Register

(RPPM) Trinket Proc Rate issues (Apr/May '13)

Face-rippin fun.
Revered
User avatar
Posts: 309
Joined: Tue Oct 23, 2012 4:19 am

(RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Alpheus » Thu May 02, 2013 6:21 pm

For the past 10 days I've noticed quite heavy fluctuations with my two trinkets, Renataki and Bad Juju (haven't seen Rune drop yet). I'm not quite sure when this started but it's been destroying progression attempts for me for the better part of the past 2 raiding weeks. I'm posting here in hopes to raise some attention to this issue and find out whether this is me doing something wrong or if there's anyone out there with the same problem. I'm from the EU region and sadly our CM/GM's are as close to a vegetable as you can get.

Setup

Renataki [541, HC+TF]
Bad Juju [522]
Meta Gem, mastery > haste > crit reforge

My armory: http://eu.battle.net/wow/en/character/d ... s/advanced
Our logs: http://www.worldoflogs.com/guilds/191352/

Will be focusing mainly on our Iron Qon wipes:
http://www.worldoflogs.com/reports/rt-0 ... boss=68078
http://www.worldoflogs.com/reports/elvx ... boss=68078

The Issue

My Juju procs are fluctuating between 15% and 45% per attempt (averaging 20%). On the lower end of the proc spectrum it rarely procs on the pull.
My Renataki procs (Blades) are fluctuating between 15% and 35% (averaging 20%). On the lower end of the spectrum it rarely procs on the pull.

Per fight table looks something like this:

Try 1 (9:08) 20% Juju, 20% Blades
Try 2 (5:41) 17.6% Juju, 17.6% Blades
Try 3 (4:21) 30% Juju, 17% Blades
Try 4 (8:37) 19.3% Juju, 15.4% Blades
Try 1* (5:13) 26.5% Juju, 19.1% Blades
Try 2* (1:37) Juju did not proc, 20.4% Blades
Try 3* (4:56) 28% Juju, 26.9% Blades
Try 4* (3:37) 18.3% Juju, 18.3% Blades
Try 5* (8:57) 25.5%, 18.3%

* - second log mentioned above

Oddly enough, Dancing Steel and Capacitance are both at their expected levels: ~40% and ~60%, respectively. There was about 3 minutes of wipe recovery between tries. I started logging out and back in after some mass rezzes/combat rezzes/sacrifices hoping it would solve the issue but I haven't been able to find any significant correlation yet.

My SimulationCraft profile lists average proc rates of slightly above 25% for both trinkets over 10000 iterations while the logs above average about 19-20% for both trinkets. Now before you jump that it's a too small sample to be statistically reliable, keep in mind that there are encounters in there in which either one or neither trinkets proced within the first minute of the fight despite having 3+ mins of downtime between pulls, which is the major DPS impact that made me look into this in the first place.
Now I won't compare the overall dps for the point of this discussion because they are wipes after all and have a different uptime on heroism/berserk/activity etc. Nonetheless, the dps burst is quite different between tries 1, 4 and 5*.


For comparison, there's three patchwerk-like kills in the first log (Ji-Kun, Durumu (boss-only) and Primordius (no mutations))

I scored top 10 ranks on Ji-Kun and Durumu and my procs were as follows:

Ji-kun: 12.9% Juju, 54.9% Blades
Durumu: 12% Juju, 41.8% Blades
Primordius: 35.2% Juju, 29.3% Blades

Am I the only one that finds these numbers terribly disturbing or am I just overanalyzing this?

TL;DR

Of the 4 RPPM procs: Capacitance, Dancing Steel, Bad Juju and Renataki the former two are where they should be on EVERY try. Why aren't the trinkets?
Image

Exalted
User avatar
Posts: 1386
Joined: Mon May 24, 2010 10:21 pm

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Tinderhoof » Thu May 02, 2013 7:29 pm

I am sorry but I do have to jump on the "too small sample to be statistically reliable" train. Looking at your logs I see 2 fights which don't have both trinkets proc'ing in the first minute of combat (Try 3 and Try 2*). I think try 2* is an exception due to how fast your raid wiped (more on that below) The rest do have both trinkets procing near the start of the fight.

