Soboro Grips

[Hakamaru]

What kind of operation he did that he needed Fourier transform? noise cleaning?

Heh, I don't completely understand it myself (it's been years since I've worked with frequency-domain). Seems like he just applied a frequency-domain transform on the data, looked at the plot and asserted that it looks like the dirac delta function because of the pronounced spike. Then he says something about how the dirac delta function is random by definition (not sure where that comes from or how it's proven), which implies that the data-set is random as well.

Looks like a bunch of smoke and mirrors for not saying much at all.

[Zarath]

What kind of operation he did that he needed Fourier transform? noise cleaning?

Heh, I don't completely understand it myself (it's been years since I've worked with frequency-domain). Seems like he just applied a frequency-domain transform on the data, looked at the plot and asserted that it looks like the dirac delta function because of the pronounced spike. Then he says something about how the dirac delta function is random by definition (not sure where that comes from or how it's proven), which implies that the data-set is random as well.

Looks like a bunch of smoke and mirrors for not saying much at all.

I think adding the Fourier transform is kinda over thinking it...

[ringthree]

Testing done with Ridill shows two things:

1. Number of hits per round is not evenly distributed. Ridill procs as a triple significantly less often than double or single.

2. Trait-based Double or Triple Attack proc first, and eliminate any possibility for the weapon to proc. This was tested with Warrior's Charge, and I believe with Assassin's Charge as well, and in both tests, Ridill NEVER proc'd anything differently than what the ability told it to.

Assuming these findings can be carried over to Soboro, which they most likely can, you see that not only does DA reduce your chance to get a triple proc, it also effectively lowers your average hits/round.

I actually posted an explanation of why people could get Ridill proc rate problems, while testing. Its on my blog too, about a month ago.

If I'm reading it correctly, you're saying that it was because rounds which kill the mob would skew it down. However, all of the tests I've read throw out all killing rounds because of that possibility, so that's not a valid argument.

I haven't seen confirmation of that being the case, just that it was suggested that it should be the case. The one test I have seen didn't make mention of it, and that test had some pretty random numbers for 1, 2 and 3 attack rounds. I would have to find it again to confirm, if you can supply the test information I would reconsider if I am wrong.

Edit: Didn't see the post above with the information, but I still doubt the frequency, so I may just test my Soboro.

Edit 2: This spread: 27% single attacks, 53% double attacks, 20% triple attacks. Could easily be 25%/50%/25%. I don't think that test "massively" favors Single over Triple attacks, just Double over both of them.

[Hakamaru]

Heh, I don't completely understand it myself (it's been years since I've worked with frequency-domain). Seems like he just applied a frequency-domain transform on the data, looked at the plot and asserted that it looks like the dirac delta function because of the pronounced spike. Then he says something about how the dirac delta function is random by definition (not sure where that comes from or how it's proven), which implies that the data-set is random as well.

Looks like a bunch of smoke and mirrors for not saying much at all.

I think adding the Fourier transform is kinda over thinking it...

I'm confused. Did you mean to make a post saying exactly what I stated in the last sentence of my post? I wasn't the one that posted the graph in question here. Just trying to understand what someone else put up on the Wiki.

[Ferien]

Testing done with Ridill shows two things:

1. Number of hits per round is not evenly distributed. Ridill procs as a triple significantly less often than double or single.

You got that test? I'm hoping it was done with a job that doesn't have double attack since if you tested it on WAR you would of course have more double attacks than single or triple (the trait processing before the multi-hit).

They were on BG a while back, and I know I've seen them on the wikia. Give me a bit and I'll dig them up. I believe at least one of the tests was done by a WAR, and at least one of the others was done by a friend of mine with RNG. Pretty sure I remember a DRK doing one, too, and they all showed that even the weapon itself favors singles and doubles over triples. The WAR's test showed that the DA trait skewed it further away from triple than it did single.

Edit: You already found the WAR's tests, this was the RNG's: viewtopic.php?p=696994#p696994

Came out as 30.6% single, 50.7% double, 18.6% triple. She was /NIN, no Brutal, and discounted all final rounds. Pretty good evidence IMO.

[VZX]

What kind of operation he did that he needed Fourier transform? noise cleaning?

Heh, I don't completely understand it myself (it's been years since I've worked with frequency-domain). Seems like he just applied a frequency-domain transform on the data, looked at the plot and asserted that it looks like the dirac delta function because of the pronounced spike. Then he says something about how the dirac delta function is random by definition (not sure where that comes from or how it's proven), which implies that the data-set is random as well.

Looks like a bunch of smoke and mirrors for not saying much at all.

Exactly what I thought

As much as I learn Fourrier Transform... it does nothing other than represents the values in frequency domain.. which makes.. some operations like noise filtering/convolution is easier :/ but looks like he didn't do anything with it.

