PSA regarding Cuchulain's Mantle

[pchan]

If your base hit rate is say 85%, and do a test with fixed mob/parmeters etc, you have a std deviation of ~ 0.35, which gives you statistical errors :

- with a sample of N=1000, you get +/- 1.8% error so if you find something to be 85% it's anywhere from 83.2% to 86.8%...

If you want the accuracte formula for acc=hitrate you'll need a huge sample size. for +/- 0.1% precision you need N=300,000 ...

So all you can humanly do is "guess", equip +10 acc and compare. If 10acc=5% then you should be getting 85% and 90%, and everyone will be convinced if you get a 0.5% error in both tests. To achieve this you need 10k+ sample sizex2 though...

[Araelus]

Fighting Lesser Colibri for parts of the pDIF function it took me ~100 swings to kill one mob.

Two sample sizes of 10,000+ attack rounds would take about 200 mobs (solo) to get. And while Greater Colibri have more HP (more swings per mob), and I won't be soloing (less swings per mob), I'll get merits in the process.

For that matter, if I make sure to note my party members' accuracy totals! Hey! Wow! This is easier than it sounds.

So this won't take a very long time, compared to other goals. The problem is fixing accuracy totals throughout and bothering to go and fight mobs, not the number of attacks required.

Nice try discouraging us from testing it with the whole "all you can do is 'guess'", though, had me going for a bit there.

Where did the "std deviation of ~ 0.35" come from, by the way?

[pchan]

Nice try discouraging us from testing it with the whole "all you can do is 'guess'", though, had me going for a bit there.

Where did the "std deviation of ~ 0.35" come from, by the way?

Elementary stats, You should know that since you are a pro tester. sqrt(0.85*(1-0.85))=0.35. A test that doesn't give it's margin of error is worth nothing.

I'm not trying to discourage you or anyone, I was just mentionning what kind of data sample you are looking at to prove the 1acc=0.5% hitrate in reference to the small data sample posted earlier.

[ringthree]

13 attack what ? Stop using NQ cerberus mantle already, amemet+1 is better and level 61... Charger mantle has 20 attack, so for my monk it'd be 7 acc 4 str vs 20 attack. I use the general rule 2 attack > 1 acc for monk whether it'd be melee or TP.. so umm, charger.

This is what I do for my MNK.

[ringthree]

Elementary stats, You should know that since you are a pro tester. sqrt(0.85*(1-0.85))=0.35. A test that doesn't give it's margin of error is worth nothing.

I'm not trying to discourage you or anyone, I was just mentionning what kind of data sample you are looking at to prove the 1acc=0.5% hitrate in reference to the small data sample posted earlier.

Also, when the 1 acc = 0.5% hitrate testing was done there was mention that it was likely two or three difference accuracy curves like with pDIF.

The higher you are the less of an effect accuracy (and well anything else) has against what you are fighting.

This is why we need more testing on pretty much everything in the game.

[hagun]

Why do more testing when theory crafting and eye-balling is so much cooler? Life is so simple when you can just retardedly apply "2 attack > 1 accuracy" all over the place.

[Araelus]

Elementary stats, You should know that since you are a pro tester. sqrt(0.85*(1-0.85))=0.35. A test that doesn't give it's margin of error is worth nothing.

I'm not trying to discourage you or anyone, I was just mentionning what kind of data sample you are looking at to prove the 1acc=0.5% hitrate in reference to the small data sample posted earlier.

My apologies, then. I'll bear with you if you'll bear with me. Your post #182 just sounded mean-spirited so I snapped back. (I'm not a "pro tester" by the way or else I wouldn't have asked where the value was from..? ah, whoops, I used when I meant )

In terms of the values, I'm used to only coming up with standard deviations once I have an actual statistical population. I expected to hit for n rounds, then take all those 1(hit) and 0(miss) and throw them into http://upload.wikimedia.org/math/f/f...e7ed4d9f42.png instead of sqrt(μ*(1-μ)) ... which would then get, um, divided by n? Used in the denominator of a z-value? I've never seen someone come up with sigma before actual samples. o_o;

Just goes to show which side of the econ / math divide I stand on.

