Formula for comparing +haste to +accuracy

[Kirschy]

Obviously not much interest that 2 acc = 1 hit rate appears to break down at the 50% mark and at lower levels because very few people are interested in idle "model building." So this kind of fake concern about the macc testing is kind of annoying especially when decisions about meleeing are based on much flimsier or nonexistent evidence.

I only used the breakdown of melee accuracy at 50% as one of my many tests that shows the 2acc->1% hit rate model doesn't hold up for all cases. I don't think it's a fake concern to warn people not to take this generality as fact. People are quick to believe what they read and pass it on as fact.

Minimizing the macc data as a "step in the right direction" is absurd when it's a reasonably solid foundation for people finally to discard unfounded notions such as "1 skill = 0.9 macc above 200," confirm (or refute) previously held ideas about accuracy and INT, that elemental staves even provide a multiplicative % bonus (if it did, then why don't changes in elemental skill in the presence of a staff reflect that?), and recognize this strange 50% accuracy (landing rate) threshold as a confounding factor in previous tests.

It's a step in the right direction as it puts forth a lot of theories to be tested. Unfortunately the 1 skill = .9 mg acc above 200 wasn't discarded or disproved. If you took the time read the information, you would know that much more data would need to be collected to ascertian if 1 skill = 1 mg acc or .9 mg acc. You would also know that the data taken on staff wasn't used on varying amounts of elemental skill... Which is why Staff bonus is unknown to be a static +Mg Acc or +% Land Rate.

True, the limited range of elemental magic skill leaves unanswered the question of what may or may not happen at higher levels. The range of dINT above the identified critical point is also limited (10-30). But it is absurd to argue that conclusions from a single mob cannot be generalized to other mobs. This is not comparing the efficacy of [insert drug here] in adults to that in children.

It is absurd to generalize the data to all mobs blindly. I'm only asking people to understand where the data came from, and that there is lot left to be tested still. Using it as a guideline is perfectly acceptable, and I do it myself.

Also, the distinction between static macc vs. direction addition to landing rate is practically meaningless unless there is some reason for the distinction I am not considering atm. You would have a marginally better case for arguing that the magic accuracy vs. landing rate was not adequately explored (if you think 1 macc from a piece of equipment is somehow not +1% landing rate, for example) or that the results are not necessarily applicable to other magic types.

There is no distinction between static mg acc and land rate if the function is linear. But that hasn't been proven yet. It could be a piecewise linear function or something such as a regression curve.

The point of my post was just to inform people. It's not uncommon to see a general guideline followed blindly. I only want people to understand that these are guidlines based on specific situations. And there are more tests to be done before we can be sure how the models work in a more general manner.

[Yugl]

I believe CDF is the guy who made (Translated?) the data in the first place. I'm pretty sure he has looked over all the data, so we can stop with the "If you read..." comments.

[Kirschy]

Re-reading I did come off sounding rude, and I'm sorry for that. (And if CDF is the one who translated the info, I apologize even further.)

The data, and the blog explaining the data both report that the Staff experiments were done with no varying skill. (Even the Blogger posed the question of it was additive or multiplicative.)

I did miss that there was a second set of testing done with varying skill and HQ Staff only, but unfortunately there was not enough data to determine if it was a direct +Mg Acc or +Land Rate%. And with only 5 points of elemental skill used which included an inflection point, the linear vs non-linear relationship is unable to be determined.

As with 1 Elemental Skill = 1 Mg Acc, the blogger stated:

Analysis Of Parameter
Estimates                     Standard  Wald 95% Confidence
Parameter  Estimate     Error        Limits
Skill           0.0096    0.0006   0.0084      0.0108

This is saying within 95% confidance, 1 elemental skill is .84-.108 mg acc.

Again I just want to say the information is increadibly useful. But I still feel it's important that people understand where the data came from, and that guideline implicated may or maynot apply to all other situations.

[Tsuki]

Yes, CDF=Robonosto

[CDF]

First, I don't mean to come across as unduly aggressive in my comments. Second, you make good points made from the standpoint of "model building," which unfortunately most people don't care about. (I mean... just look at this thread.) Third, it was not my intention to be deceptive and act as though I were a third party observer.