[Aymee]

...This has turned into a discussion about distribution of hits on multi-hit weapons and less about soboro grips, lol.

A few things I've noted...

First off, Searain doesn't mention if he has +DA merits. Assume he's lvl5. That would make his innate DA 15%, not 10% from trait. Though this doesn't matter in the end (see below).

Searain's data for the second set confused a little at first because it does have more total attacks than the first set. Then I actually divided it out and saw how there's essentially no difference:
1967 / 1020 = 1.93 swings per round.
1941 / 1021 = 1.90 swings per round.
I'm not going to run the math because my brain hurts, but the standard deviation there would be tiny. For all intents and purposes, these are equal results.
Conclusion thusfar: +DA from whatever on WAR has zero impact on Ridill's total attacks and won't change TP gain. 1000 data points seems plenty... 2000 if you include both experiments (which we can do, if there was no change).

By association of the +DA% gear, +DA merits would also be meaningless [for Ridill]. Also by association, if having DA 10% or 20% results in the same average number of swings, having DA 0% would as well. So I don't see any need for the experiment to be run by another job.
Conclusion thusfar: If we apply the logic derived from all the data, the swings/round with a Ridill would be the same for any job that can use.
HOWEVER--it's worth noting that if you use his raw %es (the ones he crossed out?), you get 1.8 swings, and 1.74 swings/round, respectively. I have no idea if he crossed them out to correct for WAR natural trait, or because of one of the people that corrected him later. Evenso, the match doesn't match... 1.8swing/round with no DA is only ~5% less swings than 1.9... not the 10% innate job trait on WAR, so I don't believe the crossed-out numbers.

People have noticed the rough spread of 30/50/20 with 10% DA and more bending towards 25/60/15 as you add to 20% DA. Theoretical result for +30% DA... 20/70/10 maybe? ...That would still be 1.9/round. I have a feeling it has to start tapering off at some point, since I'd hope the TA rate can't be 0 unless you had 100% double attack--MAYBE if you put on +DA everything imaginable and had a COR that you only tested with having the same roll number on you, there could be a point where the swings/round could change: Example, say at 30% or higher DA the linear progression doesn't work, and you end up with 15/77.5/7.5. That would mean 1.925/round... tiny change but at least it's slightly different instead. This would be really really hard to test since you'd need the same COR roll to be really picky about your data.
Thusfar: After applying the general consensus on percentage distribution, the amount of double attack one has still makes no difference in swings/round. This upholds the earlier conclusions.

Since +DA lowers singles for doubles and lowers triples for doubles... If it's a linear (or minor exponential decay) progression then 0% DA would give something like 35/40/25. If the distribution was that, we would get the exact same 1.9 swings per round as his 2 experiments.
Conclusion: Applying conclusion 2 with 3, we are reinforced that there is no real need to test the ridill with a job with no innate multi-hit traits.

After all of this I think it's a pretty fair conclusion that the Ridill average hits per round is 1.9 operating on some kind of bell curve with a push up in the center as multi-hit abilities are added. Key being, it's the same regardless of job/job traits/gear. Perhaps time to rethink some TPing gear for Ridill people? I'm not cool enough to have one, so I really wouldn't know.

Now someone just needs to do it with soboro! ...Knowing SE though, the distribution is probably roughly the same, if not exactly the same. Major problem though--Zanshin. A first-hit whiff might just be 1 hit with a Zanshin or 2 hit with one miss. Ug. It'd be next to impossible to figure out the difference between zanshin hit and soboro multi-hit, unless someone wanted to take as much data as Searian AND had the group2 merit that gives extra TP for zanshin hits. That's the only way I can think of to distinguish.

Conclusion for the OP (or uh, Soboro grips? whatever the subject is): Pole strap for +2% is likely doing absolutely nothing for your total attacks or TP build. However, it'd still proc on WS... meh. Conclusion for me? Definitely time to buy a new grip if you use soboro... probably Claymore as I said in my first post days ago, lol.

[Phraust]

Did this tonight:

Test configuration:

Lebros Cavern Assault: "Excavation Duty"
Jobs:

Drk/Drg.
Thf/Drg.
War/Drg.
War/Drg + Askar Korazin & Brutal Earring.

Merits: Triple Attack: 5/5, Sword 8/8.
(My war is pretty much unmerited... No DA merits, etc.)

Method: Enter assault and hit rocks completely naked, save for the weapon, and Korazin/Earring in the final test.

