Paralyze Testing

[reccaa]

this turned into an argument on how to run a test & a defensive one that is; instead of discussing paralyze II proc rate /sigh

[Callisto]

Anyways...headed out now as RDM/PLD, no alt b/c of the stupid POL issue, Kirschy were the crabs that you were seeing consistent MND results from @ the Tree or Kuftal(or elsewhere)?

[Snapples]

so basically what the general idea is: if you have an awesome mnd setup, dont merit slow2 or para2 beyond lvl 2 or 3 because you can land it at its capped value. if you DONT have a big mnd setup, merit the spell higher which gives you higher magic acc and potency (effectively higher mnd according to the formula). someone please correct me if ive made a drastic mistake here, and yes im making a generalized statement.

There was a clear difference in caps using Slow II /w 2 Merit and Slow II /w 3 merits in my testing. Testing /w a 4th and 5th merit hasn't been done yet, but there will probably be a noticable increase. These kinds of tests /w Paralyze probably won't provide precise data on how much increase is obtained /w further merits.

asll the testing ive seen say slow2 caps at 34% no matter what merit you have, it just takes less mnd to reach that cap

[Callisto]

Well I got MND values established for some Steelshells at least, level 73 has 76 MND, level 74 has 77 MND...some asses came in to level their NPCs on the crabs now so I guess getting actual para info is going to have to wait...

Add: Whatever level gives 130 XP has 80, and 75 has 77 MND as well.

Edit: Yeah...can't fucking test like this, their mobs' para procs are fucking my program up, guess I'll have to wait til the weekend.

[Plight]

From my own testing done a while back on Clippers in Qufim Island, with a sample size of 1000 attack rounds per Paralyze type (Para/ Para II 1 merit/Para II 2 merits) unfortunately I can't find my data for 3 merits at the moment. The first attack round after each cast and the last attack round before each 'worn off' message was ignored to account for lag. Should it be relevant, all tests were performed with 134MND 286 Enfeebling

Paralyze: 258/1000
Paralyze II (1 Merit): 314/1000
Paralyze II (2 Merits): 319/1000

Edit: found half of my 3 merit data - 159/500

[ringthree]

Talked to my friend that is a Ph.D. candidate in Comm at USC. His research is mostly in statistics and statistical analysis. Anyway, this is what he had to say about the data from the first post.

