Diablo® III

Mechanics And Game Information Compendium

(Sticky)

can someone help me test paralysis?

from my observation:

-lightning hydra seems to get the maximum 8% chance of it(seeing how often it procs)
-items with +chance to stun (gloves,source,mainhand) seems to add up to paralysis max stun chance
edit: L.hydra does not benefit from item's stun chance
-the +stun on items depends on a skill's proc so it will get deducted to the +chance to stun scenario
-i think what they mean with up to 8%chance, is that every lightning skill we cast at the same time like , electrify storm armor's lightning strike, then forklightning -- if we add all the proc to that, the maximum chance will only be 8% so items tend to increase that chance

check the build i used:

http://www.youtube.com/watch?v=mEwVelmXKcM
Edited by yems#6688 on 6/24/2013 8:51 AM PDT
Reply Quote
@yems: I just tried 200 hits with basic attack (proc coef = 1) and a lightning weapon. Got 15 procs.
That's pretty much 8%, so Paralysis might not even suck that much.

What surprises me the most is that Lightning Hydra procs so much. Usually it doesn't proc anything, so that might be some kind of crack in the space-time continuum.
Reply Quote
@yems: I just tried 200 hits with basic attack (proc coef = 1) and a lightning weapon. Got 15 procs.
That's pretty much 8%, so Paralysis might not even suck that much.

What surprises me the most is that Lightning Hydra procs so much. Usually it doesn't proc anything, so that might be some kind of crack in the space-time continuum.
Hmm... that's interesting then.

If the chance for Paralysis to proc is scaled by the proc rate of a spell (which we'd expect), then it'd make sense why it's often difficult to trigger it. For example, with Living Lightning that has a proc rate of 0.2, and most targets get hit 5 times by Living Lightning over the course of 5 seconds, then the chance is ~8% per cast of Living Living Lightning. Not very flattering! :lol You'd need a good stationary DoT like Wicked Wind (or Liquefy) to trigger it.
Reply Quote
06/24/2013 07:40 AMPosted by TekkZero
P.S. For those that might be less familiar with stats, a p-value of 0.05 means we have a 5% error of being wrong in concluding there is a difference, while a p-value of 0.01 means we have a 1% error of being wrong. Thus, the p-values I'm reporting here are very significant. The chance of the stats being wrong in concluding there is a difference is VERY low (at least statistically speaking).

Well, it´s always nice to read that my tests are quite accurate :)

I´m wondering what the next step should be. Retesting each breakpoint 10 more times would be too time consuming and I´m not sure how it would help me (us) figure out CM´s mechanics. Also I can´t reliably test lower than 1.20 aps values because there´s exactly 8 inferior 1H weapons on the EU AH lol and I can´t get to 75% CC with a two hander :/

I´ll probably try to confirm "that 6 theory" with the 24 frame breakpoint on RoF next and maybe later test the 6 frame breakpoint on Disintegrate (which has a 20/aps scaling). But from the tests so far it really seems like the best you can get is 15 CM checks/sec (3.75 CM procs/sec with 0.333 proc coefficient at 75% CC) which the >2.5 aps breakpoint comes close to with an average of 3.64 procs/sec value. The lowest is probably 11 CM checks per second although one of the undesirable breakpoints showed me even lower numbers (10.8).

I really don´t feel like testing Wicked Wind at all (sorry guys), it would be too much frame counting including between casts - I´d have to cast one at a time to get comparable results with my RoF tests.

About paralysis. I´m not sure what you expect to find out. That it gets worse if you use stun gear? It should use proc coefficients based on the usual behavior of skills but it probably doesn´t, because then Hydra would never stun (but it does). From the wording "lightning damage" it implies that each lightning damage hit should have a 1 in 12.5 chance to stun.
Edited by Nubtro#2147 on 6/24/2013 11:59 AM PDT
Reply Quote
Well, it´s always nice to read that my tests are quite accurate :)

I´m wondering what the next step should be. Retesting each breakpoint 10 more times would be too time consuming and I´m not sure how it would help me (us) figure out CM´s mechanics.
I don't think we need to retest all of them. It's good enough in my mind that if two breakpoints are shown to be statistically different, especially since it's consistent with your previous data. I just wanted to be sure the spread of the data wasn't a concern!
Reply Quote
06/24/2013 11:53 AMPosted by Nubtro
I´ll probably try to confirm "that 6 theory" with the 24 frame breakpoint on RoF next and maybe later test the 6 frame breakpoint on Disintegrate (which has a 20/aps scaling). But from the tests so far it really seems like the best you can get is 15 CM checks/sec (3.75 CM procs/sec with 0.333 proc coefficient at 75% CC) which the >2.5 aps breakpoint comes close to with an average of 3.64 procs/sec value. The lowest is probably 11 CM checks per second although one of the undesirable breakpoints showed me even lower numbers (10.8).

