic skill vs. enemy's relevant attributes/skills/levels) * (1 - (% interruption rate from gear + % interruption rate from aquaveil)) = (total interruption chance). Some other people were saying it was maybe additive and I was saying I doubt that. Pchan seems to be arguing with the people who doubt that Aquaveil is 25% (and I'm personally comfortable with saying it's 25% for now given what we have and what we could work with, as well).

i came to the conclusion that he believed them to be additive based on his last post where he was letting x be enh skill, y be ninjutsu, and z be aquaveil's effect and adding x to z as well as y to z and trying to draw a conclusion. that would only be valid if the effects of skill were additive to aquaveil's -interruption, not multiplicative unless i TOTALLY spaced on what he said. did i goof?

[Suiram]

Hmm, maybe I'm the one who goofed? In the OP pchan says that skill "doesn't provide extra spell interruption" which I took to mean that he's saying adding X skill won't accomplish the same as adding Y "spell interruption rate down" which is the same as saying it's not additive and which you and I agree on. That seems to be what he and Dawin are arguing about but to be perfectly honest it's not entirely clear to me what either of them are trying to say in their most recent posts so I could very easily be wrong here.

[Dawin]

142 skill difference gives strictly less than 1%, which means that skill doesn't affect your chance of being interrupted, at least when what you get from gear/spell is close to cap.

Jesus christ you're retarded.

What the others are saying is likely to be true.

Some formula determines how often a hit interrupts you.
Another formula determines what % of the time that is ignored.

Multiply the two together.

The second number can = 100%, so all testing where it is 100% are irrelevant, i.e. all the tests you are basing your conclusions off of.

Edit: There may also be a tier system with either or both of those variables that's truncated down to something smaller than 1/1024

[Byrthnoth]

What seems strangest to me is that it caps at 102% at all.

- clubs x2 (25%) - 64/256
- muse tariqah (10%) - 25/256
- willpower torque (5%) - 12/256
- magnetic (8%) - 20/256
- rdm AF1+1 body (12%) - 30/256
- rdm AF1 body (10%) - 25/256
- 3 belts (10%,6% and 8%) - 25-15-20/256
- karasutengu (15%) - 38/256
-8% from merits - 20/256 calculated individually or not.


102% : dual clubs,torque, earring,NQ body, 6% belt, karasutengu - 258/256 rounding down everywhere.
101% : dual clubs, earring,NQ body,10% belt,karasutengu - 256/256 rounding down everywhere.

Why is that? Seems to imply that you have to have over 100% reduction rate before it works. It seems possible that you guys are somehow perverting an equation or function via an unanticipated negative term. Like if it uses a lookup table to determine the skill/level correction and you're suddenly giving it an input value outside of its range.

[ringthree]

The point is more that just as with so many other stats in the game like haste and damage reduction, the values reflected in the item description aren't the exact values on the piece. A 10% piece is really 100/1024 or 25/256 or something. So it's more than you need 102% in item description spell interruption rate to achieve 100% true spell interruption rate.

Also it is interesting (although not entirely new) that interruption rate behaves peculiarly just below the cap, but I think the most important part of what pchan has shown is that you (some jobs anyway) can achieve a gear-only 100% spell interruption rate, and that there's not a seperate gear-cap/magic-cap thing going that caused earlier 100% spell interruption builds to fail. Whatever use that may actually be is up for future debate, I guess.

Pchan has shown that there may be a cap at some point for some spells that may or may not be dependent on skill.

We also know that below that point things go very wonky, and it could be because of a skill check difference or that the evidence from Pchan's observations does not match reality.

Definitely a good start for understanding the nature of spell interrupt, but there are many more factors at play here, and the conclusions that have been drawn do not match the observations. For example, if skill is a check and it is a comparative check (based on level or skills or something we don't know about yet) these tests only tell us about one level of those checks. To check this you would need to check on mobs of a different level range.

[Dawin]

101% : dual clubs, earring,NQ body,10% belt,karasutengu - 256/256 rounding down everywhere.

I checked your math. It seems correct.

Also if we use 1024 instead of 256, the 101% comes out to 1031/1024 rounding down everywhere which you would assume would lead to 0% interrupt.

[Dawin]

If we assume clubs are 256/1024, and the rest of the numbers are 8% = 80/1024, then it actually works out that 102 is needed and 101 is not enough. Of course rounding 8% down to 80/1024 sounds retarded on SE's part, but I wouldn't put it past them.