[09:53] ringthreetheicp: Can I ask you a question and you can help me see what test would be best to look for significance?
[09:54] ringthreetheicp: If you are busy don't worry about it.
[09:54] Paul: in social science research, it is generally defined in such a way as to say that there is 95% likelihood that differences in the independent variable are responsible for differences in the dependent variable
[09:54] Paul: sure describe the situation
[09:54] ringthreetheicp: Ok, this is game related.
[09:55] ringthreetheicp: A player casts a spell on a monster that has a chance to prevent that monster from acting.
[10:02] ringthreetheicp: The first spell is cast ten times, with the spell procing 39%, 31%, 18%, 33%, 34%, 36%, 40%, 23%, 28% and 37% of the time for each test respectively.
[10:04] ringthreetheicp: A second different spell with a similar effect is cast 5 times, with the spell procing 17%, 18%, 17%, 19%, 19% of the time for each test respectively.
[10:04] ringthreetheicp: Is there a way to tell if the first and second spell have a statistical significance in their chance to proc?
[10:05] Paul: can you collect more data?
[10:07] Paul: with a t-test, one of the assumptions is that you have equal sized groups
[10:07] Paul: but it is fairly robust to that assumption
[10:07] Paul: let me run the numbers
[10:09] Paul: ok-- so i conducted an independent sample t-test
[10:10] Paul: for an indepedent sample t-test, you want the two groups to have equal variance, so first I conducted Levene's test for equality of variances
[10:11] Paul: with an F of 6.479, p=.024, we can say that they are close enough
[10:11] Paul: now, in the first group, the average was 31.9%
[10:11] Paul: in the second group, the average was 18%
[10:12] Paul: with 13 degrees of freedom, the t value is 4.281, significant at p < .001
[10:12] Paul: which means, the odds that the difference occured due to chance is less than .1%
[10:13] Paul: i.e., highly statistically significant
[10:14] Paul: the t-test is useful for comparing means, but there are other differences between the two groups
[10:14] Paul: the first has a standard deviation of over 7, while the latter has standard devation of 1
[10:15] Paul: this means that there is a much higher degree of variability in the first group -- although it appears that only a few times out of a hundred would that first group be as low as the second group
[10:16] Paul: perhaps less than that, if, for example, there is a programmed lower limit to its effect
[10:16] ringthreetheicp: Ok, cool. Is there a different test we could use that would be better than a t-test for this?
[10:16] Paul: nevertheless, the second item is highly consistent, so it averages 18%, and then 2/3 of the time it will be between 17 and 19
[10:16] Paul: no, the t-test is just fine for this
[10:16] ringthreetheicp: Ok, cool.
[10:17] ringthreetheicp: Pure spectualation on my part, but I believe the lower spell has a lower cap, and the higher spell has a higher cap but with a higher level of variation.
[10:18] Paul: yeah thats what the data indicate
[10:18] Paul: but if there was more data, you might be able to make stronger inferences
[10:19] ringthreetheicp: What would you want for those tests?
[10:24] ringthreetheicp: Busy?
[10:24] ringthreetheicp: Just need equal results or how many more tests would you want?
[10:25] Paul: well i would say that if you found more results that generally confirmed what you already have, you arent going to get a different result
[10:25] Paul: and you don't need a 'stronger' result, as the result is quite strong
[10:26] Paul: but if you find something which contradicts the assumption, then that plus all the rest of the data in the intererim can help determine if its an abberation or if the assumption is in fact mistaken

Hope that helps. Definitely need more testing, though.

[Callisto]

Kind of a side note if anyone is interested, when I was doing the rounds on crabs last night I noticed that Para II very consistently lasted 19-21 attack rounds(this was on about 10 straight level 73-74 crabs). If the crabs have 360 Delay(I think Kirschy mentioned they do) then 20 swing rounds is 120 seconds, I think it'd be safe to say normal fully unresisted Para II duration is 120 seconds.

[Kirschy]

I had used the crabs in Kuftal and only used four crabs total, because there was no need to kill them off during the tests. Either they all have 66 MND, or I was just lucky. Crabs are faster than bees, around 240 delay.

[Callisto]

I had used the crabs in Kuftal and only used four crabs total, because there was no need to kill them off during the tests. Either they all have 66 MND, or I was just lucky. Crabs are faster than bees, around 240 delay.

The Robbers in the tree were annoying, there was a 2~ MND swing in both directions it seemed like, maybe they vary more in level there. So 240 delay would take...80~ seconds for 20~ swings, correct? Kind of odd duration time, but I'll try to keep track of that in addition to the swings/para procs.

For what it's worth the 3-4 crabs I was able to do with a 0 dMND value, Para II level 2 proc'd 3 times in the 20~ swings for all of them. Anwyays I'll post a real data set hopefully some time tomorrow assuming half the friggin' server doesn't show up to level their NPCs on Steelshells again tonight.

[Belkin]

MGS4 hinders my testing. Might be a couple days now

[Callisto]

MGS4 hinders my testing. Might be a couple days now

Screw you and your owning a PS3. G/f won't let me get one b/c she's adamant that we have to buy a couch first.

[LeifofKujata]

I'm coming in kinda late so bare with me. Has it been determined yet that dMND actually affects para proc% or are the tests still trying to find the average base percentage of dMND = 0? I personally have always felt MND had no real effect on para other then Macc and that proc rate was just a dice roll based on the base percent per duration. Example: Para1 rolls a d100 have to roll between 1-20 to proc and Para2 rolls a d100 and have to roll between 1-40 to proc(just example numbers). In no why have I tested it formally. This is what I assumed to be happening regardless of dMND, but not elemental weakness/resistance.