Disintegrate should tell is if the base frame scaling matters. I think that's would be a good next step.
I'm still trying to wrap my mind around that weird up and down of CM procs. The only time I have ever seen a similar pattern in D3 was when we were testing LoH breakpoints. The number of displayed healing numbers with increasing APS were slowly rising from 7 to 12, then abruptly falling to 7 and repeating that cycle all over again with little variation.
Pretty sure that's not relevant here and the "low" breakpoints there don't match the breakpoints here. Just thought I'd mention it.

I really don´t feel like testing Wicked Wind at all (sorry guys), it would be too much frame counting including between casts - I´d have to cast one at a time to get comparable results with my RoF tests.

I'd be surprised if they were different. In all aspects we currently understand, WW and RoF are practically the same.
I think the best option is to just cast a constant number of like 5 or 10 twisters and measure the effect it has on the cooldown. That's enough to calculate the number of procs/s based on number of twisters * twister duration and effective Archon cooldown time.
As long as the number of twisters is the same in every test, it shouldn't matter if they stack or if there's a gap between last twister cast and Archon cooldown end, right?

About paralysis. I´m not sure what you expect to find out. That it gets worse if you use stun gear? It should use proc coefficients based on the usual behavior of skills but it probably doesn´t, because then Hydra would never stun (but it does). From the wording "lightning damage" it implies that each lightning damage hit should have a 1 in 12.5 chance to stun.

I'm not sure either ;) I included it in the unsolved mysteries, because I didn't do any research on it and I don't know of anyone who did.
There's just the common opinion that Paralysis barely does anything. But that might as well be a subjective impression caused by the generally low proc coef of lightning skills and the not all too high %chance of Paralysis.
But the fact that Lightning Hydra procs it (with what seems to be a proc coef of 1), at least means that there is something going on.
Edited by apo#2677 on 6/24/2013 1:24 PM PDT
Reply Quote
@yems: I just tried 200 hits with basic attack (proc coef = 1) and a lightning weapon. Got 15 procs.
That's pretty much 8%, so Paralysis might not even suck that much.

What surprises me the most is that Lightning Hydra procs so much. Usually it doesn't proc anything, so that might be some kind of crack in the space-time continuum.


i think that 8% is pretty much ok, seeing how fast lightning strikes - lightning spells tend to take care of fast moving mobs in the game, it tends to shock them a few milliseconds and the stun is the added flavor to it. Act 4 is a very good example, seeing those teleporting mages, and that annoying succubus - fork lightning and lightning hydra takes care of them
Reply Quote
Daily dose of data...

1. Tested 1.224 aps (24 frames per LoH/APoC tick) to further confirm my undesirable "no remainder after dividing the breakpoint frame length by 6" theory...24 frames also gave me the lowest number of CM procs which means that undesirable tier now includes "12", "18" and "24".

75% crit chance Cold Blood 1.224 aps 120 sec Archon cooldown
(41) 01:38 40:01 39:23 2363 = 2.0296 procs/sec
(32) 59:41 32:17 32:36 1956 = 2.6646 procs/sec
(30) 42:07 74:47 32:40 1960 = 2.6582 procs/sec
(54) 18:40 49:42 31:02 1862 = 2.8378 procs/sec
(39) 52:46 85:38 32:52 1972 = 2.6313 procs/sec
(54) 28:31 54:44 26:13 1573 = 3.5429 procs/sec
(42) 52:42 88:18 35:36 2136 = 2.3511 procs/sec
(52) 31:04 62:40 31:36 1896 = 2.7700 procs/sec
(51) 06:16 39:32 33:16 1996 = 2.5817 procs/sec
(395) ------ avg 32:48 17714 (1968.2 avg)
64800 -395 -17714 = 46691 / 17714 = 2.6359 procs/sec avg

2. Then I tried to test Disintegrate whether it uses the same weird mechanics. I wasn´t able to channel it long enough for Archon cooldown with two APoC sources and to use a third I´d need a 5-6% crit reflection which I don´t have (last time I borrowed it from a friend and didn´t feel like pestering him again with it - my fault for feeling bad and giving it back despite him telling me to keep it). So I just channeled one Diamond Skin at a time.

75% crit chance Disintegrate 1.54 aps (20/1.54 = 12.987 = 12 frames)
06:46 12:41 05:55 355 9 procs
13:10 20:08 06:58 418 8 procs
20:51 26:36 05:45 345 9 procs
27:05 33:03 05:58 358 9 procs
38:09 45:05 06:56 416 8 procs
45:19 53:44 08:25 505 7 procs
54:09 63:07 08:58 538 6 procs

75% crit chance Disintegrate 1.668 aps (20/1.668 = 11.99 = 11 frames)
02:45 08:40 05:55 355 9 procs
35:40 40:36 04:56 296 10 procs
51:20 57:06 05:46 346 9 procs
29:45 35:17 05:32 332 10 procs
46:18 52:33 06:15 375 9 procs
03:12 08:28 06:16 376 10 procs
28:11 32:19 04:08 248 11 procs

The results aren´t very clear because the 15 CD is quite small...but to me it seems that Disintegrate is similar to RoF in the "6" area. Hopefully someone else will be able to do some tests and confirm it. I kind of need a break :)
Reply Quote
Thanks again Nubtro.

