As long as you keep the samples separate I can extract the mode from both and see if there were likely any non blocks in the face tanking portion of the fight (and thus extract shield block rate, which may be capped for Ochain at 10 lvls under).
And as for sample size, I can get within 3% certainty for damage reduction with 2k samples, but with 257, I should be within 5-6%, depending on the variance, so it should be fine.
As for pure damage reduction, I'll take a look at the data when ya post, it but the difference in means should be enough this time to adequately determine damage reduction.
Ochain
|
Ochain
Would you prefer I post the data on BG somewhere so its not all squished again? Or will here be fine?
Also my Creed hands test only got up to 200~ hits before a group showed up wanting to kill hydra. I just let them. Needed a break anyway. Should I bother posting that or hold off till I increase the sample size? If you send me a screenshot I could make a table for you so it's not all scrunched
I had already made a script to parse the data you posted last time in that format, so it's no big deal either way.
edit; And it doesn't hurt to post creed hands data, can see if there is any information to pull out. OK, I'll go ahead and post stuff. I'll get a SS made too, and maybe edit the table into my post later. Umm, how should I send said SS?
Data incoming an a few Ochain Tests
LVL 90 PLD/DNC DEF 500 VIT 86+47 No PDT gear 369 Shield skill Base Unblocked Sample Base Blocked Sample Creed Hands Blocked That's all for now. I'll increase the other samples later.  Very interesting. Note that the mode and the mean for each were well within the variance/std for that size of a data set, and that there was -no- overlap in damage between blocked and unblocked data. Given this data, it appears that you had near 100% block rate (I say near due to there being ~5% error in this size of a data set) and 63% damage reduction on shield block (again within 5% accuracy). It appears that Ochain does not have an upper bound on block rate, and perhaps stays at capped block rate until you start shifting the curve over by fighting higher level enemies (would need to see some data on some higher lvl things to see the lower bound on it's block rate curve). Pretty interesting though. ~63% damage reduction, then you get to cut the remaining 37% damage off with any PDT you have (if just going for minimal damage), meaning you could get down to ~18% total physical damage taken (with capped PDT) on any given mob. AFK cutting self with my Aegis. I'll take a look at the creed hands data later, as I need to head out today to get some things done before some RL stuff tonight. Thanx for the tests Martel. Nice. I pretty much figured the block rate bit once I looked at the unblock sample dmg. But I much prefer a solid analysis of the data than my eyeballing it.
63% dmg reduction is actually a bit lower than I thought it'd be, but still amazing combined with the block rate. I'm even more curious about the precise creed hands bonus now. Still, based on the old shield dmg- tests Ochain should have at least -70%. The tests I've seen had dmg reduction calcs as "Damage_Absorption = BASE + Shield_Defense / 2 ", with a kite(size 3) shield having a base of -50%. Assuming Ochain has a base of -50%(and it really should be higher, it's a size 6) it'd have -70%. -50% from the base and -20% from its def. Would a larger sample size on the unblock test get us better accuracy? Anyway, I'll increase those samples some time and repost. I think I'll go for Cerberus for my lvl 85 test. and we can see if the block rate drops any. I'll make the table later kinda fighting right now and haven't gone to bed yet lol
No rush.
Well, I was surprised by the actual damage reduction myself. It could be a factor of the sample size accuracy, but honestly it could also just be a different base -dmg% for size six shields (to account for block rate).
SE may have just thought that the linearly increasing base block dmg% wasn't appropriate with Ochain due to the ridiculous block rate. Though I'd say 70% may be pushing it a bit with this data set on the error. I could see 65% or so.. I need to do a similar test with 90 Aegis here sometime when I get the time. Makes sense. Ochain is pretty damn crazy as it is. And SE tries to keep things balanced.