Results:


Drk/Drg:

Triples: 17.21%
Doubles: 50.93%
Sinlges: 22.09%
Total: 430 Rounds. 795 Swings. (1.85 Swings/Round)


Thf/Drg:

Triples: 21.86%
Doubles: 47.21%
Sinlges: 39.93%
Total: 430 Rounds. 821 Swings. (1.91 Swings/Round)


War/Drg:

Triples: 15.81%
Doubles: 55.35%
Sinlges: 28.84%
Total: 430 Rounds. 804 Swings. (1.87 Swings/Round)


War/Drg + Askar Korazin & Brutal Earring:

Triples: 18.37%
Doubles: 59.77%
Sinlges: 21.86%
Total: 430 Rounds. 845 Swings. (1.97 Swings/Round)


So if you take Drk as your "Baseline Reading", when you switch to thief, you gain roughly 5% more Triples, the sacrifice of lost Singles/Doubles is directly proportionate to the pre-existing ratio, the same seems to be true with the warrior gains:

Drk: 17:51:32 [T,D,S]
Thf: 22:47:31 (+4.65%, -3.72%, -1.16%)
War: 16:55:29 (-1.40%, +4.42%, -3.26%)

WarD: 18:60:22 (+1.16%, 8.84%, -10.23%)

Still trying to figure out why the geared war had such odd results...


Edit: Fixed a finger slip, Thief had 29% singles, not 39%...

[Phraust]

Oi...

Ok.. the point of all that: basically what the dude above me said, Having a +DA grip seems to be basically a 0 gain. It neutralizes itself. So imo on the Soborro, Use the Crit+3% grip.

[Tjonke]

Question: Isn't it possible that the person reporting seeing 4 hits with Soboro missed one of his swings and that Zanshin (sp) kicked in to counter his miss. Or does Z[tab a few times] fail to effect if a multi-hit weapon swings more times in that attack round and the later attacks are actual hits?

Just a question.

[Amele]

never seen a zanshin proc on a triple with soboro and I've been using it exclusively since level 50 - although I don't miss much in exp and zanshin is based on missing (you get the most out of zanshin around 50% total accuracy, yuck!)

short version is: I doubt that zanshin can trigger on non-virtue multihit (or I would probably have noticed one by now). but unlike double/triple, will not overcome multihit weapon attributes (since the number of swing calculation comes before the number of hits calculation etc.)

I did an independent test with soboro the last time this came up (I didn't keep the original data) but also arrived at about a 1.85-1.9 swing/round (I can't control for zanshin on singles beyond trying not to miss - but I did go sam/nin so I wouldn't have double) I'm not 75 so I couldn't distinguish a zanshin vs. a regular hit, although you still wouldn't be able to control for double miss zanshins.

==
I think some people are over thinking the DA vs. ridill stuff : at best it has marginal effect; at worst it kills your multiattack rate. so the gear is blank either way.

brutal? blank (except for store tp+1) wear it in ws unless you really need that 1 tp.
pole strap? blank (period) don't wear it with soboro.

[Edrick]

I know it'd be more of a pain, but couldn't you eliminate any attack rounds where you miss any of the attacks when testing out soboro multi-hit stuff? That'll eliminate any possibility that Zanshin proc'd at all. This would be like eliminating the final attack round. Though, I guess you could always include any attack round that missed on the -final- attack... May be a bit complex, but I'm not the one that's going to test something like that. >.>

[Retrieving/C.D/Searain]

...This has turned into a discussion about distribution of hits on multi-hit weapons and less about soboro grips, lol.

A few things I've noted...

First off, Searain doesn't mention if he has +DA merits. Assume he's lvl5. That would make his innate DA 15%, not 10% from trait. Though this doesn't matter in the end (see below).

I have maxed Aggressor and Berserk, I would've mentioned the additional DA% from merits otherwise.

it's worth noting that if you use his raw %es (the ones he crossed out?), you get 1.8 swings, and 1.74 swings/round, respectively. I have no idea if he crossed them out to correct for WAR natural trait, or because of one of the people that corrected him later. Evenso, the match doesn't match... 1.8swing/round with no DA is only ~5% less swings than 1.9... not the 10% innate job trait on WAR, so I don't believe the crossed-out numbers.

The crossed percentages are wrong (hence the crossing lol, I must've mixed the data up when I copy and paste'd em the first time in the software I used to count the sheer #'s of 1's, 2's and 3's, I haven't done that manually) while the %'s right next to them are the correct ones. The strings of data however have always been correct regardless and they match the info found in my original files and they've never changed either (check the article history if you're doubtful).

[Amele]

I know it'd be more of a pain, but couldn't you eliminate any attack rounds where you miss any of the attacks when testing out soboro multi-hit stuff? That'll eliminate any possibility that Zanshin proc'd at all. This would be like eliminating the final attack round. Though, I guess you could always include any attack round that missed on the -final- attack... May be a bit complex, but I'm not the one that's going to test something like that. >.>

you would get biased results doing this. (exercise for the reader: how would the results be biased!)

