Paralyze Testing

[Kenji]

No, I'm saying that the test doesn't mean or tell us anything useful yet. A better test would be on several mobs of the same level, resulting in data that can be pooled.

I think the numbers tell us something very useful. If you can't tell that a 20-40% range is better than a 10-20 range, well, you need your eyes checked.

This was a good scientific study. He did tests against the same types of mobs, in the same level range. Most variables were accounted for. The largest change was the switch from Para I to Para II.

One can form a hypothesis based on these tests: That a larger sample size will produce the same results, ie, Para II being twice as effective as Para I. A meritable ability that's twice as good as a normal ability, is always a win in my books.

Now, a meritable ability that's *maybe* only slighter better than what another class can do naturally, would seem to be a poorer choice. Except for perhaps a short time period in a small, isolated part of the game.

[Max™]

Ladybugs are 45-50 as I recall, so not terribly large of a range, though their stats are unknown.

Still silly to say you can't work out a percentage from less than 100 samples.

3 procs of 6 hits is 50%, just has a larger margin of error until you prove the pattern holds.

[reccaa]

Nice initiative for the GP

I think you need a bigger of a sample pool for the results to be even remotely reliable.

From what I have read/heard is that 1st merit into para II is about accuracy, and 2nd merit is where you start noticing increased proc rate. This actually matches what I have experienced personally from Para I & 1 merit into Para II

From my experience (done many testing in sky which I do it regularly & on same set of mobs):
Paralyze II & Slow II (1 merit into each) have awesome landing accuracy comparing to normal spells. I even managed to stick them naked! And this is on normal mobs and not gods, just to clarify.

As far as proc rate is concerned... I noticed that Paralyze II may have 1 proc extra from normal paralyze.

That can be due to 2 reasons:
1- The effect is stronger on Para II opposed to Paralyze I
2- Paralyze II changes the formula & reduces the MA & difference of MND-caster to MND-target difference required in order to have an effect.

It seems to be that the general population believes that #2 is most likely the case. I believe that is more evident in Slow II where it requires less MND to reach the hard cap (which is higher for Slow II than Slow I) than it is needed to reach hard cap for slow I, giving room for RDM to equip non-MND gear.

Giving the randomness of Paralyze, I think a pool of 10 is way too small of a sample set. A sample set of 100 and on different mobs is required before even being able to come up with an "informed" conclusion. For the following reasons (of the top of my head)

1- Mob weakness/elemental strengh is not taken into consideration.

2- It is hard to test against such weakness (does it work like Ice potency merits? If so, how do you quantify an unknown/immesurable value?) 1 reason that this is important is probably a very simple observational fact that paralyze procs on certain mobs (like carbs) is much higher than other mobs where it rarely procs.

3- Elemenent of the day (we are talking about 5% extra effect, and probably in (Too Weak) Mobs, it does not make a big difference, but against level 70+ mobs & NMs, I bet it will differ.

4- All these tests are taken into consideration the +1 Effect, which I don't believe it is 1% extra (total bullshit). According to Slow II tests 1st merit has base of 25% opposed to 15% of slow I base effect. (10% extra). the hard cap being 37%. I believe you can reach hard cap without fully-merited Slow II but you'll need far more MND to reach the cap. So the +1 effect either reduces the MND difference needed to reach the cap OR increases the base slow %.

Applying that to paralyze... I seriously don't think the +1 effect is merely 1%. That is just wishful thinking & over-simplyfing things. It is easy to manipulate numbers/statistics to match what we wish to find.

At the moment I don't have enough time to do any testing, but you can expect within the next few months, a detailed set of tests to verify so many things in the world of end-game RDM ^_^

[Nekio]

I think the numbers tell us something very useful. If you can't tell that a 20-40% range is better than a 10-20 range, well, you need your eyes checked.

If you think that those numbers mean anything, you need to be educated on statistics. Do the math. If you don't do the math, take my word for it, because I have done the math. The probability that the average proc rate of Para II is more than the average proc rate of Para I is about 50%. That means that it's equally as likely that there is absolutely no difference at all based on this test.

Do I think that Para I and Para II proc at the same rate? Of course not. What I can say is that these numbers don't tell us anything yet, and need to be expanded considerably to make any statements like the OP made. The first thing you need to realize about scientific experimentation is that what you see is never accurate. You must do the math in order to make correct conclusions.

Again, I'm not criticizing the test... I think it was done well. I'm criticizing the baseless assumptions people are making using the extremely limited data.

[Seraph]

Again, I'm not criticizing the test... I think it was done well. I'm criticizing the baseless assumptions people are making using the extremely limited data.