If it did have a -70% dmg reduction... -70% on block, then -50% of whats left with PDT, you'd be taking 15% of the original dmg so effectively 85% dmg reduction?(if I did that right anyway^^;) That be pretty overpowered. But this also makes the shield def bonus(henceforth SDB, god I'm tired of typing that out) on creed hands even more appealing, since its a form of dmg reduction that's separate from PDT. And there's the chance we may get more SDB traits before 99. oh... wait, SDB... How much of that reduction is from the trait, and how much is Ochain? Hmmm, need a pre trait and post trait test with the same shield. Isolate the trait value then Ochain's dmg- should become clear. This, of course, is assuming that SDB functions as a direct increase to a shield's DMG-... Damn, you're right, didn't factor in SDB traits.
That makes Aegis tests hard as well.. Hmm, anyone with low lvl pld wanna do some tests? Not sure how much more useful this will be but here it goes. Believe it was the base blocked set.
Melee damage taken
The rest is already lined up and not sure how a it could be made to look better... maybe you have some ideas. That and I'm not sure what the + and ^ next to the numbers mean.
[+]
Phoenix.Neosutra said: Damn, you're right, didn't factor in SDB traits. That makes Aegis tests hard as well.. Hmm, anyone with low lvl pld wanna do some tests? Also might be best running the test for SDB with a few different shields and mobs. See if it just adds a certain percentage or dmg reduction multiple or maybe a sort of phalanx effect or maybe something silly like just playing around with Cratio/pdif.
Ooh God. I set off another whole round of needed tests.
Well, at least you don't necessarily have to have low lvl pld. As SDB is a lvl 77 trait, syncing to 75 is an option. I'd just be certain to use a lvl ~75 shield for the test as syncing a shield may reduce its def and further eff with the dmg reduction. That variation would also skew our read on the trait value. Although, if the current shield def/dmg reduction formulas are correct you could just check the sync'd def and account for the difference when determining the trait value. It'd be really great if we had a good lvl 75 DMG- test already. then we could just test the same shield at lvl 90 and compare the dmg reduction. Fenrir.Nightfyre said: Phoenix.Neosutra said: Damn, you're right, didn't factor in SDB traits. That makes Aegis tests hard as well.. Hmm, anyone with low lvl pld wanna do some tests? VNM O:< Ragnarok.Martel said: Ooh God. I set off another whole round of needed tests. Well, at least you don't necessarily have to have low lvl pld. As SDB is a lvl 77 trait, syncing to 75 is an option. I'd just be certain to use a lvl ~75 shield for the test as syncing a shield may reduce its def and further eff with the dmg reduction. That variation would also skew our read on the trait value. Although, if the current shield def/dmg reduction formulas are correct you could just check the sync'd def and account for the difference when determining the trait value. It'd be really great if we had a good lvl 75 DMG- test already. then we could just test the same shield at lvl 90 and compare the dmg reduction. My only concern is with the size of those tests. Due to doing a tests at 2 different lvls (75 and 77), you'd have to account for the lvl difference as well in the damage equation. With enough hits for each trial (blocked/unblocked at both lvls), you could pull out a pretty accurate mode/mean, but with the damage changing a good 10% just due to lvl difference, it would be challenging to nail down the additional damage reduction from SDB only. Nah, ya see, the tests are independent.
The 75 test would compare it's ownblock and unblock, then you get a resulting DMG reduction value. Then you do the higher lvl test, and do block vs unblock using only high lvl data and get a -DMG% value for it as well. The tests don't mix, and the only value compared to the other test is the final calculated DMG reduction percentage. Thus taking less dmg overall in the high lvl test shouldn't matter. And they wouldn't even have to be on the same mob or anything. You just need both sets of tests to be large enough to accurately show the difference between trait, and no trait. I just wanted to add that when I got Ochain I ran to powerlvl my shield skill from 282(290) to 361(369). I did it on ~10 VT mandragoras in Abyss - Altepa and my block rate was pretty much the same when I started and when I finished (took me 20 min overall). So even on mobs several lvl higher then you the base block rate from Ochain is very very high. Was funny too :D With phalanx, defender and semi PDT they were hitting for 0 and curing me from set proc from time to time so overall my HP was going up :D
I'll add that even if you had to mix the test from different lvls if you fight against mobs with small lvl range with known attacks you could just add/subtract some def to make your Cratio the same. Or just fight something that is EM to the higher lvl.