[Aymee]

never seen a zanshin proc on a triple with soboro and I've been using it exclusively since level 50 - although I don't miss much in exp and zanshin is based on missing (you get the most out of zanshin around 50% total accuracy, yuck!)

I did an independent test with soboro the last time this came up (I didn't keep the original data) but also arrived at about a 1.85-1.9 swing/round (I can't control for zanshin on singles beyond trying not to miss - but I did go sam/nin so I wouldn't have double) I'm not 75 so I couldn't distinguish a zanshin vs. a regular hit, although you still wouldn't be able to control for double miss zanshins.

Zanshin proc obviously can't be on the third hit, because yea we'd see some kind of habit of 4x attacks, which isn't possible... a fair guess might be that zanshin can only kick in on the first swipe. Your testing resulting in 1.85-1.9/round doesn't surprise me a bit, given the result of Searain's research all indicates that double attack is entirely meaningless on a 3x hit weapon regardless of where it comes from.

I know it'd be more of a pain, but couldn't you eliminate any attack rounds where you miss any of the attacks when testing out soboro multi-hit stuff? That'll eliminate any possibility that Zanshin proc'd at all. This would be like eliminating the final attack round. Though, I guess you could always include any attack round that missed on the -final- attack... May be a bit complex, but I'm not the one that's going to test something like that. >.>

you would get biased results doing this. (exercise for the reader: how would the results be biased!)

If you only included rounds where you had 100% accuracy, you'd get a bias toward 1 and 2-hit, since you're more likely to miss a swing if you go 3 times (regardless of your accuracy). It might be the first or second missed, but yea there's a total overall higher % chance of -something- missed with 3 swings. Furthermore, if you did specifically include rounds that missed only the 3rd attack, you'd get higher results on 3x rounds. Times where specifically the 3rd hit miss and not the first or second aren't magically special for some reason =\

Though it's probably worth noting that if you excluded then included those two different things, they might balance each other out singe both scenarios would be a very low % of rounds, lol. Totally not scientific though. I'm sure you'd get the ~1.9 even with these little issues, though yes it would be tainted.

[Amele]

yeah, I was corroborating sea's evidence. I have no reason to doubt that the code snippet that does ridill is copy/pasted onto soboro (or is a part of the native attack code and just activated by a flag on both).

zanshin procs can't be happening period when combined with a triple. (zanshin always shows as an extra swing.) since there's nothing special about 3 attacks in terms of attacks/round (contrast with 8) there's no evidence to support zanshin proc'ing -at all- on a double or triple from soboro. (since there's no reason to believe that a double is different from a triple)


(I expect it still happens on single attacks with soboro; but I don't really have a way to verify this myself - I'm not 75 nor am I planning to merit +TP on zanshin.)

they won't balance out. the likelihood of missing the first or second hit is still higher than missing the first of two hits, regardless of how you treat the third. - likewise, since you're not throwing out -any- single attacks you'll still be biased on these (unless you throw out all misses, which results in a bias in the -other- direction, especially with zanshin around.)

the best way to do this is to first confirm the zanshin rate, (minimum acc+sam/thf+non-soboro: take a blind potion to beseiged maybe)

then verify if zanshin can proc with soboro (via the merit)

then adjust your data by removing the 'expected' rate of zanshin procs from your data. (and or watch like a motherfucking hawk to check your TP after every double).

regardless. most people with soboro don't set up a zanshin set if they use it at 75; and sam can't make the base trait go away; so zanshin doesn't *really* matter.


one thing I am curious about is how zanshin combines with double attack; I've had soboro the entire time I've had double on sam, so I can't really say with any accuracy.

I suspect they stack and that a 4 swing (with 2 misses) is entirely possible (although vanishingly rare) but I haven't had the motivation to get out a different GKT and some blind potions to go test.

[Zarath]

Heh, I don't completely understand it myself (it's been years since I've worked with frequency-domain). Seems like he just applied a frequency-domain transform on the data, looked at the plot and asserted that it looks like the dirac delta function because of the pronounced spike. Then he says something about how the dirac delta function is random by definition (not sure where that comes from or how it's proven), which implies that the data-set is random as well.

Looks like a bunch of smoke and mirrors for not saying much at all.

I think adding the Fourier transform is kinda over thinking it...

I'm confused. Did you mean to make a post saying exactly what I stated in the last sentence of my post? I wasn't the one that posted the graph in question here. Just trying to understand what someone else put up on the Wiki.

I guess I did agree on you with that ^^;.

Anyways I believe zanshin wouldn't proc to make a 4th attack hit. I think Zashin works if you completely miss in a single attack round, rather than a miss with an attack in a multi hit attack round.

[Amele]

right. zanshin doesn't work with soboro (or else I would've noticed)

does zanshin work with double attack?