Forum Settings
       
This Forum is Read Only

Effect of DEX confirmed! Test inside!Follow

#1 Oct 24 2010 at 12:34 PM Rating: Good
16 posts
I decided to check the effects of DEX {especially on accurracy} for myself, after Yabusame Sage DEX Test showed, that there seems to be no effect at all.
This led to interesting discussions with different opinions.

I also had 2 problems with the test.
1. It only showed the effect on Ranged Accuracy
2. The enemy-level was much lower than his.

So i went as my Rank 20 Gladiator to Coerthas where i could fight against the arbor squirell.
They seem to be 50/50 Orange/Green-Difficulty for me. So I asume that there level ist also about Rank 20.

The Test:
I did 20 fights.
10 with 32 DEX. (5 Green - 5 Orange)
10 with 52 DEX. (5 Green - 5 Orange)

First of all: you FEEL the difference. So the results I got didn't suprise me.

32 DEX - 10 fights:
62,5%

32 DEX - 5 Orange
60,4%

32 DEX - 5 Green
65%

As expacted: the accuracy against stronger enemys is much lower.

52 DEX - 10 fights:
70,9%

52 DEX - 5 Orange
67,4%

52 DEX - 5 Green
75%

We dan now see a higher accuracy for both "stong" and weaker monsters.
We also have everywhere a accuracyboost of 7%+.

Yabusame Sage proofed in his STR Test that +20 STR are about +11-12% more Damage.
I assume that its very similar with DEX.
+20 DEX = +12% more accuracy.
____________________________

#2 Oct 24 2010 at 12:57 PM Rating: Excellent
Scholar
**
364 posts
I'm not sure the sample size is large enough or the test exhaustive enough to call it definitively, but I have more confidence in this than the Archer test. The Range system was/is immediately suspect. People seemed to take a little offense to me pointing that out in the other thread, not sure why.
#3 Oct 24 2010 at 1:42 PM Rating: Decent
Scholar
***
1,104 posts
Furia wrote:
I'm not sure the sample size is large enough or the test exhaustive enough to call it definitively, but I have more confidence in this than the Archer test. The Range system was/is immediately suspect. People seemed to take a little offense to me pointing that out in the other thread, not sure why.
I personally agreed with you in that thread and I still do. ARC range mechanics change everything and there is no way to determine just HOW. Plus the projectile itself has an accuracy rating that SHOULD make archers the most accurate class in the game by far - but it doesnt seem to do that.

I dont think the effect of DEX on melee accuracy was ever in dispute, was it?
____________________________

Melaahna Valiera
#4 Oct 24 2010 at 1:43 PM Rating: Good
***
1,083 posts
I also concur with the results as they make logical sense. However, I'm seeing too many 'tests' done with a sample size that are, for all intents and purposes, statistically meaningless. This test doesn't really confirm anything.

Sample sizes of 100+ should be the baseline. It takes time, yes, but we all know how easy it is to go 0/50 on VE or LL and for the next guy to say he went 1/1 (that was me btw ^.^ ). You just don't see good results without a really large sample.

Edited, Oct 24th 2010 3:45pm by rikkuotaku
____________________________
Wada: "There may be some areas of testing that were lacking or too rigid."
#5 Oct 24 2010 at 1:51 PM Rating: Decent
*****
11,576 posts
rikkuotaku wrote:
I also concur with the results as they make logical sense. However, I'm seeing too many 'tests' done with a sample size that are, for all intents and purposes, statistically meaningless. This test doesn't really confirm anything.

Sample sizes of 100+ should be the baseline. It takes time, yes, but we all know how easy it is to go 0/50 on VE or LL and for the next guy to say he went 1/1 (that was me btw ^.^ ). You just don't see good results without a really large sample.


And just to be picky, even a sample size of 100 is statistically insignificant. About the best you could call a sample size of 100 might be "proof of concept". Until people with a ton of time on their hands and rank capped classes get down to some serious number crunching or SE comes out and tells us explicitly, nothing is "confirmed".
#6 Oct 24 2010 at 2:06 PM Rating: Good
rikkuotaku wrote:
I also concur with the results as they make logical sense. However, I'm seeing too many 'tests' done with a sample size that are, for all intents and purposes, statistically meaningless. This test doesn't really confirm anything.

Sample sizes of 100+ should be the baseline. It takes time, yes, but we all know how easy it is to go 0/50 on VE or LL and for the next guy to say he went 1/1 (that was me btw ^.^ ). You just don't see good results without a really large sample.

This. It's definitely a good start though, and more testing (by multiple people) is really the only way to prove the theory short of SE coming out and telling us. Honestly, the DEX issue is probably less of a debatable issue (it's already pretty commonly acknowledged that it improves accuracy, based on extensive testing in other games), but it also brings up the importance of using the scientific method in any sort of testing. It is the basis for every piece of research conducted for a reason, so take advantage of it and do things right! There's still a lot to learn in this game, and the only way to do it is through SCIENCE.
____________________________
IRC: Kagutsuchi

FFXI:
Tolite
Formerly of Lakshmi, now in Asura

FFXIV:
Kagu Tsuchi
Ultros
#7 Oct 24 2010 at 2:25 PM Rating: Decent
8 posts
Good work. This is a suggestion to anyone running tests in the future and trying to fight the same level mobs: you can still have multiple levels of a mob in a color category, such as "green." However when you kill a mob, it awards a certain number of physical experience. The same level / same type mob will always give you the exact same physical experience (unless you physical level up). So when running tests is a good idea to break down numbers by mob experience level, that way we can be sure the control and test cases were both tested on the same thing.


