Extensive Paralyze I Testing

[Auspice]

Assuming mobs don't have MND values of like 120+, then this number is ridiculously easy to hit. As an Elvaan RDM/WHM (no merits), I had 75 base. With a good amount of HQ gear, I had something around +60 MND. Talk about overkill...

Wait, isn't it your mind minus the mob's mind? You're not going to hit that 120 cap unless the mob has very low mind.

So, if a DRK throws Absorb-MND for you, the paralysis should last longer?

DRKs never do this for me so I have to rely on when I go /DRK to events. *IF* I can land it, the extra MND is always nice. :D

[Byrthnoth]

He concluded that potency is based at least partially on total MND, not dMND. That's where the first comment came from. However, test that showed that also had a 30% resist rate and no one has ever definitively dissociated resist rate and potency.

Random stuff:

dMND=0 (EM crab) - Data is fine, proc rate of 13.7% okay
sqrt[.13731*(1-.13731)/(1056)]*1.96 = 95% confidence interval [11.65%; 15.8%]

dMND=0 (+15 Skill) - Your numbers are off on this one. Your raw data has 82 Procs and 534 hits. That still doesn't make your numbers work though. (Proc rate of 15.34%)
95% confidence interval of [12.3%; 18.4%]

dMND=0 (High level crab) - Your numbers are fine for this one, though the 30% resist rate is disturbing. Proc rate of 16.6%
95% confidence interval of [14.0%; 19.2%]


Good job collecting the data, and I'd agree that Paralyze was procing more on the higher level crabs if it wasn't for their resist rate. It would have been better to do both tests with enough Skill gear to floor resist rates on the higher level crabs.

[McGoonagle]

DRKs never do this for me so I have to rely on when I go /DRK to events. *IF* I can land it, the extra MND is always nice. :D

As an enthusiastic casting DRK, Absorb-MND always struck me as a bit of an oddity. If it can help WHMs, I'll start throwing it.

[Masamune]

Resistances were based off simple criteria over the course of my test. I noted that the max duration was in the region of 120 seconds, unresisted. I also noted that 20 seconds was the lowest recorded time for a long time, until the odd 10 second value turned up. I theorise that unresisted Paralyzes will do a number between 20-120 and 1/2 resist between 10-60. Knowing this, any time between 10 and 20 are classified as half resist, and any above 60 are full hit, with 20-60 being unknown and potentially either. It's worth knowing that regardless of resistance or not, the potency remains the same.

I don't understand this paragraph, which represents the very basis of all the tests that followed...

I'm ok with "Max duration=120secs" and "Min duration=10secs" but not with how you "classify" other durations, especially this :

20-60 being unknown and potentially either

That can't work in terms of programmation, ALL durations should be classified somewhere.
For example, if 120secs is the max of the "non-resisted" range of durations,
then an half resist range would max at 120/2=60secs,
then a 1/4 resist range would max at 120/4=30secs,
then a 1/8 resist range would max at 120/4=15secs,
then would have to prove if duration can dip below 120/16=7.5 => 8secs ?

That makes you 5 categories where to sort all your procs, for each dMND test. I don't see any categorization or i missed something ? looks to me you just added all procs together, regardless half resist, full hit or w/e ?

2nd point that disturbs me kinda is the level of confidence of such tests involving determination of an unknown randomized % value, should always be specified like Byrthnoth said earlier. That's big difference with testing a static randomized value (like pDIFs). Why: because when testing a static value you know when to stop sampling, while when testing a %, only the error on the current obtained % gives you an idea when to be confident enough to stop. That error is displayed nowhere in your tests :s
So 1st question a decent reader will wonder is "are all those numbers enough for the conclusions following to be believed as fact ?"

But don't get me wrong, i'm just pointing at what i think being flaws (i might be wrong) to help you make your tests even stronger.

[Masamune]

Also, i got a problem when i tried to confirm the MND values for your tested mobs:

Race Crab : PLD/pld, MND ranked D
PLD job: MND ranked C
MND=fMNDRace(lvl, RaceMNDrank) + fMNDMainJob(lvl, JobMNDrank) + fMNDSubJob(lvl, SubJobMNDrank)

lvl66 Robber Crab
MND=floor(3+(66-1)*0.35) + floor(4+(66-1)*0.4) + floor((4+(66-1)*0.4)/2)=25+30+15
MND=70

lvl76 Steelshell
MND=floor(3+(76-1)*0.35) + floor(4+(76-1)*0.4) + floor((4+(76-1)*0.4)/2)=29+34+17
MND=80

...are closest mnd values i could get of yours. If i tried RaceMNDrank=C -> MND=75 and 85 for resp. lvl66 and 76.

I dunno for Banish2 dmg, but i read Kirschy mentioning Banish1 dmg=14+dMND, so might be worth verifying that ?

And did any of your tested crabs used that MdB Shell TP move ? that might skew some of your procs potencies :s

[CDF]

The assertion is that the rate of paralyze procs is independent of the consequences of resistance, which determines only whether average duration is halved or the spell resists outright. I would argue that's what's really going on. Therefore, you can pool the potency data to obtain an average proc rate.