So I guess the big question now is "why?".
What could cause the proc rates to drop by so much, on very few specific breakpoints?
Maybe again some kind of rounding issue?
Hmmm...
Reply Quote
06/23/2013 09:28 AMPosted by apo
I meant this is big news. People are chasing higher breakpoints in part because they expect a significant increase in CM and Shocking Aspect procs. From what you've found out so far, the only benefit of a higher breakpoint might be fast Energy Twister cast rate. And of course more LoH/APoC. but after all, most want to optimize for maximum DPS, not for sustain and AP management.


Higher APS also shortens the amount of time between Chain Reaction pulses, which should also affect freeze strength I would think. At 1 APS it's about 1 second between pulses. Not sure of the exact scaling, though.
Reply Quote
Shouldn't make a difference. It's always 3 hits per cast, multiple casts stack. Shorter delays might be better in the first few seconds because you get your hits right away and don't have to wait for multiple casts to stack up. But in the first few seconds CC resistance hasn't fully built up, so that is probably also not relevant.
Of course, damage-wise it's better because it's the same amount of damage in a shorter period of time. But that's only relevant for short fights, over time it evens out as well.
Reply Quote
Even if you had, say, 5 CRs stacked up, them pulsing faster should still make a difference beyond the beginning of a fight. You're getting your CM procs from CR in a shorter amount of time, thus the cooldown on FN is back up faster, right?

Like, if you had 1 APS and it takes 1 second between pulses. Let's say you have 5 CRs stacked up, it would take 3 seconds for all of them to pulse, for 15 total pulses in 3 seconds. Compared to, say, if you had 3 APS and all 3 pulses of each CR cast went off in 1 second, for a total of 15 pulses in 1 second. I think I'm explaining that correctly.

I know that you're never going to have that many CRs perfectly stacked on each other, but it should still make a difference. I would compare it to something like Electrocute I suppose. Higher APS means you're hitting with it more often, thus CM procs happen faster.
Reply Quote
Well, it's 3 hits per cast. And cast rate is only affected by your ability to reduce the cooldown.
No matter how fast or slow those hits happen after you press the button, in total it's always 3.
So yes, you might get those hits later, but you are not missing out on anything.
Reply Quote
First - great thread - I wonder how many hours of testing it took to collect all those knowledge. I recently realized that i spend more time reading, testing, switching then actually playing the game. I spend more gold on !@#$ty test gear than on upgrades :)

Anyway - speaking about CR i have a question too. I play cm wiz with very high cc, meteor reduction cost gear, 30 apoc and skills which allows me to spam Moltens or Liquefy without bigger problem.

So lets say i'am in the middle of monsters gropup, i only spam Prism and Moltens on those monsters and my AP is always top. So theoretically i should reach max Moltens spam speed only limited by casting time. And then i start spamming CR too. Meteors are spamming even faster.....Why? Tested in single player with 40-50 fps. Well even more - i see it everyday for a long time now. It would be nice if someone could clarify it.

edit: Not only CR but other EB runes or Shards too.
Edited by Oxan#2794 on 6/27/2013 1:55 PM PDT
Reply Quote
06/27/2013 01:53 PMPosted by Oxan
First - great thread - I wonder how many hours of testing it took to collect all those knowledge. I recently realized that i spend more time reading, testing, switching then actually playing the game. I spend more gold on !@#$ty test gear than on upgrades :)

For me, testing is just a fraction of my total ingame time. It's definitely several hours, maybe a few days, but that's nothing compared to the hundreds of hours I spent actually playing.
Even if you include forum time, spent on reading, writing and thinking about stuff, I think it wouldn't come close.

Anyway - speaking about CR i have a question too. I play cm wiz with very high cc, meteor reduction cost gear, 30 apoc and skills which allows me to spam Moltens or Liquefy without bigger problem.

So lets say i'am in the middle of monsters gropup, i only spam Prism and Moltens on those monsters and my AP is always top. So theoretically i should reach max Moltens spam speed only limited by casting time. And then i start spamming CR too. Meteors are spamming even faster.....Why? Tested in single player with 40-50 fps. Well even more - i see it everyday for a long time now. It would be nice if someone could clarify it.

edit: Not only CR but other EB runes or Shards too.

I have read something like that before, but in the end I think people agreed it might have been a lag/FPS drop issue that makes it seem like the meteors are casting faster when in fact the animations are just trying to "catch up" with what already happened on server side.