Furia wrote:
I'm not sure the sample size is large enough or the test exhaustive enough to call it definitively, but I have more confidence in this than the Archer test. The Range system was/is immediately suspect. People seemed to take a little offense to me pointing that out in the other thread, not sure why.


One thing I'd like to see is a rudimentary test of if there is even a "range system." When grouping I've tried standing varying distances from the mob and I haven't noticed any difference in damage or felt any more/less accurate. If range does have an impact on something, it isn't obvious to me.


Edited, Oct 24th 2010 4:27pm by Zerrius
#8 Oct 24 2010 at 2:31 PM Rating: Good
Scholar
***
1,104 posts
Zerrius wrote:
One thing I'd like to see is even a rudimentary test of if there even is a "range system."
Play an archer for 10 or 15 levels; The range system will make itself apparent. It doesn't apply to melee though - if you're close enough to hit then your accuracy is static anywhere within that distance. Not sure why the only class that pays for damage needs to be penalized also but thats another issue entirely.
____________________________

Melaahna Valiera
#9 Oct 24 2010 at 2:52 PM Rating: Good
16 posts
maybe the 20 fights are a little misslleading, because it does not sound so much, but 20 fights are 350 Attacks.
More fights sure would bring more precise numbers, but it was never my intention to offer a 100% accurate result.
The important conclusion it, that 20DEX will give you a decent accuracy advantage. If its 5%, 10 % or 15% is not very important.
20 DEX + are 3-4 physical Level Ups (with only investing in DEX).
If that had no effect at all in 350 Attacks (even due to bad look )would it still show, that DEX cant be important.

____________________________

#10 Oct 24 2010 at 3:30 PM Rating: Good
***
1,083 posts
Aurelius wrote:
rikkuotaku wrote:
I also concur with the results as they make logical sense. However, I'm seeing too many 'tests' done with a sample size that are, for all intents and purposes, statistically meaningless. This test doesn't really confirm anything.

Sample sizes of 100+ should be the baseline. It takes time, yes, but we all know how easy it is to go 0/50 on VE or LL and for the next guy to say he went 1/1 (that was me btw ^.^ ). You just don't see good results without a really large sample.


And just to be picky, even a sample size of 100 is statistically insignificant. About the best you could call a sample size of 100 might be "proof of concept". Until people with a ton of time on their hands and rank capped classes get down to some serious number crunching or SE comes out and tells us explicitly, nothing is "confirmed".


I agree, I just didn't want to sound like an elitist XD I hate it when people at least devote some time to trying to form a theory and someone on this forum just point blank shoots them down without even reading what they wrote saying "sample size too small" or "test meaningless". I mean, I'm happy at least someone put forth the effort to form a hypothesis. Gotta start somewhere. Still, agree that nothing is confirmed at all.
____________________________
Wada: "There may be some areas of testing that were lacking or too rigid."
#11 Oct 24 2010 at 3:42 PM Rating: Decent
***
1,004 posts
Owasephiroth wrote:
maybe the 20 fights are a little misslleading, because it does not sound so much, but 20 fights are 350 Attacks.
More fights sure would bring more precise numbers, but it was never my intention to offer a 100% accurate result.
The important conclusion it, that 20DEX will give you a decent accuracy advantage. If its 5%, 10 % or 15% is not very important.
20 DEX + are 3-4 physical Level Ups (with only investing in DEX).
If that had no effect at all in 350 Attacks (even due to bad look )would it still show, that DEX cant be important.



thats all well and good, but people like to know as close to exact numbers as possible, especially in a game like this. That way you know if you need more or if you have to much.
#12 Oct 24 2010 at 3:51 PM Rating: Default
Scholar
***
3,226 posts
Quote:
Effect of DEX confirmed!

Oh boy!
Quote:
Test inside!

Surely, if the effect of DEX is confirmed, the amount of data here reaches into the hundreds--absolutely necessary when exploring hit rates, proc rates and anything that depends on a percent chance!
Quote:
The Test:

MY BALLS ARE BURSTING WITH EXCITEMENT!
Quote:
I did 20 fights.

Rate down, banned, FBI notified, citizenship revoked, basic human rights denied.

I went to public school--an American public school. Even I know "20 fights" is too low to make any conclusion about the nature of DEX v. Hit rate.

Look, I am happy there are people who want to test game mechanics, but don't let the excitement of discovery affect good judgment. This is the third game mechanics thread I've seen that makes bold assertions with limited test data.

"Good starts" and "preliminary tests" aren't helpful anymore. You can state an idea, like a plan to test DEX v. hit rate, and receive criticism before you begin testing. That is fine. But once you actually begin the testing session, you don't stop as soon as you're finished jacking off and name your piddly sample size "DEX confirmed!"

There will always be some caveman who wanders into these threads, scrolls to the bottom of the OP, and uncritically designates the OP correct beyond a shadow of a doubt. He tells his dullard LS, they believe it, and it propagates itself from moron to moron, server to server. And then we have a superstition that never goes away, even after SE says "No, DEX does not increase drop rates."