Again, the behavior of melee accuracy in the specific cases you describe seems to be of relatively little concern compared to where 2 acc -> 1% hit rate actually holds. This is not an unconditional fact in the same way that classical mechanics cannot unconditionally describe physical "systems." But so what? Describing a general model to cover all situations is one thing, but the chief preoccupation here and elsewhere is with optimality at level 75, where 2 acc -> 1% hit rate applies.

Drawing analogies between melee mechanics and magic mechanics (either direction) doesn't seem like something that you would approve, but the 1 magic skill -> 0.9 macc assumption is a typical, even classic example of that. Yet the alleged inability to "prove" statistically that 1 skill -> 1 macc and not 0.9 macc is more of a problem? (I made an argument based on statistical power and sample size. I could just as easily have done an equivalence test if I really wanted to be even more anal about it.)

As I mentioned in passing, it is difficult to say whether the elemental staff accuracy bonus is additive or multiplicative based on direct inspection of the estimates of staff accuracy. Indirectly though, one would expect the change in skill to be higher in the presence of a staff than without. If you believed that 1 skill corresponds to 1% "hit rate" or 0.5% "hit rate" and that a HQ staff really conferred a 15% accuracy bonus akin to he damage bonus (you have to start somewhere), there is no compelling evidence, given the sample sizes involved and the data itself, that a 1-point increase in magic corresponds either to a 1.15% increase or 0.575% increase in "hit rate."

As for the alleged absurdity of generalizing to all mobs, what is a good reason not to? Unlike other things in life, we should be pretty confident that there is a functional relationship between Input X and Output Y in a MMORPG. I am straining to think of any hypothetical differences conditional on mob type that that don't involve (1) outright immunity or (2) capped or floored hit rate.

Finally, don't flip the relationship between regression and "reality." The point of regression in this case is to model a causal relationship. It is curious to think that the probabilistic mechanisms in the game must adhere to a contrived regression model, the point of which is to approximate the causal relationship that is limited by the scope of the data.

[FailureMidgard]

Ahlah is his own league. Don't even try to compare him to someone else unless it's senseiFF11 (Closest comparison by far).

Really? I think I almost prefer Ahlah's shit to pchan. If anything, I wouldn't be surprised if they were brothers, or the same person. Guess we'll just have to disagree on this one.

[Kirschy]

I apologize again for sounding rude in my initial response. And yea, it seems that I'm alone in feeling on the matter~

I think the classical mechanics anology is a bad one. I feel that 2acc->1 hit rate% only applied to one DEX/AGI relationship, and at a specific accuracy/evasion relationship. I think if you carried it to mobs such as an IT thf Mamool you would need a different amount of accuracy to increase your hit rate by 1%. I don't have any proof or evidence though. Thats why I'm asking people be open minded, and understand there's still a need for more testing.

I've probably been too vauge and not articulating myself well. But I don't believe the ele skill +1 = .9 mg acc. The main reason for mentioning mg acc, was that I felt it paralelled the lack of information on melee hit rate%. (At least when compared to fleshed out models like calculating Magic Damage or Enmity values.)

I'm suggesting it's possible that there's a magic evasion trait that is compared to magic accuracy which plays a role in calculating land rate%. If the relationship is a ratio, than a monster with a higher mg evasion than an elemental would see different relationship between ele skill->land rate%.
It's also possible that calculation involves a more intricate/complex relationship between the variables and lends itself to a non-linear graph.

The only thing I think is absurd is to take the test and say "Ok this is how it works for everything." (Which you haven't done at all.) I just feel it needs a little more testing to check other possibilities.

Also, I used regression poorly before. I was thinking of the curve from a log regression.
http://www.codecogs.com/users/378/reglogistic-378.png

[FailureMidgard]

I apologize again for sounding rude in my initial response. And yea, it seems that I'm alone in feeling on the matter~

I think the classical mechanics anology is a bad one. I feel that 2acc->1 hit rate% only applied to one DEX/AGI relationship, and at a specific accuracy/evasion relationship. I think if you carried it to mobs such as an IT thf Mamool you would need a different amount of accuracy to increase your hit rate by 1%. I don't have any proof or evidence though. Thats why I'm asking people be open minded, and understand there's still a need for more testing.

I've probably been too vauge and not articulating myself well. But I don't believe the ele skill +1 = .9 mg acc. The main reason for mentioning mg acc, was that I felt it paralelled the lack of information on melee hit rate%. (At least when compared to fleshed out models like calculating Magic Damage or Enmity values.)