Although arguing from analogy has its limits, it seems similar to Drain and Aspir, for which it's obvious (at least to me) that resistance manifests in reduction of average potency (average based on a distribution of possible values) by half or worse.

[Byrthnoth]

Yeah, I was just worried that potency and magical accuracy of Paralyze might (for whatever reason) be randomized within a certain range. In that case you would get a measured Paralyze potency higher than it should be, because the less potent/accurate paralyzes would be resisted more frequently.

Like, imagine if every melee hit had a slightly different chance of landing directly proportional to the pDIF value. The lower pDIF was, the lower the odds would be of landing it (and, in this case, the lower the duration would be if it landed). So higher pDIF values would be overrepresented.

I have tried to make histograms of:
Individual Cast's Average / [Condition's population average]
so I could look for a normal distribution, but I don't think the sample size is high enough.

[Siniroth]

Wait, isn't it your mind minus the mob's mind? You're not going to hit that 120 cap unless the mob has very low mind.

estimate your paralyze potency using the following (rough) equation:

Potency = (2*(MND+dMND))/10 %

e.g. You have 70 MND, the monster has 65MND:

Potency = (2*(70+5))/10 % = 15%

so your mind plus (your mind minus mob's mind)

[Spagnutty]

If you have the chance, could you test if Sub-Zero Smash, Sub-Zero Smash, Sub-Zero Smash and Sub-Zero Smash have static paralyze potencies?

Understand if you've got better stuff to do. Great work regardless.

[Masamune]

The assertion is that the rate of paralyze procs is independent of the consequences of resistance, which determines only whether average duration is halved or the spell resists outright. I would argue that's what's really going on. Therefore, you can pool the potency data to obtain an average proc rate.

Although arguing from analogy has its limits, it seems similar to Drain and Aspir, for which it's obvious (at least to me) that resistance manifests in reduction of average potency (average based on a distribution of possible values) by half or worse.

Yes me too, just read fast cast Kegsay results, looks like a static potency so far, but since all his data is mixed i might have misread. That's why i explained this "categorisation" needed to prove the asssertion of independency between resistrates and potencies.
The direct consequence of skipping this step is what Byrthnoth explained in his post just above: Higher measured proc%.

Personnally, i try to avoid using too much analogies but only to get "ideas" and stop there nothing more, especially dont make asumptions based on those analogies. I know that's so convenient to just test a few wisely chosen points then generalize for the rest from that... but no.

[Kayatani]

If the results show anything, it's that potency isn't static. You can't possibly catagorise 1/2, 1/4, 1/8, etc. resists because an unresisted paralyze (level 0 mobs) can last anywhere from around 30~120 seconds. As the OP said, the only resists you can be sure of are those that last <30 seconds.

[Masamune]

erm.... im wondering what is the real definition of "resist" then ? cuze what you wrote don't make sense at all if i were the programmer.

For me, i always thought that resist = spell feature have been reduced from its max observed features.

For example, someone observed his max drain draining 320 max under normal conditions, so if if he see his Drain draining less = "resisted" ?
I'll continue with nukes calculations around always consider as facts the resist rates of 1/2 1/4 1/8 and 1/16, which have been proven in many situations.
Now in the case of a spell with a random duration, affected by resists, then best thing we can do is combine the 2 known statements above into "categories" i explained in earlier post.

So what's the problem with that? wrong definition of "resist" ?


Direct consequence of this is can't conclude anything about procs reported ie static or not proc'ing.

[Kayatani]

In the case of enfeebles, a resist is commonly accepted to be a reduction in duration in the spell, as is the case with Silence, Slow and Sleep spells.

While you can't establish that resists do not affect potency of paralyze using the available data, it's a pretty logical assumption. Suggesting that potency can be resisted based on observations of dark and elemental magic is not.

[Masamune]

In the case of enfeebles, a resist is commonly accepted to be a reduction in duration in the spell, as is the case with Silence, Slow and Sleep spells.

yes that's on what i based my categorization proposal (and Kegsay too but only 2 categories).

While you can't establish that resists do not affect potency of paralyze using the available data, it's a pretty logical assumption.

Why not ? that's the whole test objective Kegsay clearly explained : proving if the potency are subject to resists. or did i misread something ? and no its not "pretty logical" and need testing like ANY asumption.

Suggesting that potency can be resisted based on observations of dark and elemental magic is not.

Ok there i admit those examples shouldnot be taken as examples to base asumptions for enfeebs. np my bad.

[Byrthnoth]

It's commonly accepted that it's a change in duration, but it has never been proven that it doesn't also change potency. Most Paralyze tests are done at floored resist rate. We could group the data here by duration and look for a correlation between duration and potency, but I'm not sure we have enough data for it and it'd probably be simpler to just repeat with higher MAcc.

I've been plotting the durations and am getting slightly lumping looking results. However, those trends go away if I pool with other samples, which either indicates that the other samples have a different maximum duration and it's fractional or that the lumpiness was due to low sample size.

[CDF]

If you really don't want the effect of skill to be confounded with resistance, you will just have to ignore all the ambiguous data (anything with duration above 60 seconds) and perhaps use Poisson regression or something similar to model the proc rate while controlling for duration.