That said, I encourage OP to pop open a cream soda, get a bag of Pirate's Booty and spend the rest of the day in Coerthas punching vermin.
____________________________
w(°o°)w
#13 Oct 24 2010 at 3:58 PM Rating: Excellent
*
145 posts
Just to point out that the sample size should not be based off the number of mobs killed, but the number of swings taken. This is really a test of the probability that at any given swing it will either hit or miss. So assuming the mob is the same level for each battle, the sample size would be the number of swings he made on that level specific mob. So if he did in fact swinging a significant number of times on the mob, then the sample size could be rather large, and therefor statistically significant.
#14 Oct 24 2010 at 4:10 PM Rating: Decent
*****
11,576 posts
rikkuotaku wrote:
I agree, I just didn't want to sound like an elitist XD I hate it when people at least devote some time to trying to form a theory and someone on this forum just point blank shoots them down without even reading what they wrote saying "sample size too small" or "test meaningless". I mean, I'm happy at least someone put forth the effort to form a hypothesis. Gotta start somewhere. Still, agree that nothing is confirmed at all.


It takes quite a lot to convince me that something is "confirmed" when it comes to "hidden" game mechanics. And it's not because I'm trying to be mean guy, it's because I don't really want this to turn into another case of FFXI theorycrafting run amok. The last thing this community needs is a bunch of players running around trying to run into a dodo's left shoulder to interrupt Rancid Belch while facing west-northwest because dodos are cloudkin which are based on wind. We're already facing the uphill battle of trying to teach tanks that Provoke and Taunt don't generate as much enmity as everyone wants to think and crafting facility support doesn't do much unless a recipe calls for it. Which is not to say that things like the OP's conclusions following their test are flat out false simply because the sample size wasn't really significant, but to be careful about labeling things as "confirmed" or "fact" when there is still plenty of margin for other alternative conclusions to be reached.
#15 Oct 26 2010 at 6:32 AM Rating: Decent
16 posts
@Almalexia Scholar

What the **** is your problem?
If you don't like the result fine, but why do you have take it out on me?
I can't care less if you want to believe it or not. I did the test for myself and i thought it would be a nice gesture to share my results as a help for other players.
Maybe it was a mistake considering how insanly "nice" this comunity seems...
Also it's very strange to bash someone that hard without having any numbers or data. My suggestion: Make youre own test if youre disliking others that much...

And angain. We're talking here about 350 Attacks. Why should 350 Attacks be to low to determine if 20 (!) DEX have any effect at all?
As an example: With my 52 DEX, 1 out of 10 fights had over 6 "misses" (10%). With 32DEX i had 6/10 fights with over 6 "misses" (60%)
So the 10 vs 60 are also just coincident?
The data makes in every espect perfectly logical sense and i'm a 100% certain that 100 times more data will bring only slitely more precise numbers but they will not change the outcome.
If i wanted to proof the size of the effect I would have done more fights but i only wanted to proof, if there is a effect at all.

Quote:
I went to public school--an American public school. Even I know "20 fights" is too low to make any conclusion about the nature of DEX v. Hit rate.

I've studied statistics and scientific methodology for four semesters (BA sociology/psychology) - not in America.
But i'm not shure why this is important.




Edited, Oct 26th 2010 8:54am by Owasephiroth
____________________________

#16 Oct 26 2010 at 7:01 AM Rating: Good
***
1,083 posts
Owasephiroth wrote:

If i wanted to proof the size of the effect I would have done more fights but i only wanted to proof, if there is a effect at all.


But 20 fights, even 340 attacks, really isn't proof of even that - it could just be that it was some elemental day that we don't even know about yet or the moon phase (not joking). I only bring XI into this because its the same company, but for every person who did 100 synths to prove X there was another that did the same 100 synths to disprove X. You need MOUNTAINS of data to prove something in an SE game. XD You have a hypothesis, but that is not "confirmed!" nor is it a proof.

Quote:

I've studied statistics and scientific methodology for four semesters (BA sociology/psychology) - not in America.


Go forth and put it to use my son...

Edited, Oct 26th 2010 9:02am by rikkuotaku
____________________________
Wada: "There may be some areas of testing that were lacking or too rigid."
#17 Oct 26 2010 at 8:34 AM Rating: Decent
*
70 posts
Thanks for the effort, OP.

Using some margin of error calculator online, I found out that at 350 samples your margin of error is ~5% assuming a 95% level of confidence. Since the difference between 32DEX and 52DEX is only 7%, I think you would need to bring the margin of error down to 3% or less to convince the skeptics here, and for that you need to do 650 more swings. Do eeeet!

#18 Oct 26 2010 at 8:51 AM Rating: Decent
Scholar
*
247 posts
So...I understand the naysayers. SE is ridiculous at times when it comes to what influences certain game mechanics.

To "confirm" something you need to test:

- On different IG days
- Different mob type
- Elemental influence
- Day elemental influence
- So on
- And so forth

SE puts all types of crazy mechanics in place. This game should be no different. However, the test does go to show that at least something happened when more DEX was added. Is DEX 100% worthless? Nope. Is mine still 40+ points below my STR? Yes.

-Kash
____________________________
I "BAM!" therefor I AM!
#19 Oct 26 2010 at 11:28 AM Rating: Excellent
***
1,021 posts
From the OP, we have:

 
         Total  Orange  Green 
Dex32    0.625  0.604   0.650 
Dex52    0.709  0.674   0.750 


Also, a later post clarified that there were 350 attacks in total (I'm assuming this is approximate).

For further analysis, we need the actual number of trials for each condition. Unfortunately, from the information posted, this can't be determined exactly. Assuming that all mobs take the same number of hits to kill, we can solve for the number of hits and from that calculate the number of trials. I am assuming here that the Orange mobs took the same number of hits as the Green mobs which is probably a bad assumption, but is unlikely to affect the final outcome of the analysis.

Let H=number of hits to kill an Arbor Squirrel.

