Enhances the effect of...

[epigram]

I reported the initial findings on Pln. Body and reported it as around 2.5% most likely 2-3%. A more conclusive test followed after the original reporting, this test had a greater sample of ~2000 hits.

The tests were done over several goobbue of the exact same level with assault jerkin on, and pln. body on.

The assumptions were:

- A critical attack is a Bernoulli trial.
- p-hat can be estimated by (# of crits) / (total number of attacks).
- We can restate the question as, what is the expected number of critical attacks per 100 attacks.

Proceedure:

-Determine p-hat with Pln. body -> p-hat1
-Determine p-hat without Pln. body (assault jerkin) -> p-hat2
-Determine var(p-hat1) and var(p-hat2).
-Use CLT
-Construct 95% CI and test.

Final results over 2000 hits showed:
1. Pln. body did increase critical hit rate
2. Pln. body had an increase in crit rate around 2.5%

[Tarumara]

if noone has posted, gloom breastplate adds 2% to souleater.

[Asherek]

does cover last for a set amount of time, or for an amount of hits covered ??

It's a set amount of time. Without any merits, I'm pretty sure it's 30 seconds.

[Apollyon]

Roshijinpachi - Enhances Effect of Rice Balls

This is the same effect granted by the SAM AF items.

The effects from this item, and from the SAM AF STACK, so you will get double the bonus. This information is already on FFXIclopedia.

[BerenTebogo]

Roshijinpachi - Enhances Effect of Rice Balls

This is the same effect granted by the SAM AF items.

The effects from this item, and from the SAM AF STACK, so you will get double the bonus. This information is already on FFXIclopedia.

With possibility of SAM tanks in update it brings up the question:

"Is SE bringing back riceball fun?"

Roshi + SAM AF hands = 2x Effect rice ball, ~DEF +100ish / Double ATT+2% (long time since I used rice balls, confirm?) and pretty good ATT+

Just of the top of my head, but before sushi was made rice balls were a pretty good choice, who knows, maybe theyll make a comeback. lol

[Ferien]

Cover: 15 seconds unmerited with no Cover+

Will edit this post once I pull my Coronet out >.>

[Asherek]

Cover: 15 seconds unmerited with no Cover+

Will edit this post once I pull my Coronet out >.>

Are you sure it's 15 seconds? I tested it the other day without wearing a single piece of Cover+ (and no merits into it) and I swear I timed it as 30 seconds. >.>

Hell I'll Fraps it when I get home tonight.

[divisortheory]

Have a BLU equip and try Jettatura (120 second recast) would be the best way.

Jettatura/Temporal Shift
2:00
1:54 H.Legs (3% + FC)
1:56 H.Hands (3%)
1:53 H.Legs + N.Feet (4% + FC)
1:55 H.Hands + N.Feet (4%)
1:50 H.Hands + H.Legs (6% + FC)
1:49 H.Hands + H.Legs + N.Feet (7% + FC)
1:21 H.Hands + H.Legs + H.Feet + N.Body + W.Turban + Swift Belt + Refueling (31% + FC)
1:20 H.Hands + H.Legs + H.Feet + N.Body + W.Turban + Swift Belt + Refueling + Locquacious Earring (31% + FC + FC)

This seems like enough information to figure it out conclusively, I only have one question which I don't have the time to figure out on my own right now. Say you've got an ability with Recast Timer X. You have Haste+N%. To determine the new recast timer, which of the following formulas are used?

a) X - Floor[N*X]
b) X - Ceil[N*X]
c) Floor[X - N*X]
d) Ceil[X - N*X]

[shotgn]

I reported the initial findings on Pln. Body and reported it as around 2.5% most likely 2-3%. A more conclusive test followed after the original reporting, this test had a greater sample of ~2000 hits.

Would you link or display the data for this test?

Proceedure:

-Determine p-hat with Pln. body -> p-hat1
-Determine p-hat without Pln. body (assault jerkin) -> p-hat2
-Determine var(p-hat1) and var(p-hat2).
-Use CLT
-Construct 95% CI and test.

What do these mean?

[epigram]

I might have the data still, most likely I deleted it when I was done. If I have free time I could run it again.

p-hat is an approximation for the probability of an event occuring, here it is the probability of a critical attack. To make any valid assumption based upon an increase in critical attack you really need the variance of p-hat, or var(p-hat).