There's no mechanic or effect we know of that might cause the behaviour you describe.
And I choose not to believe it, until I see a video with a steady 60 FPS framerate and measurable differences in meteor cast rate ;)
Reply Quote
So without tests for WW if we try to directly apply the results to WW, we'd want to do something like assume 1 WW cast is the same as 6s of RoF channel, right? That means WW has 360 damage tics per cast. Over a long fight, like a Ghom test, we'd expect the crits to average out quite well so with 50% CC we'd expect about 180 crits per WW. That means we'd expect about 180/8 = 22.5 SA procs per WW.

In reality we only see about 5-6 SA procs per WW, which is about 2.5x what we'd see using the LoH mechanics but about 1/4 what we expect from the 60 tics per second behavior.

So with all that RoF testing, does any of it explain what's going on with SA? And with that same line of thinking does it also explain the about 2xCM procs we see using WW? If I didn't miss anything with the data collection then it seems RoF doesn't directly translate into WW behavior since it overpredicts how many SA and CM procs we see.

Also, let me see if I understand how everything is supposedly working. Every LoH tic, which occur about 2*APS times per second, but uses breakpoints, the game checks a crit. If the crit succeeds, then every frame until the next LoH tic results in a crit and you get returned AP = APoC*Spell_Coef. Then, and this part I'm not sure about, if a crit occurred, a check is made against the spell coefficient to determine if the crits proc CM, and if it passes, all crits proc CM until the next check. Presumably all procs succeed if the first one succeeds, but there might be separate proc rolls per proc, so just because CM procs doesn't mean SA has to proc or the other way around, maybe.

If that's true then CM and SA procs should occur at an average rate of 360*Spell_coef*CC per WW and basically be about equal in procs. Granted some CM procs are likely to be wasted if everything is already on CD so that could explain why it seems like CM procs about 2x while SA procs more like 2.5x the LoH tic rate would indicate. The RoF tic rate indicates it should be more like 10x though.

I hope that all makes sense, and some of it I'm not sure about the actual math behind it since I haven't been able to follow the recent results as closely as I'd like.
Reply Quote
@Loroese: Actually the data by Nubtro predicts the SA procs!!!! ^_^

He showed for Ray of Frost that you have ~2CM procs per second at 50%cc and 3CM procs per second at 75%cc. If take into the proc coefficient difference (0.125/0.333 = 37.5% less procs for WW), then we guess the expected procs per twister is simply: (6 seconds)x(37.5%)x( [2 or 3] CM+SA procs per second) = 4.50 or 6.75 CM+SA procs per twister. Now Nubtro's numbers are far from perfect, since the variability is pretty large with 50% and 75%cc, but it's at least in the ballpark of what actually expect.

This is what's got me so fascinated by Nubtro's results, since they actually predict the "2x" multiplier that we've always thought was there. We expect 2.25 procs per twister if CM procs behaved like LoH (ie 50%cc and 36 ticks (for 2.73aps)): (Proc Coefficient)x(Crit Chance)x(Number of Ticks) = 0.125*0.5*36 = 2.25. But that's too low. Most of the data we have suggests the presence of a "2x" multiplier and Nubtro's research confirms that. And so it looks like CM procs might be independent of tick rate, and instead behave very differently than LoH.
Edited by TekkZero#1963 on 6/27/2013 4:00 PM PDT
Reply Quote
Is there an explanation for the CM procs in there somewhere that I missed? Or was it just a test to see how many CM procs you get in xx seconds to refresh Archon?
Reply Quote
There's no mechanic or effect we know of that might cause the behaviour you describe.
And I choose not to believe it, until I see a video with a steady 60 FPS framerate and measurable differences in meteor cast rate ;)


The video would be here already if I would know a magic trick that let me fraps this game with 60 fps because right now i can't even get 30fps while spamming and recording....
Reply Quote
06/27/2013 04:02 PMPosted by Loroese
Is there an explanation for the CM procs in there somewhere that I missed? Or was it just a test to see how many CM procs you get in xx seconds to refresh Archon?
It's the latter. To me it's VERY interesting because it's experimental proof of what we already expected to see. But the observation is novel too, in that the CM procs don't seem to scale with attack speed. That's quite tantalizing. And while we're still in the dark right now on the actual mechanism, it's clearly not the same breakpoint mechanics as LoH.
Reply Quote

Please report any Code of Conduct violations, including:

Threats of violence. We take these seriously and will alert the proper authorities.

Posts containing personal information about other players. This includes physical addresses, e-mail addresses, phone numbers, and inappropriate photos and/or videos.

Harassing or discriminatory language. This will not be tolerated.

Forums Code of Conduct

Report Post # written by

Reason
Explain (256 characters max)
Submit Cancel

Reported!

[Close]