5*H/.604 + 5*H/.65 + 5*H/.674 + 5*H/.75 = 350

solving for H, H=11.645. Back calculating the number of trials (swings) in each condition gives:

 
         Total  Orange  Green 
Dex32   185.98  96.40   89.58 
Dex52   164.02  86.39   77.63 


The question of interest is: Does going from Dex32 to Dex52 increase the hit rate of a random swing when fighting a random Arbor Squirrel? From the OP, we know that the sampled % chance of fighting an orange squirrel is the same as the population % chance of fighting an orange squirrel (with some minor error which again will not affect the analysis). Therefore, the samples gathered are distributed in the same manner as the population we wish to sample from.

Applying a two-sample unpooled t-test with unequal variances (Welch's t-test) we get the following.

Let population 1 be the Dex52 case and population 2 be the Dex32 case. The null hypothesis is that there is no increase in hit rate due to Dex changing from 32 to 52. x is the probability of hitting. xbar is the mean of x. xbar1 is the estimated mean of x for population 1. var1 is the estimated variance of x for population 1. n1 is the number of samples in from population 1.

t = ((xbar1 - xbar2) - d0) / sqrt(var1/n1 +var2/n2)

xbar1 = .709
xbar2 = .625
n1 = 164.02
n2 = 185.98
d0 = null hypothesis difference = 0.0

var1 = E(x^2) - E(x)^2

since x only takes on the values 1=hit and 0=miss, x^2=x for all possible values of x.

var1 = E(x^2) - E(x)^2 = E(x) - E(x)^2 = xbar1 - xbar1^2 = .709 - .709^2 = .206
var2 = .625 - .625^2 = .234

t = (.709 - .625) / sqrt(.206/164.02 + .234/185.98)
t = 1.6739

df = degrees of freedom
df = (var1/n1 + var2/n2)^2 / ((var1/n1)^2/(n1-1) + (var2/n2)^2/(n2-1))
df = 346.65

Looking up on a t-distribution table for a 1-sided test (since the null hypothesis is that dex does not increase accuracy), we see this is significant at the 95% confidence level.
#20 Oct 26 2010 at 11:48 AM Rating: Good
Scholar
**
805 posts
Thank you Red Hobbit for providing analysis instead of just spewing hate like other posters more interested in +1 than providing any real contribution to the dialog. And thanks to OP for providing the data that appears to be "statistically significant".
____________________________
Lokithor
Shiva Server
Linkshell: Soulfire
Rng/Blm/Thf/Brd/Dnc
#21 Oct 26 2010 at 11:57 AM Rating: Good
****
9,526 posts
Thanks for the test OP. Don't get discouraged.
____________________________
lolgaxe wrote:
When it comes to sitting around not doing anything for long periods of time, only being active for short windows, and marginal changes and sidegrades I'd say FFXI players were the perfect choice for politicians.


#22rikkuotaku, Posted: Oct 26 2010 at 12:24 PM, Rating: Sub-Default, (Expand Post) I don't think this is in dispute. Its only a 95% confidence level for the specific situation the OP was in, which has so many variables, that using it for the application of a generic DEX statement is statistically meaningless.
#23 Oct 26 2010 at 12:44 PM Rating: Good
***
3,530 posts
Furia wrote:
I'm not sure the sample size is large enough or the test exhaustive enough to call it definitively


This. If we had much higher numbers the difference, if any, would be readily available; a good start, though!
____________________________
"... he called to himself a wizard, named Gallery, hoping by this means to escape the paying of the fifteen hundred crowns..." (Machen 15)

"Thus opium is pleasing... on account of the agreeable delirium it produces." (Burke para.6)

"I could only read so much for this paper and the syphilis poem had to go."
#24 Oct 26 2010 at 1:02 PM Rating: Default
*
64 posts
wow, the amount of flame a player gets for trying to help the community. why do all forums have so many asshats? who cares if the sample range is low? if many players do the same the sample size grows exponentially. Kudos to OP.
#25 Oct 26 2010 at 1:13 PM Rating: Decent
**
445 posts
Almalexia wrote:

I went to public school--an American public school. Even I know "20 fights" is too low to make any conclusion about the nature of DEX v. Hit rate.


Do you mean an American public school fail to teach what is "appreciation"?
____________________________
Timfung (Alexander) 75WHM72RDM46PUP Cooking81 (retired)

http://lodestone.finalfantasyxiv.com/rc/character/status?cicuid=1662475
#26 Oct 26 2010 at 1:24 PM Rating: Excellent
****
7,106 posts
No one here has a problem with the Op's contribution. The Op is just wildly overstating what he achieved.

He gave some data to the community. That's awesome. If more people did that, things would be much easier to prove. The data isn't quite as complete as one might hope, but it's still data.

He did not, however, prove anything. Nothing has been "confirmed." And, he's getting a little butthurt when people point that out, which maybe isn't surprising since he was proud enough of his results that he needed to leave the ongoing thread about DEX and make a whole new one declaring the mystery completely solved. If he'd kept his contributions in the existing thread, and not announced that with only 20 mobs he confirmed! something that no one else had yet confirmed, he probably would have gotten a much warmer reception.

The Op should be applauded for his contributions. He just needs to get over himself a little.

*edit* And I suspect he's probably right. My guess is that DEX does roughly the same thing in XIV as it did in XI. But, even when the Op's conclusions match my own, I can still see that his data didn't prove anything.