CLT is Central Limit Theorem, it's the easy way to measure the difference between two point estimates, assuming that they meet a number of conditions. It allows you to use normality assumptions essentially.

If none of this makes sense to you, then it is unlikely that you can make any statistically valid statement.

[divisortheory]

Also, 95% CI = 95% confidence interval. It's basically a range of numbers that statistically there is a 95% level of confidence the true number falls somewhere within that range. As your sample size goes up, your CI shrinks and you can close in on the real number. You can make a 99% CI too, but it would take between 10 and 20 thousand samples to get a decent interval.

This is the same way I performed my test on the Thief's Knife and concluded that the increase was around Drop Rate+1%.

Edit:
(TotalCrits*[TotalHits^2 - 3*TotalHits + 3*TotalCrits] / TotalHits^3, TotalCrits*[TotalHits^2 + 3*TotalHits - 3*TotalCrits] / TotalHits^3)

gives you a 99% confidence interval unles i messed up the simplification. Change the 3* to a 5* in the 4 places to get a 95% confidence interval.

[shotgn]

Confidence interval..that's what i was interested in. According to your formula then, using Genome's test with Pln. Body (TotalCrits=117,TotalHits=718) that gives a 16.24%..16.35% confidence interval. I must say that's a lot better than I thought, if the formula is correct. That would basically be saying that most likely we can trust his numbers to within a few tenths of a percent. I.e. Pln. really DOES lower your crit hit percentage.

I must be missing something here because basically this is saying that Genome calculated the effect of Pln. to be -2% and epigram did 2.5%, both with 99% confidence.

[divisortheory]

I fucked up the formula, my bad. Left out a square root trying to simplify it, and reversed the 5s and 3s. Maybe this is easier, I just wanted a one step formula:

CritPct = TotalCrits / TotalHits
Deviation = Sqrt[CritPct*(1-CritPct)/TotalHits]

99% Confidence Interval = [CritPct - 5*Deviation, CritPct + 5*Deviation]
95% Confidence Interval = [CritPct - 3*Deviation, CritPct + 3*Deviation]


Edit:
Genome's Results

With Pahluwan
CritPct = .163
TotalHits = 718

Deviation = Sqrt(.163*.837 / 718) = .0137

99% Confidence Interval = [9.41%, 23.19%]
95% Confidence Interval = [12.17%, 20.44%]

Without Pahluwan
CritPct = .181
TotalHits = 635

Deviation = Sqrt(.181*.819 / 635) = .0153

99% Confidence Interval = [10.46%, 25.74%]
95% Confidence Interval = [13.52%, 22.68%]

[Ferien]

Cover: 15 seconds unmerited with no Cover+

Will edit this post once I pull my Coronet out >.>

Are you sure it's 15 seconds? I tested it the other day without wearing a single piece of Cover+ (and no merits into it) and I swear I timed it as 30 seconds. >.>

Hell I'll Fraps it when I get home tonight.

I promise. Wore off with 2:45 on the recast.

[Khamsin]

If it hasn't already been mentioned, be aware that Warrior's Calligae and Lorica (Enhances Berserk/Aggressor) only maintain their effect while they are equipped. They cannot be macroed.

If you equip Warrior's Calligae and use Berserk, you will receive the lesser Defense penalty. However, unequip the Calligae, and your Defense will go back down to the full penalty. Re-equip the Calligae, and your Defense will go back up again to reflect the reduced penalty.

They must be left on to retain the penalty reduction.

On the other hand, Warrior's Mask can be macroed. If you equip the Mask, and use Warcry, your Warcry will last the entire extended duration even if you take the Mask off immediately after using Warcry.

[epigram]

115/635 = 18.11 (af)
117/718 = 16.3 (pln)

Quicky 95% CI
1.96 * SE(af) = 3.0%
1.96 * SE(pln) = 2.7%

95% CI for AF (15.11% - 21.11%)
95% CI for pln (15.6% - 20%)

You really can't show anything either way.

[shotgn]

And for 2000 trials the confidence interval is much tighter?

[Ferien]

And for 2000 trials the confidence interval is much tighter?

Indirectly, yes. A higher #trials gives you a smaller Standard Deviation, which gives you a tighter confidence interval.

[epigram]

The 95% CI didn't overlap for my test, it was around 9 percent for assault jerkin and around 12 percent for CI. The real increase was most likely within a 1-5% range, I just estimated it at 2.5%

[Scythiroth]

This topic is making me think too much, I'm going to stare at NWS topics some more. =/