The problem with your assertion is you are attempting to compare SimC with real world execution. The big difference is you need boss uptime to get proc's (Try 2* likely had very little uptime). Iron Qon is not a great fight for melee uptime (well everyone up time really). Between Tornado's, Fire/Ice/Lighting on the ground, dog changes, and the stupid shield there are plenty of stretches which you do not have contact on the boss. Fights like Durumu have much higher uptime. Comparing the two is apples and oranges. Same thing with comparing uptime to SimC which expects 100% uptime.

Also all 4 of your RPPM items/enchants have different proc rates. From my personal experience Renitaki does tend to proc a little later in my opening then Rune does (first few hits). While I have JuJu I haven't had a chance to use it yet so not sure there. Have you considered taking some of your rather massive Mastery pool and pushing it into some more Haste? This would likely help you see more procs.

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Thu May 02, 2013 7:31 pm

If you change zones or log in, this resets the "bad luck streak" timer. So your logging out was hurting you. Releasing after a wipe also hurts you. At your haste level, renataki is a guaranteed proc after 5.3 minutes and juju after 6.4.

In general though, you're going to see much larger fluctuations in juju/renataki uptime than in capacitance/dancing steel uptime just because of sample size. The latter two effects proc so many times per fight that the average proc interval rapidly approaches the expectation. That's not the case with the trinkets.

Revered
User avatar
Posts: 309
Joined: Tue Oct 23, 2012 4:19 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Alpheus » Thu May 02, 2013 7:38 pm

Tinderhoof wrote:I am sorry but I do have to jump on the "too small sample to be statistically reliable" train. Looking at your logs I see 2 fights which don't have both trinkets proc'ing in the first minute of combat (Try 3 and Try 2*). I think try 2* is an exception due to how fast your raid wiped (more on that below) The rest do have both trinkets procing near the start of the fight.

The problem with your assertion is you are attempting to compare SimC with real world execution. The big difference is you need boss uptime to get proc's (Try 2* likely had very little uptime). Iron Qon is not a great fight for melee uptime (well everyone up time really). Between Tornado's, Fire/Ice/Lighting on the ground, dog changes, and the stupid shield there are plenty of stretches which you do not have contact on the boss. Fights like Durumu have much higher uptime. Comparing the two is apples and oranges. Same thing with comparing uptime to SimC which expects 100% uptime.

Also all 4 of your RPPM items/enchants have different proc rates. From my personal experience Renitaki does tend to proc a little later in my opening then Rune does (first few hits). While I have JuJu I haven't had a chance to use it yet so not sure there. Have you considered taking some of your rather massive Mastery pool and pushing it into some more Haste? This would likely help you see more procs.


I tried looking it that way as well, but consider this: if the main factor on Iron Qon was activity/uptime on boss, why are Dancing Steel and Capacitance at their expected average values consistently? Shouldn't they be lower as well to offset for the lower activity (the same things that proc dancing steel proc the trinkets as well)?

To my knowledge the lack of actual procing events (ie. not making melee hits) should not in any way be noticable on a 10 minute fight on for an RPPM trinket (like it isn't for Capacitance and Dancing Steel), especially so since it's meant to have a no-proc-failsafe and 10% increased proc chance in ToT.

This still doesn't explain why I have a very low uptime on Bad Juju on Durumu (which I had for 3 consecutive weeks).

Edit: There's also the alleged increased proc chance on pulls.
Image

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Thu May 02, 2013 7:48 pm

DPS downtime only matters if you're off the target for >10 seconds, which is only during wind storm. So I don't think that's a major concern. The difference between juju/renataki and capacitance/dancing steel is just that dancing steel procs 4 times as frequently and capacitance procs 100 times as frequently as the trinkets. Hence, the uptime of these effects approaches the theoretical expectation much more rapidly.

Revered
User avatar
Posts: 309
Joined: Tue Oct 23, 2012 4:19 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Alpheus » Thu May 02, 2013 7:53 pm

Thanks for the replies so far guys, the logging in and out does make sense as you mentioned in hindsight, but I guess it doesn't really matter then because I'll be forced to do either one or the other (release/get ressed) after each wipe during progression so there's no way around that.

In way I'm not really looking for any kind of rationalization of why this "would" be happening, I'm challenging the notion whether it is actually acceptable that it DOES behave this way. The original PPM+ICD system got to the point where people could (and were thus "forced to") time ICD's to maximize certain burn phases (especially so with snapshotting dot classes). From my understanding the idea behind RPPM was to randomize when these events happen but over the following weeks we noticed that it not only affected the distribution of the proc but also the overall quantity of it. There were some fixes, and some more fixes. And then another and another. The situation now is that most RPPM trinkets have an internal CD to prevent wasted double-procs, a no-proc-failsafe which should nullify bad streaks, 10% increased proc chance on strength and agi trinkets in ToT and true RPPM taking account only haste rating (and not attack speed).

