Zanshin proc rate and accuracy bonus

[Caiyuo]

This is really interesting, thanks for all the testing, Mote. I remember You and Taro discussing it on FFXIAH awhile ago and hoping for something like this to turn-up just for the sake of finally knowing more about Zanshin.

[Byrthnoth]

Could you post the exact gear and your dagger skill/Dex again? Specifically, did adding the 20 Acc reduce your Dex?

I know I've done tests multiple times and gotten funny results, stared at what I was doing without seeing what was wrong, and then had an LS mate take a 10 second look at the situation and tell me what was up.

[Motenten]

Taru 75 Thf/War
1 dagger merit, 278 total dagger skill
72 Dex

Gear:
Ceremonial Dagger
Tiphia Sting (+2 acc)
Wahlara Turban
Brutal Earring
Raparee Harness
Dusk Gloves
Ulthalam's Ring (+4 acc)
Woodsman Ring (+5 acc)
Steppe Rope (-4 Str to be certain of fStr)
Homam Cosciales (+3 acc)
Homam Gambieras (+6 acc)
All other gear slots empty

Should be:
200 + (78*.9) + (72/2) + (5+6+4+3+2) = 200 + 70 + 36 + 20 = 326 acc

I have already verified that at 306 acc the lv 75 goldfish cons neutral evasion, and at 307 acc it cons low evasion, but going to double-check...

... OK, I'm wrong. I know for sure that I verified that number yesterday. Just rechecked, and +7 acc gave me neutral and +8 acc gave me low evasion. Starting from 270 acc from skill and 36 from dex, that's 314 total acc needed to get it to low evasion. That means I'm at +12 over Low Evasion, which should put me at 86% or 86.5% hit rate.

Both the one I checked yesterday and the one I checked today are Even Matches, I'm sure of that. I'm also sure of the accuracy I needed yesterday, since it was exactly +10 over the value from the crit thread that had level 75's at 297 evasion. So....


Also irritating because now I'm 1 accuracy off from what I wanted, and this test won't tell me anything.

[Byrthnoth]

Well, the Crit Hit thread has:

...

- Om'Hpemde lvl74:
VIT=77
AGI=70
DEF=303
EVA=292

- Om'Hpemde lvl75:
VIT=77
AGI=70
DEF=308
EVA=297

- Om'Hpemde lvl76:
VIT=80
AGI=72
DEF=314
EVA=302
...


I did not verify with checks though so take care :s

If 314=80% Acc, then the monster's evasion is 304, which doesn't match up with any of Masa's Hpmedes. I dunno! Did you have Signet or something during a part of the test?

[Motenten]

Well, I do have signet on me now. Had made sure to not have it on during previous tests, but forgot to take care of that today. On the other hand, accuracy is not one of the listed modifications for signet -- it should only be boosting defense and evasion. Still, will try again later tonight; going to go in without signet, check, then use a signet staff and check again.

Edit: Actually, if signet had boosted my accuracy the problem would have shifted in the opposite direction.

[Byrthnoth]

Yeah, I was hoping you had it on during the first test and not the second.

[Motenten]

Ok, I'm stupid and can't count. Thf has A- dagger, not A+ >.> -7 acc there. I checked using pld with sword for A+ skill. Used to pld/mnk/sam with 276 capped skill.

[Byrthnoth]

Harharharharhar. Does your test mean something now?

[Motenten]

Unfortunately the data I was running was useless. I regeared and now have a true +21 over Low Evasion (328 total acc). I should have results by tomorrow afternoon if the CI is favorable; may have to run overnight tomorrow as well, though.

[Amok]

Ok, I'm stupid and can't count. Thf has A- dagger, not A+

That's not you being stupid, that's SE being stupid!
[/rant]

If I may ask a stupid question: how does the question of floored accuracy relate to the Zanshin testing? I'm missing how things moved in this direction. Btw, please keep it up, I enjoy this thread and your testing a lot.

[Motenten]

It determines the apparent accuracy bonus Zanshin gets. If accuracy is floored then I can't work with 0.5% increments for the base value or the Zanshin hit rate, which influences the apparent range of the confidence interval, and thereby the final estimated bonus gained on the Zanshin hit.

[Masamune]

please help me confirm those Om'Hpemde evasion values Mot...

Atm i'm getting stuck after checked stats for GreatBirds and Mandies families.... so getting serious doubts on the method used to interprete "Low Eva/def" and "High Eva/def"...

[Nameless]

Yea, im totally agreed with Tarowyn in terms of Zanshin proc'ing method: whenever your rounds gets 2 or more attacks already -> no Zansh, no matter which type of weapon or underwear you have.
Which means Zansh check is most likely done after DA and TA correct?