I'm suggesting it's possible that there's a magic evasion trait that is compared to magic accuracy which plays a role in calculating land rate%. If the relationship is a ratio, than a monster with a higher mg evasion than an elemental would see different relationship between ele skill->land rate%.
It's also possible that calculation involves a more intricate/complex relationship between the variables and lends itself to a non-linear graph.

The only thing I think is absurd is to take the test and say "Ok this is how it works for everything." (Which you haven't done at all.) I just feel it needs a little more testing to check other possibilities.

Also, I used regression poorly before. I was thinking of the curve from a log regression.
http://www.codecogs.com/users/378/reglogistic-378.png

I saw you mention the dex:agi in another post elsewhere, and after that, I went out and did some really quick tests that didn't conclusively prove anything, but left me wondering where you got the idea to begin with. I tried a few different dex values (with +/- acc to counter +/- dex), and never once saw differing accuracy values to change the evasion /check.

Did you do some test to ever once find that varying dex/agi has any effect whatsoever on accuracy other than Xagi=1eva, and Ydex=1acc?

Personally, I'm not about to go test whether 2 accuracy is 1% hitrate or not, but I don't mind testing things that are a little more concrete, so just wondering where you're even getting this from...

[Kirschy]

With low level characters fighting IT+ mobs, and using Abs-AGI + DEX Etudes there was a noticable impact on hit rate. Also when messing with blindpots, DEX seems to make a more noticable impact.

I didn't collect data for either, and the sample size wasn't very large. I was just playing around testing some ideas for myself. The dex/agi relationship is just a theory I have. It could be wrong.

[Thekender]

With low level characters fighting IT+ mobs, and using Abs-AGI + DEX Etudes there was a noticable impact on hit rate. Also when messing with blindpots, DEX seems to make a more noticable impact.

I didn't collect data for either, and the sample size wasn't very large. I was just playing around testing some ideas for myself. The dex/agi relationship is just a theory I have. It could be wrong.

It's something to look into. And definitely a noticeable difference depending on mob and level from experience.

[pchan]

I'm suggesting it's possible that there's a magic evasion trait that is compared to magic accuracy which plays a role in calculating land rate%. If the relationship is a ratio, than a monster with a higher mg evasion than an elemental would see different relationship between ele skill->land rate%.
It's also possible that calculation involves a more intricate/complex relationship between the variables and lends itself to a non-linear graph.

I've done a lot, and I mean, a lot of tests, on qiqirn poulterer which don't have special resistance to stone I think, and the same pattern was observed : at "low rate" and dint=0 the HQ staff or the Alkalurops (20 Macc 10 INT) would give 15% in landing rate while at "high rate" the same weapons would give something close to 30%. The only part where I don't aggree is the scission with under 50% and above 50% as with "medium rates" I got 20 ish %. Both magic accuracy and HQ staves give of an increase in % of something that I don't understand. Also HQ staff seems to ignore the dint while magic accuracy seems to be tight to dint in some way.

For example : if you would be at dint=0 20 magic accuracy w/o any other stats would add 10% hitrate (1) while alkalurops itself is more than 25% (2) and HQ staff are ~30% (3) ; If HQ staffs would give at straight + m.acc it would then be around 60 m.acc according to (1) while it would be hardly more than 20 m.acc and 10 int according to (2). Test here :

Spoiler: show

dint=0 skill=281 mac=2
no staff : 44.5%
+10 m.acc : 50.3%
+18 m.acc : 55.7%
+alkalurops : 70.9%
+ HQ staff : 72.4%

Sample is 1k for each line and the % is the proportion of non-resisted spells which is the only relevant data.

Got plenty of other tests with various skill if interested, but I didn't really understand a pattern.

[JigglyJam]

Because I'm all about the giving back:

http://venomweb.150m.com/ffxi/haste_vs_acc.xls

Thanks for this.

[CDF]

On the role of magic evasion in affect landing rate:

This suggestion of a ratio of magic accuracy to magic evasion is a bit inchoate. Ideally if current knowledge is inadequate to explain some phenomena that people are really interested in or make a solid prediction within the game, the evidence that players gather (anecdotal or otherwise) would help direct further investigation. It is kind of unproductive just to throw out some speculative ideas about how something might work when the current base of knowledge is "good enough" for many.