I'm forgetting, there's some better special equation since the hit/miss is binary, isn't there? It's been a while since I had to do something as "simple" as finding a P-value, so I am rummaging through my old books and whatnot.

Why do more testing when theory crafting and eye-balling is so much cooler? Life is so simple when you can just retardedly apply "2 attack > 1 accuracy" all over the place.

As an aside, hagun, I know you're being sarcastic, and I'm as guilty of that as the next guy, but we will actually be testing this stuff.

Some grumbling gets results inadvertently (eg. when I got mad at pchan and he clarified the point) or because eventually someone finds the data (eg. Suiram's links to the 1acc=0.5% hit posts) but sometimes not being constructive is simply not constructive.

[mdkuser]

If I can add something to the acc=hit rate testing thing... remember that one single level difference can bring up to 5% hit rate difference, it's huge. At one point I wanted to do some acc tests, so I chose a high eva mob.

I Decided to test on those THF goblins in bibiki bay. They pop at three different levels, with 4-5 difference in evasion ( tested it with /check...took me some time). Also one level difference means 4 accuracy penaly ( 2% hit rate). So in total you can have 4-5% hit rate difference, depending on level, and 10% between the lowest level and the highest level. I believe the G-colibris do have different levels as well. For perfect tests you'd have to fight the same mob of same level over and over, which sucks lol. I gave up my tests after noticing that the second THF gob actually was a shared pop with a drk gob.. and of course the drk gob kept poping...fuck it.

[Araelus]

Regarding level difference, you're right - and that's why it's vital to use a parser that sorts by mob level (actually by their xp return but the effect is the same). :D Of course, beastmen having different jobs throws a wrinkle into that, and I'm going to go for single-job monsters.

[pchan]

. I expected to hit for n rounds, then take all those 1(hit) and 0(miss) and throw them into http://upload.wikimedia.org/math/f/f...e7ed4d9f42.png instead of sqrt(μ*(1-μ)) ... which would then get, um, divided by n?

This is the empirical standart deviation ; In practice both are the same.

[Shinryuu]

Then I'd do the BC, sell the mantle, and buy more useful things. Don't get me started on the whole "if you farm it/get the drop, then the monetary value is irrelevant". That is the stupidest argument anyone can ever make. It pisses me off even more when shells use it as justification to fund a relic for a leader. If it isn't a r/e item you can sell it and buy something else. I'd cry if I saw some guy walking around with a Cuchulain's Mantle and a Chiv Chain, then they said "Well, I got the Cuch from the BC".

A thousandfold this. Compare getting Cuchulain's mantle drop to winning a Mog Bonanza prize of exactly Cuchulain's price's worth of gil. What are you going to spend it on? If the answer is Cuchulain's, don't sell. If the answer isn't Cuchulain's, sell it, buy something that increases your overal utility more. And I mean utility in the broad, economic term.

Why do more testing when theory crafting and eye-balling is so much cooler? Life is so simple when you can just retardedly apply "2 attack > 1 accuracy" all over the place.

Lol'd. Where'd the 2 attack > 1 accuracy come from anyway? Theorycrafting at least has merit when your assumptions are correct, you're taking all variables into account and you're doing the math correctly (you wouldn't need to parse anything ever). But the attack for MNK thing sounds like someone's been thinking with their heart instead of their brains again. This isn't fucking Captain Planet, people.

[pchan]

Where'd the 2 attack > 1 accuracy come from anyway?

Because 1 attack is slightly more than 0.25% increase on the things I punch ?

[Shinryuu]

So you found a place to punch stuff where mobs don't spawn with varying levels and where you don't ever have to resort to killing the harder mobs surrounding it?

Get out of Ronfaure, the bunnies hate you.

[Araelus]

This is the empirical standart deviation ; In practice both are the same.

I have no experience with the truly theoretical side of stats, only empirical testing. So thank you for the info on getting a standard deviation before actual testing.