[pchan]

Jesus christ you're retarded.

What the others are saying is likely to be true.

Some formula determines how often a hit interrupts you.
Another formula determines what % of the time that is ignored.

Multiply the two together.

The second number can = 100%, so all testing where it is 100% are irrelevant, i.e. all the tests you are basing your conclusions off of.

Edit: There may also be a tier system with either or both of those variables that's truncated down to something smaller than 1/1024

Nah. That's not what you were arguing. Let me quote you :

Stoneskin(76% + aquaveil) : 17 / 6

Utsusemi(76% + aquaveil) : 4/2

Utsusemi(101% + no aquaveil): 7/4

These all should be 99% (not the ratios, but per hit). When you factor in that stoneskin takes 2.5x as long to cast, it should have MANY more interruptions, but instead has far less (17/6 vs 11/6). This seems to be proof that skill level does matter.

Do you use any logic at all when making your conclusions?

You are clearly saying that at 99% the skill difference has an impact whereas it has none (lol sample size).

[pchan]

it seems most likely to me that the way it works is (chance to interrupt based on magic skill vs. enemy's relevant attributes/skills/levels) * (1 - (% interruption rate from gear + % interruption rate from aquaveil)) = (total interruption chance)..

Maybe, but impossible to test. In practice spell interruption is only usefull when stacked anyway.

[Dawin]

You are clearly saying that at 99% the skill difference has an impact wherease it has none (lol sample size).

If the higher skill spell had 75% as many interruptions, you would be right: lol sample size. But when combined with the fact that it takes 2.5x as long to cast, then the sample size becomes pretty close to sufficient to show there's something different going on.

Factor that in with basic observation that you should have had over the years playing FFXI to show that you get interrupted less after skilling up, and a reasonable conclusion can be drawn with very little doubt.

[Kirschy]

Interesting test + results~ Figured I would play around /w possible values on the equipment based on my experiences with haste and damage reduction:

77% : club+shield,torque,earring, ,6% belt,karasutengu

First assuming the /256 system:
(Estimating values of course.)

Club: (25%) 64/256
Shield: (10%) 25/256
Torque: (7%) 17/256
Earring: (8%) 20/256
Belt: (6%) 15/256
Feet: (15%) 38/256
Merits: (8%) 8/256

Total: (79%) 187/256(~73%)
(BTW, You said 77% but it adds to 79%. Did you write down the numbers wrong on forum or did you do the test with 79% accidently? I've done that before =x.)

Aquaveil would need to be like... 69/256+ (~27%+)

Assuming the /1024 System:
(Again estimating values):

Club: (25%) 256/1024
Shield: (10%) 102/1024
Torque: (7%) 70/1024
Earring: (8%) 80/1024
Belt: (6%) 60/1024
Feet: (15%) 152/1024
Merits: (8%) 80/1024

Total: (79%) 800/1024 (~78.125%)

Aquaveil would need to be... 224/1024+ (21.875%)

It sounds likely that Spell Interruption works on the /1024 scale. But it's possible that my estimated values are off. I'm gonna mess around /w that 20% ninja earring and see what haste it gives~

[Dawin]

Shield: (10%) 102/1024
Torque: (7%) 70/1024
Earring: (8%) 80/1024

These numbers don't make sense unless you are assuming SE error/weirdness in rounding (which isn't unreasonable).


Like I said above, if only 25% = 256/1024, and all the rest of the numbers are truncated as if its x/1000 (i.e 10% =100/1024), then Pchan's numbers work out so that 102% is actually needed with his gear choices, and 101% is not enough.

[pchan]

What seems strangest to me is that it caps at 102% at all.

- clubs x2 (25%) - 64/256
- muse tariqah (10%) - 25/256
- willpower torque (5%) - 12/256
- magnetic (8%) - 20/256
- rdm AF1+1 body (12%) - 30/256
- rdm AF1 body (10%) - 25/256
- 3 belts (10%,6% and 8%) - 25-15-20/256
- karasutengu (15%) - 38/256
-8% from merits - 20/256 calculated individually or not.


102% : dual clubs,torque, earring,NQ body, 6% belt, karasutengu - 258/256 rounding down everywhere.
101% : dual clubs, earring,NQ body,10% belt,karasutengu - 256/256 rounding down everywhere.