[Ambir]

A couple of things to consider when tesing (I did a lot of slow II testing back in the day and got quite a bit of good advice...)

For testing the cap:
Grab a mind setup and go test on a Level -1 Ronafore bunny. Don't leave any doubt that you're at the MNDdiff Cap.

For acquiring a large number of data points:
Use the same mob. Pick a mob and use the same physical mob to remove the chance of MND variations from mob to mob. There's no reason you can't use the one mob over and over unless you think it could build a resistance (and theres no evidence that happens with Para). Also, you can use data from more than one casting to increase the number of data points (Assumption: partial resists affect duration but not potency. Definitely true for Slow, probably certain for Para)

Testing the difference between Para and Para II:
Again, use the same mob. Get sets of 100 (say) data points on the one mob for both Para and Para II and compare. The more points you have, the less uncertaintly there is in the measurement.

Measuring the cap:
The cap won't be the highest number of Procs you get in 100 tests, it will be the average. Para has an equal chance of Procing on every normal attack. For a cap of 40% Proc rate, there is a 1% chance of Para procing 5/5 attacks, but that does not give a cap of 100%. Obviously, the larger the number of tests in a set, the closer to the true average you will see for that set. Overall though, the average is the same for 20 sets of 5 data points or 1 set of 100 data points.

Duration:
Measure duration. Slow/Slow II has a fixed duration (176 sec) which only changes if there is a resist. A half resist is 88sec and there is no slow that will stay on between 88 and 176 sec (unless something removes it). Para/Para II are variable duration enfeebles, and it's still uncertain whether MND, Skill, or indeed anything other than randomness affects the duration outside resists. More data about duration vs potency may help to clear that up.

[Nateypoo]

I had an argument before in a thread, that was very similar to this. I was trying to figure out whether the puppetmaster's attuner attachment affected a ranged weaponskill's damage. I several dozen data points for two tests. In the first test, all the data was lower than the second test, so it looked something like this:

Test 1: 500, 500, 500, 500, 500, etc
Test 2: 700, 700, 700, 700, 700, etc

To me it was pretty obvious it was affecting the damage, but people tried to argue over it like in this thread. Going back to the coin toss scenario Kenji corrected...

If you assume the outcome should be the same for both, then it would be equally likely that either one would be higher than the other. The chance of having every pair of data points higher for one test would be (0.5)^X, where X is the number of pairs. I can only assume it's a MUCH lower probability, since all the data points in test 1 are lower than all the data points in test 2 (not just the pairs). I'm not sure the specifics of how a t-test works, but you don't need a p.h.d. to figure out the answer to a yes/no question. The OP's test isn't exactly the same as mine, since there's more variance and one of this data points dropped pretty low, but if you apply the process I described, you'll get about a 1 / 2^5 = 3% probability of this data being lucky. So it makes sense that ringthree's t-test showed even less than that (0.1%).

The question for this thread is: "is the paralyze II cap higher?" The answer is yes. How much higher? We have no clue, but we can at least say that it is higher. I think that's what this data tells us. I'm not sure why anyone would flaunt their statistician credentials, when a simple reality check implies one is stronger than the other.



With that said, I'm not sure how helpful this data is yet. It needs to be tested on a mob of value. For example, what if it was tested on VT or IT mobs and the difference in proc rate was no longer noticeable? I've had paralyze II merited since the week the tier II merits came out, and I had never been particularly impressed with it. In fact, the only time it was phenomenal was against TW mobs. I think anyone who has used the spell can attest to the incredible proc rate on weak mobs and barely noticeable proc rate on strong mobs. The test shows that it has nearly 15% more proc rate, yet often I do not see even a 15% proc rate against strong mobs.

So obviously when the op claims:

If anybody can suggest a better testing method, I'm all for it. But I hope this gives some RDMs some peace of mind knowing that Paralyze II is about 15-20% better than Paralyze I.

it needs to come with the disclaimer: "when dMND is capped" and probably even "against this specific mob".