Ragnarok.Martel said: Nah, ya see, the tests are independent. The 75 test would compare it's ownblock and unblock, then you get a resulting DMG reduction value. Then you do the higher lvl test, and do block vs unblock using only high lvl data and get a -DMG% value for it as well. The tests don't mix, and the only value compared to the other test is the final calculated DMG reduction percentage. Thus taking less dmg overall in the high lvl test shouldn't matter. And they wouldn't even have to be on the same mob or anything. You just need both sets of tests to be large enough to accurately show the difference between trait, and no trait. Oh I know the tests are independent, I meant more that the decreased damage variance would impact the standard deviation for each test and make a 1 to 1 analysis less accurate (without a significant sample size, since the shield def boost is either a small % pdt in a test that already contains a few % PDT error, or it's a straight -dmg phalanx effect, which also can get lost in 1-3% error). I prefer ~2000 blocked/non blocked hits if someone can manage it. Should be able to find the shield def contribution within that information. It may be awhile before I get around to doing a Shield defense Bonus test. I want to raise the sample sizes on the lvl 80 unblocked and creed hands tests first. Although, someone's actually been killing hydra recently, so I haven't really had a chance.
If I need to get ~2k unblocked hits on the hydra test for good accuracy... It's gonna take a long time. Both my main, and my whm alt completely run outta mp after a few minutes of unblocked testing. I get maybe ~120 hits or so each time, then my whm has to rest, and I have to face the mob and let my mp recover. Now that I think about it, is the 2k hits just for telling the dmg reduction, or is that the amount needed to accurately tell when you're blocking or not in a mixed sample? If it's the latter, at least on hydra with Ochain we shouldn't need that. Now.. if we do need that many hits just to see the full dmg distribution for accurate dmg reduction values... well, then I need to find some help curing. ^^; Perhaps I should have picked an abyssea mob to test this on, but I couldn't think of anything with a static lvl that hits hard enough and wouldn't get me GM'd for holding. If anyone can think of any good test mobs for lvls, 85, 90, 95, etc let me know. For lvl 90 at least, I suppose I could find some EM normal mobs and tank a few at a time for faster sample collection. Just how hard does a mob need to hit for good accuracy on dmg reduction, anyway? I assume higher dmg hits would yield greater accuracy, but how high do we need to go, reasonably speaking? If it's not too high, using an abyssea mob might let me supertank to gather data. It'd be alot faster... Not a big statistics wiz but I'd say a good 100+ dmg normal hits should be sufficient. Which isn't so bad with phalanx and triple regen atmas
Was for accurate damage reduction Martel, not shield block rate (since it appears to be capped at 100% on Hydra).
1000 hits is probably enough though. Well, I think I can live with that. It'll take some work, and I need to catch Hydra up again but it's less painful than 2k. I'll post again once I've got the full unblocked and creed hands sets ready.
OK, improved samples are in. 1K'ish on each.
Unblocked Sample. Blocked, Creed Hands +2 I got bored a decided to try to figure out the -DMG values. We'll see if they match up with Neo's once he gets to them. I think these numbers make things come out to Approx.: Ochain Damage reduction: -63% Ochain + Creed hands: -66.5% Creed hands Value: -3.5% Although I still need to find out how much of that 63% is from Shield Def Bonus. I'm kinda thinking it's that 3%, cause 63% is a weird number.. That or 8%, putting it at 55% before SDB. But in anycase, 60%'ish still seems kinda low. But honestly I can't recall what kinda -DMG values other shields have.
[+]
I may have missed this and I've been out of the game for over year so my memory is a little rough.
Was Aegis 75% dmg blocked @ lvl 75 before shield defense bonus? What % does it block with SDB and without creed? Sounds like Ochain has the traits of a smaller size shield(not 6) just with a massive amount of defense. I'm not sure of the formula's for any of this tbh. Just commenting. |
All FFXIV and FFXI content and images © 2002-2024 SQUARE ENIX CO., LTD.
FINAL FANTASY is a registered trademark of Square Enix Co., Ltd.