Not to bust balls, but what is with all the ballbusting?

People are so quick to say all these tests are shit, but I think they are a fastastic step in the right direction. For a good many months, all I have ever heard around here is "Paralyze II sucks so bad Slow/Phalanx/Bio/Dia gogogogo!" At least people are starting to take the steps to get a better understanding.

I agree that saying that Para II is twice as good as Para I this early in the testing isn't a smart idea, but give it some time people. Give some suggestions on improvement instead of the 'yertestssuxcox' retorts all the time.

[Nekio]

Give some suggestions on improvement instead of the 'yertestssuxcox' retorts all the time.

Again, I'm not criticizing the test

A better test would be on several mobs of the same level, resulting in data that can be pooled.

[Seraph]

Give some suggestions on improvement instead of the 'yertestssuxcox' retorts all the time.

Again, I'm not criticizing the test

A better test would be on several mobs of the same level, resulting in data that can be pooled.

Sorry, didn't mean to direct it at you specifically, was more of a general observation.

[Kenji]

If you think that those numbers mean anything, you need to be educated on statistics. Do the math. If you don't do the math, take my word for it, because I have done the math. The probability that the average proc rate of Para II is more than the average proc rate of Para I is about 50%. That means that it's equally as likely that there is absolutely no difference at all based on this test.

Do I think that Para I and Para II proc at the same rate? Of course not. What I can say is that these numbers don't tell us anything yet, and need to be expanded considerably to make any statements like the OP made. The first thing you need to realize about scientific experimentation is that what you see is never accurate. You must do the math in order to make correct conclusions.

If you think those numbers don't mean anything, you need to be educated on statistics. If the difference between the two was a few percent, you might have a case. But since there is a such a vast difference between the percents, we can conclude that Paralyze II is vastly better at freezing up a mob.

A larger sample size would just narrow down exactly how good it is. If you feel your hpyothesis is superior, then we shall expect that you will return with your large test.

[Nekio]

Seriously, Kenji, just stop. You obviously have no idea what you're talking about. Until you show me your Student t-test calculations yielding a p-value that's statistically significant, you're just a retard on the internet spewing bullshit.

[terranova]

@ People borderline shooting the mans' test down. Please post your own >_>

[Nekio]

@ People borderline shooting the mans' test down. Please post your own >_>

Don't need a new test, just need more samples. The reason Para testing was never quantified wasn't because it was difficult, it was because you need a huge number of attack rounds from identical monsters to determine proc rates, and that's tedious.

[Kirschy]

If it helps. You can use Banish to the test the MND of your target.

Damage = 14 + dMND

If you're rdm/whm, and not using any MAB or staff, adjust your MND until you hit 14 damage /w Banish. That will be dMND = 0. Adding 1 MND should raise the damage by 1 point each time. (When dMND is lower than 0, you need about 1-3 MND to raise the banish damage by 1.)

Wouldn't a RDM need to take their MAB traits into account when doing this? Or is that included here?

You're right. I used a WHM/THF to test Banish, so MAB wasn't an issue. RDM/WHM has MAB II.
dMND = -1, 13. 13*1.24 = 16 Damage
dMND = 0, 14. 14*1.24 = 17 Damage
dMND = 1, 15. 15*1.24 = 18 Damage

So it should still be relatively easy to find the dMND=0 point on a target mob.

[Tubbers]

Seriously, Kenji, just stop. You obviously have no idea what you're talking about. Until you show me your Student t-test calculations yielding a p-value that's statistically significant, you're just a retard on the internet spewing bullshit.

You'd need to know more about his actual test to do the math out, I'm not quite sure how you are doing any math at all based on the 15 percentage values he listed without the number of procs/total number of attacks to look at.


For example assume one test shows a 15% proc rate and another shows an 80% proc rate. What is the average proc rate? Pro tip: it isn't 55% unless the base number of attacks is the same for each test. e.g. 3/20 + 4/5 = 7/25 = .28.


Belkin:
Chigoes (the low level ones in WoTG) are a good test subject. They should hit very fast, and for 0 damage through Phalanx, taking some of the time out of testing. Also, can you post the #procs/#attacks as well? Thanks for testing!

[ambutter]

Not to beat a dead horse here, but for those of you struggling to see why these numbers don't really tell us anything that is statistically significant...

If I flipped a quarter 10 times, and it came up heads 6/10 times... then I flipped a dime 10 times, and it came up heads 4/10 times... can I clearly conclude, based on my tests, that a quarter is more likely to come up heads than a dime is?