[Ryushii]

Nose game for testing ice spikes! (finger on nose)

P.S. Anything a semi-retired RDM could do to help with testing?

[Kenji]

With that said, I'm not sure how helpful this data is yet. It needs to be tested on a mob of value. For example, what if it was tested on VT or IT mobs and the difference in proc rate was no longer noticeable? I've had paralyze II merited since the week the tier II merits came out, and I had never been particularly impressed with it. In fact, the only time it was phenomenal was against TW mobs. I think anyone who has used the spell can attest to the incredible proc rate on weak mobs and barely noticeable proc rate on strong mobs. The test shows that it has nearly 15% more proc rate, yet often I do not see even a 15% proc rate against strong mobs.

So obviously when the op claims:

If anybody can suggest a better testing method, I'm all for it. But I hope this gives some RDMs some peace of mind knowing that Paralyze II is about 15-20% better than Paralyze I.

it needs to come with the disclaimer: "when dMND is capped" and probably even "against this specific mob".

Thank you for your input. As people have mentioned in this thread, if we work under the assumption that Para II merits work like Slow II merits, then more merits simply decrease the MND you need to reach the cap. So not only test against a mob of value, but test with a full MND setup and full Para II merits.

Is it possible to freeze up most mobs of value 30-40% of the time? If so, full Para II merits may be quite valuable indeed.

[Ryushii]

Thank you for your input. As people have mentioned in this thread, if we work under the assumption that Para II merits work like Slow II merits, then more merits simply decrease the MND you need to reach the cap. So not only test against a mob of value, but test with a full MND setup and full Para II merits.

I believe over in the Slow thread they're now saying that merits also raise the cap, at least the first 3 merits.

[Kenji]

Thank you for your input. As people have mentioned in this thread, if we work under the assumption that Para II merits work like Slow II merits, then more merits simply decrease the MND you need to reach the cap. So not only test against a mob of value, but test with a full MND setup and full Para II merits.

I believe over in the Slow thread they're now saying that merits also raise the cap, at least the first 3 merits.

Yeah, I just noticed that. But it would be curious if the last 2 merits didn't function the same as the first three. The only other merit category that works like this, is Overwhelm, and SE specifically singled that one out for commentary. You'd think they might inform us of it's not the case with a few other merit groups.

Thus, Para II is even more intriguing. It might be hard to hit a raised cap, but could you imagine freezing up really tough mobs 45% of the time? Might get easier, with new gear that has more +MND on it.

[Auspice]

Thank you for your input. As people have mentioned in this thread, if we work under the assumption that Para II merits work like Slow II merits, then more merits simply decrease the MND you need to reach the cap. So not only test against a mob of value, but test with a full MND setup and full Para II merits.

I believe over in the Slow thread they're now saying that merits also raise the cap, at least the first 3 merits.

Yeah, I just noticed that. But it would be curious if the last 2 merits didn't function the same as the first three. The only other merit category that works like this, is Overwhelm, and SE specifically singled that one out for commentary. You'd think they might inform us of it's not the case with a few other merit groups.

Thus, Para II is even more intriguing. It might be hard to hit a raised cap, but could you imagine freezing up really tough mobs 45% of the time? Might get easier, with new gear that has more +MND on it.

I'm holding off on capping out my Para and Slow II with the new update until I hear some results, although Slow II is going to rock if the cap was also raised. My full MND setup already puts me at +68MND and 310 skill and I've got some pretty key upgrades that I'm missing(like Salvage gear >_> and Celestial Earrings) so it'll be interesting if potency is raised and the MND differences are lowered with those last two points. I almost hope the results show otherwise so I can put points in some of the other spells instead of going 5/5 on Slow and Para but I'll do what is going to make me the best.

[Yekyaa]

Yeah, necrobump, this is the only Paralyze Testing thread I could locate with the search function that caught my interest.
I'm wondering if anyone was able to determine a cap/percentage regarding Paralyze II merits? If anyone has a more relevant thread that I can bookmark instead of this one, please let me know.