Because 1 attack is slightly more than 0.25% increase on the things I punch ?

And right when I thought we were making some progress, I guess those charts I posted didn't work for you. Because for +1att to be that much of an increase, you have less than 100 base attack, or the mob is Jormungand.

A thousandfold this. Compare getting Cuchulain's mantle drop to winning a Mog Bonanza prize of exactly Cuchulain's price's worth of gil. What are you going to spend it on? If the answer is Cuchulain's, don't sell. If the answer isn't Cuchulain's, sell it, buy something that increases your overal utility more. And I mean utility in the broad, economic term.

Getting people to agree on what an item is worth is a terrible struggle. There's a lot of "worth it?" thrown around, and while you and I know that if someone has the money and wants to spend it, it's worth it to them, even if they are misjudging the value ... found the post I was looking for.

But the attack for MNK thing sounds like someone's been thinking with their heart instead of their brains again. This isn't fucking Captain Planet, people.

http://kingbehemoth.com/icon/14634.png+http://kingbehemoth.com/icon/14630.png+http://kingbehemoth.com/icon/14636.png+http://kingbehemoth.com/icon/14632.png+http://kingbehemoth.com/icon/14642.png = http://kingbehemoth.com/icon/15458.png

[Shinryuu]

Well what I meant is this: In economics, utility is a measure of the relative satisfaction from, or desirability of, consumption of various goods and services. Given this measure, one may speak meaningfully of increasing or decreasing utility, and thereby explain economic behavior in terms of attempts to increase one's utility. In FFXI terms it basically means that a fully informed person can calculate any item's increase in usefulness but usefulness would still vary depending on the person, the person's LS, etc. In before someone *cough* pointing out the awful irony of me quoting a Wiki.

And I might just have to sig that.

Edit:

1. You have included 7 images in your message. You are limited to using 4 images so please go back and correct the problem and then continue again.

Images include use of smilies, the BB code [img] tag and HTML <img> tags. The use of these is all subject to them being enabled by the administrator.

KWEH!?

[Araelus]

I know - the "warm glow" benefit of any item is difficult enough to measure in the real world; somehow it's even worse in a game. At least with things like attack and accuracy we can figure out the impact items have on damage output and then individuals will have to weigh their cash flow and such on their own. I can't do that for them.

And I might just have to sig that.

My work here is done. :D Wait, no, it's not. Still need to pin down accuracy lines, and then get a working auto-chart for plugging in mob stats, and not to mention...

[Bakthi]

Slightly off-topic, since we don't (yet) have a back piece with haste on it, but mostly on-topic at this point:

Araelus: I really liked the charts way back on page 3 comparing the benefits of adding acc/attack. Any chance of throwing haste in there, or just doing haste versus acc? I typically TP in a haste setup in merits on COR, and my acc is reasonable there, but tends to be far less reasonable in Einherjar, and I'm curious about the benefits of haste versus accuracy given various hit rates/haste totals.

[Mayareira]

Well, this is turning into an unnecessary information trash. Mantle before pieces like PCC/Bomblet indicates a poor decision/judgement, however, it cannot be used as an example in this situation. It belongs to the gimp thread, not the efficiency of C. Mantle.

"Hey, C. Mantle doesn't worth the gil because a random x in x server may have a rdm without x piece which could be bought with x amount of gil."

[Shinryuu]

Yeah the steep pricetag of Cuchulain's warrants a thorough "is there nothing I could spend that money on that makes me better off?" and for me I'm almost at the point I'm buying Cuchulain's but for people with less gear and/or more jobs (I have the basic purchasables and some luxury, 4 melee jobs and a BRD) it might not be the case for a bit.

[FailureMidgard]

Regarding level difference, you're right - and that's why it's vital to use a parser that sorts by mob level (actually by their xp return but the effect is the same). :D Of course, beastmen having different jobs throws a wrinkle into that, and I'm going to go for single-job monsters.

Grand Marids in Wajaom only spawn as level 80, BTW. May need someone to whack chigoes for you though.