Edited, Oct 26th 2010 3:26pm by Caesura
#27 Oct 26 2010 at 1:28 PM Rating: Excellent
*
66 posts
The weakness of this analysis isn't the number of attacks, per se, but the possible existence of oh-so-many conflating factors. Biggest one is probably going to be the possibility of gear damage effects. If all of the tests were performed with all gear items in the clear, well and good; if some gear damage was sustained during the fights, and not repaired until after, that would also affect the data involved.

The other question is, what's the width of the level range for "green" and "orange"? Is it just one level? If that's the case, then we should expect pretty consistent results. If there are two levels of mob that fit in that range, then the samples could be affected by a significant number of mobs being "low green" or "high green", which you wouldn't have any way to determine.

There's also the question of the soft cap, but that's not necessarily an issue here - for a combat class, at rank 20, it's probably not at 50. You'd need a lot of data points to really show that sort of thing, not just two.

Props to RedHobbit for busting out the statistical analysis. Keep in mind the limits of what he's demonstrated - assuming the numbers we have from Owasephiroth are correct, assuming that gear damage wasn't a conflating factor, assuming that the dex 52 sample didn't include a lot of "high green" or "high orange" versus the dex 32 sample including "low green" or "low orange" (if that's even possible)... then the data shows that DEX has an affect on accuracy, to better certainty than we have for global warming predictions. ;p
#28 Oct 26 2010 at 1:39 PM Rating: Decent
16 posts
@rikkuotaku Sage
Accuracy isn't a hidden effect of dexterity like the moonphases for crafting. It says so in the descrpition.
And why should - in this case - the moon or some other hidden effect have more influence than the "official" stat?
Sure, their might be a posibility that there are hidden factors which we dont know yet, but it's still far far more likely that DEX has the most influence on the accuracy.

@RedHobbit Sage
First of all i would like to thank you very much for youre time and very good contribution.
And to shortly clear something up:
My goal was to fight against 5 Green and 5 Orange Enemys from the start (No coincident).
Also everything was based on the number of attacks:

184 Attacks 69 Misses with 32 DEX (Orange 101 / Green 83)
165 Attacks 48 Misses with 52 DEX (Orange 89 / Green 76)

@Wolfenstein
95% means something different.
It means that my 69 Misses with 32 DEX could be 66-72 Misses and my 48 Misses with 52 DEX could have been 46 to 50.
It means, even if we calculate the error witch is caused by the "low" number of attacks as high as possible, we would get 66 Misses with 32 DEX and 50 Misses with 52 DEX.
You would still miss 16 times less with 20 more DEX. Even if my resualts would be as faulse as they can get.

@Caesura
I think people could have missunderstud.
I surely am not proud of the results. why should I? The reason why I started a new thread with that headline (which seems to upset a lot of people) is that i was excited and happy.
I simply feard that DEX had no effect, that it could be broke and this (in my opinion) was proven wrong.
I was just happy, that I now CAN do something against my poor Hit Avr. Other than waiting for my next Weapon.
Thats all their is.



Edited, Oct 26th 2010 3:56pm by Owasephiroth
____________________________

#29 Oct 26 2010 at 2:57 PM Rating: Excellent
First of all, I applaud the effort put forth by the OP here, and the spirit of exploration. It is a very small sample size he used however, and I am doing a test very similar to my first one on a rank 14 gladiator hopefully on Wednesday. The only difference between the archer test and the gladiator one will be the weapon; same gear, same mobs, same level.

I will also be testing the effects of VIT on damage taken and the effects of DEX (if any) on evasion. Following that, I will be testing a 14 CON and 14 THM and the INT, MND, and PIE stats (hopefully this weekend).

Everyone keep up these helpful threads!
____________________________
#30 Oct 26 2010 at 5:07 PM Rating: Good
*
66 posts
The real question is, does DEX have a greater effect than +accuracy? Or is there a combined limit, where your hit percentage won't appreciably increase even if you stack further DEX and +acc?

Wish I was still working at the university, where I had access to Stata...

Owa, what the 95% confidence means is that there's a 1 in 20 chance that every bit of data you collected was correct, and that there weren't any conflating factors involved, but your conclusion is flat wrong based on pure luck. (Hey, it's possible to flip a coin twenty times and get twenty heads; it's not likely, but it's possible.)

That's why more observations would improve the quality of the data. The larger the number of observations, the less chance that you're looking at the twenty-heads situation - your observed mean is more likely to reflect the theoretical mean.
#31 Oct 26 2010 at 7:07 PM Rating: Excellent
Scholar
***
3,226 posts
Owasephiroth wrote:
What the **** is your problem?

I am not convinced that you have "confirmed the effect of DEX" based on the data you have presented.

But you don't care what I believe, though you should. Whether you like it or not, I am your ideal audience. I play FFXIV. I play exclusively DoW classes. I am deeply interested in the workings of DEX. If you can convince me that something is the case, you would change the way I build my character. And not just mine, but thousands of other characters. Tens of thousands.

Maybe you have not have played FFXI. Maybe you weren't around during the last 8 years as people tested this mechanic and that trying to make sense of a game as super-secretive as FFXI. Maybe you are not aware of the systems of belief that sprung from incorrect information, some of which are still hailed today as correct.

Intuition. Superstition. Many people too easily believe what they see, read and hear. But this is a video game, and "feeling" doesn't increase hit rates, drop rates, damage and so on (not in any games I've played, anyway). Yet some people cling to feeling and wrong knowledge regardless; this being an MMO, we are bound to run into those people.