Big question now is: Can Zansh proc on a missed DA hit? (or 2x miss TA round?)

From my personal experience, zanshin doesn't proc when you DA or TA. I have never in my playing career seen a triple attack from missing either 1 or 2 hits of a double attack. I have also never in my playing career of using an Algol (there was a period that I fulltime it for a year or so in major raids with a 6 hit build in case you are wondering, and I now use Naglering for those situations) seem a 4 hit round where I miss 1 to 3 attack on a triple attack proc from the sword.

I have actually talked about it in the LS before. However, like we all were before this test, thought that the proc rate is only 10%, so the chance of triple attack (3%), and have it miss, and have a 10% chance for zanshin, might be too small to bother to begin with. Now that we know it is 25% as sub job then it definitely confirms that it can't proc during any DA or TA since the chance of me seeing a 4 hit attack during that year would have been much greater.

[Masamune]

Ok thank you for answering Nameless, even if it's eyeballing :/

[Byrthnoth]

I also don't remember ever seeing a triple attack (miss/hit/hit or hit/miss/hit) as WAR/SAM.

However, keep in mind that it looks like this trait has changed fairly recently based on comparisons to Lyra's test. Lyra used a small sample size (~100), but if you calculate the 95% confidence interval off that you still exclude 25%.
Heck, as far as we know Zanshin either is "broken" right now or was then. xD

[Amador]

Hello, I don't know if this will contribute to this thread much...

However, after obtaining my Gungnir my accuracy hit rate on parser's has reached 96.X%with varrying .X%'s.

Unfortunately I don't have a parser save for this x.x; However I don't consider %'s very accurate unless they are a large sample size over 1500... it was an insanly long merit party wiith over 1500~ melee hits. Don't think I'll be able to get my parser working for a bit, untill I can find out what the new memory offset number combo is.

However, someone in party mentioned that the 96.X% was due to Zanshin, so,yep.

[Kirschy]

Hello, I don't know if this will contribute to this thread much...

However, after obtaining my Gungnir my accuracy hit rate on parser's has reached 96.X%with varrying .X%'s.

Unfortunately I don't have a parser save for this x.x; However I don't consider %'s very accurate unless they are a large sample size over 1500... it was an insanly long merit party wiith over 1500~ melee hits. Don't think I'll be able to get my parser working for a bit, untill I can find out what the new memory offset number combo is.

However, someone in party mentioned that the 96.X% was due to Zanshin, so,yep.

Parsers only record melee hits landed/total melee hits. 95% is your cap, the extra % you're seeing is due to the error margin. The "Higher than 95% acc due to Zanshin" is due to Zanshin proc'ing on those 5% misses, giving you an extra attack you wouldn't have had before. This does not show up in a parser since it only checks total hit rate.

[Motenten]

@Masamune: The evasion for the level 75 hpemde is indeed 297 (assuming Low Evasion is equal to, not greater than, +10 over mob evasion). Verified on both pld and, after fixing my numbers, thf. The other levels you listed should be correct as well, then.

@Amador: The margin of error on a 1500 hit sample size of a 'true' 95% hit rate is 1.1%, so it is indeed perfectly possible to see 96%+ hit rates in the parser. And that's only the 95% confidence interval. One in twenty 1500-hit parses have a chance of falling outside that. The 99% confidence interval is 1.45%, which means up to 96.45% reported hit rate is theoretically possible without disproving the base hit rate cap.

Unfortunately, to definitively prove that, someone would need to do a 100k-200k hit parse as sam main (double it for sam sub). I don't know if anyone is up for dedicating that amount of time to it. Would basicly have to afk in sea for a week solid.

[Masamune]

Phew, thank you Motenten.

And yes, the low evasion check is the only one i don't use +1 or -1 to calculate real defense/evasion value:
RealEVA=Skill+DEX/2+LowEVAcheckMax +10 (don't use -1)

The other checks i used are :
- RealEVA=Skill+DEX/2+high eva check Min+1+30
- RealDEF=floor( (LowDefCheckMax-1) / 1.25)
- RealDEF=HighDefCheckMin+1

[Tarowyn]

I also don't remember ever seeing a triple attack (miss/hit/hit or hit/miss/hit) as WAR/SAM.

However, keep in mind that it looks like this trait has changed fairly recently based on comparisons to Lyra's test. Lyra used a small sample size (~100), but if you calculate the 95% confidence interval off that you still exclude 25%.
Heck, as far as we know Zanshin either is "broken" right now or was then. xD

I dunno about that, the original Japanese testing for zanshin dates back to 2006 and SE has at least not publicly changed it since then. Personally I haven't seen any period where it felt like the proc rates dropped to the ground.