At any rate, are players being misled by this "newer" (hello, 2007) knowledge about magic accuracy? At least it can be seen that sitting below 50% overall accuracy is even less desirable than just being resisted more frequently (equipment bonuses less efficient). Elemental staves almost unequivocally reign supreme still and so on.

On attributing inconsistencies to dINT 0:

Keeping in mind the supposed 50% "landing rate" threshold, suppose your landing rate with no staff was really 45%. Under the current "model":

10 magic accuracy => 50% (+5%)
18 magic accuracy => 58% (+5% +8%)
Alkalurops => 70% (+5% +20%)
HQ Staff => 70% (+5% + 20%)

[pchan]

On the role of magic evasion in affect landing rate:

This suggestion of a ratio of magic accuracy to magic evasion is a bit inchoate. Ideally if current knowledge is inadequate to explain some phenomena that people are really interested in or make a solid prediction within the game, the evidence that players gather (anecdotal or otherwise) would help direct further investigation. It is kind of unproductive just to throw out some speculative ideas about how something might work when the current base of knowledge is "good enough" for many.

At any rate, are players being misled by this "newer" (hello, 2007) knowledge about magic accuracy? At least it can be seen that sitting below 50% overall accuracy is even less desirable than just being resisted more frequently (equipment bonuses less efficient). Elemental staves almost unequivocally reign supreme still and so on.

On attributing inconsistencies to dINT 0:

Keeping in mind the supposed 50% "landing rate" threshold, suppose your landing rate with no staff was really 45%. Under the current "model":

10 magic accuracy => 50% (+5%)
18 magic accuracy => 58% (+5% +8%)
Alkalurops => 70% (+5% +20%)
HQ Staff => 70% (+5% + 20%)

Yeah it corresponds more or less, I'm going to check my other tests. But what would the 50% threshold mean then ? Is is the rate without gear, without stats, without magic accuracy or skill gear. It's not clear.

[Starr]

How did Ahlah get banned and not Pchan anyway?

[CDF]

Very loosely speaking, landing rate could follow

rate = .50 + X([magic accuracy] - [magic evasion])

where

if ([magic accuracy] - [magic evasion]) < 0, then X = 0.5
if ([magic accuracy] - [magic evasion]) > 0, then X = 1.0

(Edit: I switched the X values.)

Here, "magic accuracy" is defined as any positive contribution (equipment bonuses, etc.) and "magic evasion" is defined as any negative contribution (equipment penalties, level penalty, elemental resistance, etc.)

Even if this is really how it is, if you're sitting below 50% rate on something (this you will never know without gathering a sample) there are probably more pressing issues than whether you can clear 50% or not. Likewise, poor melee hit rate and whether 2 macc -> 1% hit rate.

P.S. This is even less relevant, but the information about partial resists can be used if you want to use a more efficient estimator (basically, smaller standard error for a fixed sample size) than just the number of non-resists over the total number of casts. You can PM me if you want the details.

[pchan]

Very loosely speaking, landing rate could follow

rate = .50 + X([magic accuracy] - [magic evasion])

where

if ([magic accuracy] - [magic evasion]) < 0, then X = 1.0
if ([magic accuracy] - [magic evasion]) > 0, then X = 0.5

Here, "magic accuracy" is defined as any positive contribution (equipment bonuses, etc.) and "magic evasion" is defined as any negative contribution (equipment penalties, level penalty, elemental resistance, etc.)

Even if this is really how it is, if you're sitting below 50% rate on something (this you will never know without gathering a sample) there are probably more pressing issues than whether you can clear 50% or not. Likewise, poor melee hit rate and whether 2 macc -> 1% hit rate.

Ok, I checked out on all my tests, some were done with low rate some high rate (the varying rate essentially resulted of magic skill change). That's 43 different tests with 1K+ sample size in each with total magic accuracy ranging from 240 to 330 and that seems to suit perfectly. This model works almost perfectly with a magic evasion of ~292 (probably including level correction) for those qiqirns. So basically this also confirms that HQ staves are static +30 mac.

I used 1int=1 mac (I didn't explore didn't >20 anyway), 1 skill=1mac.

[Saithorx]

Because I'm all about the giving back:

http://venomweb.150m.com/ffxi/haste_vs_acc.xls

Nice, you made this a lot more user friendly then the one I threw together when the question was asked. I would probably need to write a manual to try and explain what I was doing here.
http://www.geocities.com/saithorx/ff...-accuracy2.xls (I know, lolGeocities)