So what the **** is my problem? It is not you personally, OP. And it is not necessarily your work, which may very well prove correct. It is, as Caesura points out, another thread--one more voice in a cacophony of voices from people who "know how X works now," but lack a convincing amount of data. There are so many people on these forums and the internet in general who are quick to bob their heads and change their game plan. Whether the information is correct or not is inconsequential to them; it is the latest test, it has a bunch of replies and it has math in it, so that is good enough for them.

Do you, OP, honestly want to appeal to them instead of us who rightly ask for more?

If you want to be persuasive, you need to thoroughly stand out. Kanican's enmity testing for FFXI is exemplar. Granted, testing something like enmity is more complex than DEX v. hit rates, but you can see what it takes to become a voice of authority in the community rather than just a voice. I am not saying you must make a blog and spend months on testing to make a valid point, but it goes to show how far we as a community have come from "carrying a four-leaf mandragora bud adds TH because I got the item I wanted."

Also: don't let the cuddly, smiley-facey, ratey-uppy image of ZAM convince you that everyone who posts here is a polite automaton looking for someone to congratulate. Some of us are interested in the gritty workings of the game and are also courageous enough to disagree with good reason or criticize. If you made this thread specifically for a treat and a pat on the head, it would have been better if you had never made the thread at all.

tl;dr--The years of bold "testers" who have come before you have made me hypercritical, sarcastic and bitter. Don't take it personal, OP, but test more if you actually care about the discovery instead of the glory of discovery.

Edited, Oct 26th 2010 9:08pm by Almalexia
____________________________
w(°o°)w
#32 Oct 26 2010 at 8:14 PM Rating: Good
***
1,021 posts
AvatarexADV wrote:
The weakness of this analysis isn't the number of attacks, per se, but the possible existence of oh-so-many conflating factors. Biggest one is probably going to be the possibility of gear damage effects. If all of the tests were performed with all gear items in the clear, well and good; if some gear damage was sustained during the fights, and not repaired until after, that would also affect the data involved.

The other question is, what's the width of the level range for "green" and "orange"? Is it just one level? If that's the case, then we should expect pretty consistent results. If there are two levels of mob that fit in that range, then the samples could be affected by a significant number of mobs being "low green" or "high green", which you wouldn't have any way to determine.

There's also the question of the soft cap, but that's not necessarily an issue here - for a combat class, at rank 20, it's probably not at 50. You'd need a lot of data points to really show that sort of thing, not just two.

Props to RedHobbit for busting out the statistical analysis. Keep in mind the limits of what he's demonstrated - assuming the numbers we have from Owasephiroth are correct, assuming that gear damage wasn't a conflating factor, assuming that the dex 52 sample didn't include a lot of "high green" or "high orange" versus the dex 32 sample including "low green" or "low orange" (if that's even possible)... then the data shows that DEX has an affect on accuracy, to better certainty than we have for global warming predictions. ;p


These are all excellent points. Gear damage and mob level differences would both violate the assumptions that the statistical tests require that the samples are independent and identically distributed.

I've seen a few posts about other possible confounding factors such as day of week, moon phase, etc. Unless these variables are different between the tests (which is possible since I'm not sure how long it takes to respec the dex), it doesn't invalidate the test. Of course, the test tells us nothing about conditions different from what was tested. It seems most reasonable, though, to prefer a simpler explanation in the absence of evidence to the contrary (and preferably control for as much as possible while testing). I can't count the number of theories that people came up with in FFXI with basically no evidence at all.
#33 Oct 26 2010 at 9:00 PM Rating: Decent
Scholar
*
87 posts
The fact that we even need to test what the statistics of this game is insulting. Why can't they be like every other game developer and show what the stats do? Calculating Min/Max, "Breakpoints", and tiers is something entirely different, but when I invest points in dex I should see the physical manifestation of it in my accuracy stat. There's no reason not to.
____________________________
Score: Good [4.20] <--Best rating ever
#34 Oct 26 2010 at 9:48 PM Rating: Good
**
518 posts


Quote:
Accuracy isn't a hidden effect of dexterity like the moonphases for crafting. It says so in the descrpition.
And why should - in this case - the moon or some other hidden effect have more influence than the "official" stat?
Sure, their might be a posibility that there are hidden factors which we dont know yet, but it's still far far more likely that DEX has the most influence on the accuracy.


Sounds like what you have is a hypothesis there. Dex affects accuracy. Now your theory based off of your original post is that
20 DEX=7% higher hit rate

Follow me here, it'll get complicated.....

Theory: Increasing 20 DEX Will increase your hit rate by 7%
Hypothesis: DEX affects ACC
Dependant Variable: Hit Rate
Independent Variable: DEX

Sample size 350 Margin of error + or - 5%

Varriant on 20DEX affecting ACC= +12%-+2% increase of hit rate

independent Variable(a): ACC on weapon
independent Variable(b): Level penalty on mob (possible)
independent Variable(c): Crit hits vs non-crit hits resulting in more or less swings in sample

That's plenty

So lets go back and check our hypothesis
Dex affects hit rate
except for the fact that SE has said it does as such I would say your data set doesn't even prove your hypothesis. A possible +2% hit rate on +20DEX is not even close to significant if you consider those (a,b,c) factors. Those alone can explain the +2% hit rate on the two different groups.

Since we can't even conclude that there is a correlation between DEX and hit rate other than the fact that SE said so, I can't see you hypothesis even coming close to validating your theory.

I'm sorry but based on your results more research is needed and anyone who says that DEX effects hit rate better be ready to admit there is nothing other than the word of SE to back that statement up.

Good for trying, stop getting but hurt.

