Magic Accuracy - Why so serious?

[Starr]

While on the subject of JP tests, I'd like to add: http://wiki.ffo.jp/html/805.html. Which suggests that the hidden +M.Acc on HQ Staves would be roughly +25-30.

If the test I posted, which is from a BG thread that is linked on the first page, is accurate and 1 Magic Accuracy gives 0.71% to landing rate, 25-30 would give 17.75% to 21.3% landing rate, which is pretty close to the 15% people get with HQ staves.

[Tricen]

No staff, base skill=322 enfeebling, minor stat bonus (+0 INT +7MND) => 47.9% (500+ sleeps I)
HQ staff, base skill=322 enfeebling,minor stats bonus => 63% (300 sleeps I)

Not to sound pedantic, but that is a large gap :<

[Spekkio]

No staff, base skill=322 enfeebling, minor stat bonus (+0 INT +7MND) => 47.9% (500+ sleeps I)
HQ staff, base skill=322 enfeebling,minor stats bonus => 63% (300 sleeps I)

Not to sound pedantic, but that is a large gap :<

well he admitted that his sample was insufficiently large to draw any absolute conclusions from. it does give us some direction on it though. these are the tests i remembered that i referenced before. for some reason i thought they were done by pchan. my apologies to starr for the incorrect attribution.

[Tricen]

Well, in the next 200 casts, you don't know if 100 of them would be resisted, and it would drop the %'s down. Sample sizes should be close to exact to draw conclusive evidence from, in this game. imo at least

[Spekkio]

Well, in the next 200 casts, you don't know if 100 of them would be resisted, and it would drop the %'s down. Sample sizes should be close to exact to draw conclusive evidence from, in this game. imo at least

sample size only affects margin of error, it does not result in an apples to oranges comparison. flip a coin 500 times and about 250 will be heads and 250 tails. flip another coin 30 times and about 15 will be heads and 15 tails. the % of heads on the smaller test will not be quite as close to 50% as with 500 flips (hence the margin of error) but both are centered around 50% odds. i know suriam will be along shortly to give a much better example than mine. ^^

[Starr]

300 is plenty enough for an accurate result

[Tricen]

Then compare it to something with a sample size of 300? You could flip a coin 10 times and get 10 heads, then compare it to a sample size of 100 where you got 50/50.

[Suiram]

I don't really know much statistics ;x My main issue with that test is that by using sleep, the data you get is not checking the entire spectrum of resist outcomes, but rather unresisted + partial resists vs. full resists. I guess from an enfeebling standpoint that might be more important to know, but it makes it more difficult to draw broad conclusions.

[Spekkio]

I don't really know much statistics ;x My main issue with that test is that by using sleep, the data you get is not checking the entire spectrum of resist outcomes, but rather unresisted + partial resists vs. full resists. I guess from an enfeebling standpoint that might be more important to know, but it makes it more difficult to draw broad conclusions.

while it's less informative, it's still valuable. if you roll a D6 knowing how many 1's 2's 3's etc come up is more specific and valuable, but simply knowing how many rolls come up as 1 or something else (2-6) assuming sufficient trials to produce a meaningful result it is possible to discern if the die is weighted and roughly to what degree. while it leaves us wanting a lot more data regarding the bias of the die, it does give us a metric to measure if the die is biased and by how much.

[Haroun]

If there's a BRD around with access, it might be interesting to test the resist rate of Alkalurops + Shadow Coat vs. Apollo's Staff + Osode using Lullaby with Nursemaid's Harp on Greater Colibri (while under 120 total CHR). From that it shouldn't be difficult to equate HQ Staff success rate to an exact value of +M.Acc (it may be a direct % increase in success rate, but converting this to +M.Acc would be more useful for gear comparisons).

EDIT: The control would be using Osode but neither staff.

[Starr]

Then compare it to something with a sample size of 300? You could flip a coin 10 times and get 10 heads, then compare it to a sample size of 100 where you got 50/50.

wut

[Sekundes]

In regards to tiered resists... Is there any testing how the game mechanics decides a full resist compared to a half, 1/4th, 1/8th? My understanding of magic in this game is more limited because i'm a blu but my guess would be that its just a linear model shifted based on stat comparison between you and the mob + other factors.

Also I know weather can affect spell damage but can weather day or other odd variables affect macc?

[Kaylia]

Then compare it to something with a sample size of 300? You could flip a coin 10 times and get 10 heads, then compare it to a sample size of 100 where you got 50/50.

You're joking, right?

Comparing a 500 sample to a 300 sample is justified. The precision on 300 won't be as high, but it doesn't really matter unless you're trying to draw a conclusion on a interval smaller than the precision.

[Kaeko]

You're joking, right?

Comparing a 500 sample to a 300 sample is justified. The precision on 300 won't be as high, but it doesn't really matter unless you're trying to draw a conclusion on a interval smaller than the precision.

You can quantify how justified it is to compare the 2. You can use a variation of a 2-sample T-test of unequal variances (for any statistically inclined people out there).

Generally speaking though, it sounds like it's a reasonable assumption they'll both be comparable.

[DaGod]

/laugh kaeko. Right answer, but hell, i won't use those shitty formulas just to prove "Comparing a 500 sample to a 300 sample is justified"

[Shinta]

I'm mostly interested in the M.Acc testing because I have a feeling that it's better for WHM enfeebles, but HQ Staves are better for RDM. Kind of like how MAB is better until you pass a certain point, and then HQ staves are better. I think testing M.Acc on RDM is not going to be enough.

For example, I'm pretty sure a Dorje would boost a WHM's Banish more than Apollo's if they had no other source of Magic Attack Bonus. Maybe Alkalurops is similar?

[Deader2818]

Speaking of Magic Acc, has it ever been proven to affect Quick Draw because ive never really seen anything that proves Magic Acc helps with the resists on Quick Draw.

[Haroun]

Speaking of Magic Acc, has it ever been proven to affect Quick Draw because ive never really seen anything that proves Magic Acc helps with the resists on Quick Draw.

I don't think any real tests have been done, but there's quite a bit of anecdotal evidence to suggest that it does. Namely, there are alot of CORs saying they've had a noticeable increase in QD accuracy after using Warlock's Roll.

[Auspice]

I don't think any real tests have been done, but there's quite a bit of anecdotal evidence to suggest that it does. Namely, there are alot of CORs saying they've had a noticeable increase in QD accuracy after using Warlock's Roll.

This is the only "evidence" that I've found too sadly.

[Deader2818]

Just seems odd that macc would increase quick draw acc since its stated AGI is what affects that.

I know the argument is that since mattk boosts the dmg on quick draw, that macc should boost the acc but since INT doesnt affect the dmg, I think it kinda makes that seem false. But I dont know, I havent really noticed an increase in acc with macc.