Just because results of a handful of tests appear to say something doesn't mean you can take them as gospel. In the same way that flipping 2 coins 10 times each doesn't accurately determine anything... doing a small number of Para tests doesn't accurately determine anything. Sure, the small sample looks like something we'd expect, but the OP tests are in no way definitive.

I don't think anyone's criticizing the OP's effort, just pointing out that more tests are needed to make any statistically significant statement.

[ringthree]

Seriously, Kenji, just stop. You obviously have no idea what you're talking about. Until you show me your Student t-test calculations yielding a p-value that's statistically significant, you're just a retard on the internet spewing bullshit.

I think that your assumptions may be wrong, as I made the same mistake trying to make the new damage formulas work (which by the way, for two-handers the pDIF function is incorrect on the wiki), in that it is not a normal distribution. If that is the case then a t-test may not help to discern the existence of statistical significance. In fact, it would be difficult to create a null hypothesis for the test results above without an unreliable standard deviation (probably because SE doesn't use normal distribution to determine their variables).

Remember, you are not observing occurrences and trying to find a correlation. You are looking at a mathematical formula with limits and a specific function, and a very narrow distribution.

Also remember that a t-test is just one test that can be used and isn't always necessary to derive the significance of the results. There are other much more simple tests that can confirm or deny the significance of the results.

I basically slept through all of my stats classes in college, so I could be wrong on some of this, and of course you can use many different tests that will not result in significant results, but honestly, I don't think you should (at least yet) be calling internet spewing bullshit at someone else when you sound like you are just regurgitating your 300 level stats class.

[Kirschy]

Being generous and assuming Paralyze I/II has an average duration of 60s. And continuing to be generous and assume the mob has a base delay of around 360(attacks once every 6 seconds.) That would average around 100 attacks for the first 10 samples, and around 50 attacks for the next set of 5.

There was a much larger proc% in the first set using Paralyze II compared to the second set /w Paralyze I. Up until reading this post, I had suspicions that the two tests would be much closer. While the test isn't signifigant enough to stamp an exact percentage, it is enough to show that the enhanced potency of Paralyze II is easily seen in capped conditions.

As Belkin already said, this is the start to further testing. I myself am very curious to see the results of testing /w Para vs Para II at dMND=0.

[Callisto]

As Belkin already said, this is the start to further testing. I myself am very curious to see the results of testing /w Para vs Para II at dMND=0.

This might actually be difficult to do, lol. I have 70+ MND naked, the mobs that I can actually test on without getting my head kicked in will likely have a chunk less than that, I'll have find a Spike Necklace and some other MND- crap. I'll try to get a MND value on Robber Crabs and start w/ testing on them if I get no PLD invites tonight.

Edit: I actually can't think of any other easily obtainable MND- gear for RDM off the top of my head, lol. Any suggestions?

[Sargas]

Seriously, Kenji, just stop. You obviously have no idea what you're talking about. Until you show me your Student t-test calculations yielding a p-value that's statistically significant, you're just a retard on the internet spewing bullshit.

Well, I'm no statistician, but if I naively apply the t-test outlined on this page (or the one in my TI-83) to the paralyze 1 and 2 tests given in the OP, I get a t-value of 6.077, which looking over some significance tables seems quite significant (my calculator gives a p-value of 0.0001, but I'm not sure if this is useful). Granted there are some assumptions I'm making (all the tests were at capped potency, there were no partial resists, normal distribution) and having the actual number of procs vs. attack rounds would be way more useful, but I think this test shows pretty well at least that paralyze 1 and 2 have different proc rate caps. Saying what those caps are is a bit premature at this point; I too would like to see a much larger number of tests, but this is a good start.

[Souldeity]

Actually since the paralyze II tests were done 10 times, doesn't that warrant a proportion? lol Plus like previously said, it was a very significant increase between paralyze I and paralyze II. It may not be enough to make a final decision on what the actual caps are, but its fairly accurate to state that its greater than 10%.

[Turambar]

Edit: I actually can't think of any other easily obtainable MND- gear for RDM off the top of my head, lol. Any suggestions?

Lust Dagger -1
Sloth Wand -1
Sinister Mask -2
Combat Caster's Cloak -2 (easier to get than Sinister Mask)
Blue Cotehardie -4 (both NQ and HQ are -4)
Forest Belt (-5), Jungle Sash (-4), Steppe Sash (-4) if you can wear them
Galliard Trousers -2
Custom (-1) or Marine (-3) boots, again if you can wear them
Malflood Ring -3
Flame Ring -2

Lots of AH-able stuff available. Max of -15 for Galka, -18 for Hume, -19 for Mithra and Taru, and -20 for Elvaan.

There are also a number of foods which lower MND by 1-5 points, although many of them are raw fish.