For future tests

Why not test low level mobs, i.e newb mobs....take a base sample of 600+ fights to find the base hit rate. Add 10 DEX, repeat experiment above. Add 10 more DEX, repeat experiment above.
Find mob of same level reduce DEX-20, repeat experiment...you get the point..You need first a baseline of a easy to hit mob. One who's level should only be 1. So we don't even have to consider level correction.
#35 Oct 27 2010 at 7:45 AM Rating: Default
2 posts
If he had actually faught 40 mobs at ~the same level but titled this, '1,000 attacks at 32/52 dex.' would we be treating this guy differently?

Also, some say could be 2% according to the +/-5% results, it also could be 12% the other direction am I right?

Also, Why assume his gear damage could make the results skewed towards DEX having less effect? I'm asking what I think should have been asked very early on with this discusion. Was your gear in better shape, or worse shape with 32 DEX versus 52 DEX, EI: Did you start your test fully repaired @ 32 dex, and finished slightly damaged with 52 dex, or was it the other way around? If your gear was fully repaired with the lower dex number and damaged with the higher dex number it leans even more to the possibility that DEX does effect acc.

I myself would not hang my hat on a sample size of ~600, that's just me. I do think that a few people here are looking at your test feeling extreamly optimistic looking from the far left wanting to beleive anything that supports the outcome, thinking, "I knew it!" or being complete kill-joys from the extream right (no political views intended) ready to discount any bit of information you could provide, siting any reason they can think of.

With that said: Nice attempt at getting information out there, poor assumtion thinking/saying anything as been proven. I do hope more people will look at test like these 'down the middle' and not pick a stance so suddenly.
#36 Oct 27 2010 at 10:27 AM Rating: Good
The OP should have seen people telling me I needed bigger sample sizes to satisfy them (my sample size was 300 attacks each at 20, 40, 60, and 80 DEX...1200 total attacks) and gone at least as far, or else be prepared for the skepticism, particularly because he was refuting the evidence in my post and the prevalent opinion of the community. That being said, everyone needs to play a little more nicely. We can flame each other to **** in the "I'm quitting, this game sucks" threads.
____________________________
#37 Oct 28 2010 at 7:33 PM Rating: Default
*
116 posts
Almalexia wrote:
Owasephiroth wrote:
What the **** is your problem?

I am not convinced that you have "confirmed the effect of DEX" based on the data you have presented.

But you don't care what I believe, though you should. Whether you like it or not, I am your ideal audience. I play FFXIV. I play exclusively DoW classes. I am deeply interested in the workings of DEX. If you can convince me that something is the case, you would change the way I build my character. And not just mine, but thousands of other characters. Tens of thousands.

Maybe you have not have played FFXI. Maybe you weren't around during the last 8 years as people tested this mechanic and that trying to make sense of a game as super-secretive as FFXI. Maybe you are not aware of the systems of belief that sprung from incorrect information, some of which are still hailed today as correct.

Intuition. Superstition. Many people too easily believe what they see, read and hear. But this is a video game, and "feeling" doesn't increase hit rates, drop rates, damage and so on (not in any games I've played, anyway). Yet some people cling to feeling and wrong knowledge regardless; this being an MMO, we are bound to run into those people.

So what the **** is my problem? It is not you personally, OP. And it is not necessarily your work, which may very well prove correct. It is, as Caesura points out, another thread--one more voice in a cacophony of voices from people who "know how X works now," but lack a convincing amount of data. There are so many people on these forums and the internet in general who are quick to bob their heads and change their game plan. Whether the information is correct or not is inconsequential to them; it is the latest test, it has a bunch of replies and it has math in it, so that is good enough for them.

Do you, OP, honestly want to appeal to them instead of us who rightly ask for more?

If you want to be persuasive, you need to thoroughly stand out. Kanican's enmity testing for FFXI is exemplar. Granted, testing something like enmity is more complex than DEX v. hit rates, but you can see what it takes to become a voice of authority in the community rather than just a voice. I am not saying you must make a blog and spend months on testing to make a valid point, but it goes to show how far we as a community have come from "carrying a four-leaf mandragora bud adds TH because I got the item I wanted."

Also: don't let the cuddly, smiley-facey, ratey-uppy image of ZAM convince you that everyone who posts here is a polite automaton looking for someone to congratulate. Some of us are interested in the gritty workings of the game and are also courageous enough to disagree with good reason or criticize. If you made this thread specifically for a treat and a pat on the head, it would have been better if you had never made the thread at all.

tl;dr--The years of bold "testers" who have come before you have made me hypercritical, sarcastic and bitter. Don't take it personal, OP, but test more if you actually care about the discovery instead of the glory of discovery.

Edited, Oct 26th 2010 9:08pm by Almalexia


PWND!
#38 Oct 28 2010 at 8:19 PM Rating: Decent
*****
11,576 posts
Yabusame wrote:
The OP should have seen people telling me I needed bigger sample sizes to satisfy them (my sample size was 300 attacks each at 20, 40, 60, and 80 DEX...1200 total attacks) and gone at least as far, or else be prepared for the skepticism, particularly because he was refuting the evidence in my post and the prevalent opinion of the community. That being said, everyone needs to play a little more nicely. We can flame each other to **** in the "I'm quitting, this game sucks" threads.