Why is that? Seems to imply that you have to have over 100% reduction rate before it works. It seems possible that you guys are somehow perverting an equation or function via an unanticipated negative term. Like if it uses a lookup table to determine the skill/level correction and you're suddenly giving it an input value outside of its range.

I'm just guessing gear x% spell interruption gives x*10/1024. It's very likely seeing how haste and fast cast works. It would mean that if aquaveil is 250/1024 you are 1020/1024 in which case you would need an enormous sample size to see an interruption. For all we know, aquaveil could be a straight 256/1024 which would put you at

-1026/1024 with 77% in gear. (77*10+256)
-1016/1024 with 76% in gear.(76*10+256)

[Yugl]

I'm guessing much like haste, spell interruption rates are rounded values, so the gear won't reflect the true value. Not of much important since you that in total, you need 102% or 77% Spell interruption. Now the question becomes, what can we change about FFXI in light of this new information?

Edit: Have you done tests on higher level mobs as well?

[pchan]

Factor that in with basic observation that you should have had over the years playing FFXI to show that you get interrupted less after skilling up, and a reasonable conclusion can be drawn with very little doubt.

That's not proof and i've never experienced that.

[pchan]

Interesting test + results~ Figured I would play around /w possible values on the equipment based on my experiences with haste and damage reduction:



First assuming the /256 system:
(Estimating values of course.)

Club: (25%) 64/256
Shield: (10%) 25/256
Torque: (7%) 17/256
Earring: (8%) 20/256
Belt: (6%) 15/256
Feet: (15%) 38/256
Merits: (8%) 8/256

Total: (79%) 187/256(~73%)
(BTW, You said 77% but it adds to 79%. Did you write down the numbers wrong on forum or did you do the test with 79% accidently? I've done that before =x.)

torque is 5%

[Gergall]

Now the question becomes, what can we change about FFXI in light of this new information?

RDM/NIN: The new supertank?

[pchan]

Edit: Have you done tests on higher level mobs as well?

I can hold a level 83 MMJ lurker with utsusemi, stoneskin, cures for 30 minutes without interrupts at which point I got bored.

[ringthree]

it's worth noting that in my experience, a hit for any non-zero hp from a higher lvl target will generally interrupt a spell, while a hit for a lot more from a lower level target will usually not. in addition, what's been said about 75 skill vs 37 when casting utsu and interruptions is true from what i understand. a nin/war has a good shot at casting through getting hit while a war/nin has a very poor chance.

i would instead postulate that there's first a check to see if it wants to interrupt your spell based off lvl, skill, and maybe damage, then it tallies your spell interruption down gear and rolls a saving throw against it. if you pass the saving throw, the interruption prevention check succeeds despite the lvl/skill/etc roll failing. IE, in order to interrupt, you must both fail your first roll against your stats and a second roll against interruption % down. this is just a hypothesis, but it would tend to support what we've seen.

an easy way to test this would be to try to cast reraise with say 50% spell interruption gear on both whm and /whm with the main job in the latter case not possessing any healing magic skill. if skill has no impact on interruption, both the whm main and sub interruption against a finite number of tests should be fairly close. if there's a marked difference between the two, then we can conclude that there are likely two checks involved and not just a single one, or at the very least we can affirm that skill in some way does affect interruption rate.

I have noticed (with no evidence to support it) that on my NIN I was usually able to keep casting if I got hit for less than 100 damage but usually got interrupt when I got hit by more. This could have been because lower level mobs do less damage too. Just wanted to voice support for your opinion.

[Kirschy]

These numbers don't make sense unless you are assuming SE error/weirdness in rounding (which isn't unreasonable).


Like I said above, if only 25% = 256/1024, and all the rest of the numbers are truncated as if its x/1000 (i.e 10% =100/1024), then Pchan's numbers work out so that 102% is actually needed with his gear choices, and 101% is not enough.

It's how haste and fastcast works like pchan said. 10% is 102, 5% is 50, and Blackbelt 12% is 122/1024. I'll test Ninja Earring (20%) tonite after class.

I missed a bunch of the other posts, so my post was a little outdated. Also I was wrong on the torque as pchan pointed out -_-

780/1024 means ~244/1024 is needed on aquaveil. If Aquaveil was 250-256/1024 (25%), then it would fit perfectly~ But I imagine in the posts I missed and the posts being posted now, someone has said this!