There's merit and benefit in challenging the results of other peoples' "research". It happens consistently in academic circles as just part of the process. I think some people here are starting to take it a bit too far, however, and discrediting the base of the concept "just because" because the sample sizes haven't been large enough or some other factor that is relevant in academic circles, but we're not trying to cure cancer here. Some people are making a habit of it but they aren't making a contribution of their own and it starts to make them seem a bit...I dunno...useless is a word that comes to mind.
#39 Oct 29 2010 at 2:49 AM Rating: Excellent
Sage
****
8,779 posts
Quote:
Rate down, banned, FBI notified, citizenship revoked, basic human rights denied.


funny, but a bit much.

now, as a guy that comes from a forum who named someone "Mittens, Destroyer of Worlds," i have a good sense of humor, and a good idea of when someone is being facetious, or sarcastic, or purposefully stupid or something. however, a lot of people dont, and especially when its somebody trying to go out of their way to help the community, regardless of their sample size, you'll need to be careful with how you phrase any criticism.

otherwise youll get ******** like me coming in and telling you to get off your lazy *** and start expanding the sample size.
____________________________
Quote:
The thing about me is that apparently it's very hard to tell when I'm drunk. So I feel like I'm walking sideways on a UFO and everyone else sees me doing the robot like a pro.
- MojoVIII
i have bathed in the blood of many. my life was spent well.
feral druids do it on all fours.
The One True Prophet of Tonkism.

http://therewillbebrawl.com/
#40 Oct 29 2010 at 5:09 AM Rating: Decent
Scholar
Avatar
****
4,148 posts
Quor wrote:
funny, but a bit much.

now, as a guy that comes from a forum who named someone "Mittens, Destroyer of Worlds," i have a good sense of humor, and a good idea of when someone is being facetious, or sarcastic, or purposefully stupid or something. however, a lot of people dont, and especially when its somebody trying to go out of their way to help the community, regardless of their sample size, you'll need to be careful with how you phrase any criticism.

otherwise youll get @#%^s like me coming in and telling you to get off your lazy *** and start expanding the sample size.


I can't +1 this enough. When looking at something tedious like this, getting 10-20 people to chip in a couple hours would improve the sample size by 10-20x satisfying pretty much anyone.
____________________________
Mishana: DRG | THF | RDM | NIN
#41 Oct 29 2010 at 5:24 AM Rating: Decent
*
73 posts
I've been reading this thread for a while - and have to agree with most of the posters.

I do commend the OP for doing his testing, however with all the flaws in his methodology - it's not conclusive by any stretch of the imagination. Too many variables such as armour wear, weapon wear and the infamous "sample size" come up to derive any sort of conclusion about dex and its effects.

If you simply didn't title this thread as you did, I would hope your tests would have recieved a better response. I hope someone with more time can take your idea - improve upon it (do it naked with a weapon at 90%+ all the time), and post a more in-depth series of tests to help the community better understand some of the attributes in the game.

Then of course, SE will release a patch the next week and totally change how everything works. :)
____________________________
FFXI: Rison of Alexander (retired)
FFXIV: Rison Srithuram of Palamecia
LS: Omnipresence [ http://omnipresencels.com ]
#42 Oct 29 2010 at 6:25 AM Rating: Decent
Scholar
31 posts
Quote:
Then of course, SE will release a patch the next week and totally change how everything works. :)

Yup! I like to tell ya all I like to explore the game's inner workings and I went crazy with crafting, orb colors and success rates of actions used in the past. I put many hours into taking notes, tracing results. Then right before I wanted to post something here on Zam, SE decides to patch how crafting works, making standard crafting the new rapid crafting, and basically turning around all the inner workings, so I had to start over again ^^

Thanks to OP for supplying a little data on DEX and hit/miss. It is good to keep it in mind. Agree we need more data to tell what DEX does (I can imagine it is used in many more calculations than just hit ratio, although hit ratio is the most obvious) And I'm also looking forward to find other things that influence hit/miss.

It gonna be tough to prove anything since it seems (speculation!) there are these trending variables at work that by example make you miss 5 times in a row on a blue mob and then the next 3 mobs you hit everything. And this trend can be observed over the course of a grind many times over and over. I am still scratching my head on that one.

Edited, Oct 29th 2010 8:30am by ShayAmora
#43 Oct 29 2010 at 7:19 AM Rating: Decent
Scholar
*
229 posts
I didn't see it posted yet but in FFXI distance had a lot to do with your ranged acc and atk. It was called the "sweet spot" for ranged attacks. Has anyone done any testing on "the sweet spot" yet and if so would that not also have to be considered in this testing? Just curious.

Edited, Oct 29th 2010 9:20am by ISystemXI
____________________________

How I mine for fish!? O.o
Solace <<<Hades>>> RIP
Vivik <<<Cerberus>>>


#44 Nov 09 2010 at 4:28 PM Rating: Default
40 posts
Props for OP getting some preliminary numbers on the DEX mechanics. It is kind of sad that some people on the forum have nothing better to do than to belittle somebody's contributions without adding their own. (BTW, yeah even a junior high school kid knows the sample size is small, thank you Captain Obvious for pointing out stuff we already know.)

If there's a parser program like for FF11, then it would be much easier to get data. I'm a pug, which would give me twice as much data, but unfortunately I can't write twice as fast. When I start leveling lancer I'll definitely start logging by hand though.

One thing I have to ask is: are WS's counted toward hits? For pugs, they give ~ 4 times more SP than a single hit (2hits/round). I wouldn't be too surprised if it affected other things like accuracy and SP gain.
____________________________
!!KEEP THE CHATLOG CLEAN CAMPAIGN!!
Don't trade with shout spammers.

Raphael Hosokawa
Hotel Moscow LS
Wutai
This forum is read only
This Forum is Read Only!
Recent Visitors: 16 All times are in CST
Anonymous Guests (16)