Why is it then that despite all of these measures the trinket procs still fluctuate between 10 and 50% on completely average boss fights (despite or in spite of activity)? Personally, I don't find it acceptable that my major dps procs start going haywire as soon as we're doing progression content (ie. quick succession of wipes) and I'd rather have the option of a classic PPM+ICD trinket or an on-use one if that meant that my dps stops acting like a monkey on a bunjee rope.
Last edited by Alpheus on Thu May 02, 2013 7:55 pm, edited 1 time in total.
Image

Exalted
User avatar
Posts: 1386
Joined: Mon May 24, 2010 10:21 pm

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Tinderhoof » Thu May 02, 2013 7:53 pm

Alpheus wrote:
I tried looking it that way as well, but consider this: if the main factor on Iron Qon was activity/uptime on boss, why are Dancing Steel and Capacitance at their expected average values consistently? Shouldn't they be lower as well to offset for the lower activity (the same things that proc dancing steel proc the trinkets as well)?

No. As both Stenhaldi and I have pointed out those RPPM rates are MUCH higher then the trinkets. They ALWAYS proc on the pull (Dancing Steel procs from Healing Touch as well).
Alpheus wrote:To my knowledge the lack of actual procing events (ie. not making melee hits) should not in any way be noticable on a 10 minute fight on for an RPPM trinket (like it isn't for Capacitance and Dancing Steel), especially so since it's meant to have a no-proc-failsafe and 10% increased proc chance in ToT.

It's not really a 10 minute fight though. When you spend a full minute not attacking the boss cause he is in the air, and you have to avoid tornado's you aren't doing melee attacks, and you aren't leaving combat. Now you do have to understand that you are going to have encounters where your trinket procs are shitty. That's why they had to put protection on them to begin with. It is going to happen. Now if you really want to get something closer (really stressing closer) to SimC, do 10, 10minute pulls on training dummies and then look at the proc rates there.

Alpheus wrote:This still doesn't explain why I have a very low uptime on Bad Juju on Durumu (which I had for 3 consecutive weeks).

Edit: There's also the alleged increased proc chance on pulls.

Like I said, you are going to have shitty pulls, and you are going to have amazing pulls. I mean really, 50% uptime on Blades pretty darn good. Take your wins where you can.

Exalted
User avatar
Posts: 1386
Joined: Mon May 24, 2010 10:21 pm

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Tinderhoof » Thu May 02, 2013 7:56 pm

Stenhaldi wrote:DPS downtime only matters if you're off the target for >10 seconds, which is only during wind storm.

I only brought up down time as he is still doing progression on the fight. A farm kill I totally agree with you. Still learning the fight tends to produce less uptime while learning to stay out of the bad. With more experience comes better uptime.

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Thu May 02, 2013 8:05 pm

Alpheus wrote:Thanks for the replies so far guys, the logging in and out does make sense as you mentioned in hindsight, but I guess it doesn't really matter then because I'll be forced to do either one or the other (release/get ressed) after each wipe during progression so there's no way around that.

Getting resurrected doesn't reset the timer, so if you can get a resurrection then that's optimal.

Revered
User avatar
Posts: 309
Joined: Tue Oct 23, 2012 4:19 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Alpheus » Thu May 02, 2013 8:08 pm

Tinderhoof wrote:It's not really a 10 minute fight though. When you spend a full minute not attacking the boss cause he is in the air, and you have to avoid tornado's you aren't doing melee attacks, and you aren't leaving combat. Now you do have to understand that you are going to have encounters where your trinket procs are shitty. That's why they had to put protection on them to begin with. It is going to happen. Now if you really want to get something closer (really stressing closer) to SimC, do 10, 10minute pulls on training dummies and then look at the proc rates there.


I might as well try that over the weekend, I'm not convinced that the difference in RPPM should matter (the dancing steel vs. renataki discussion). Now I'm in no way too math-savvy but I do not believe that higher RPPM does in fact attribute a statistically closer result in practice. From my understanding the only thing that should affect statistical correctness is the number of "attempts" or events to trigger the proc, not the actual chance of triggering it (in this case it's melee hits, melee abilities and pretty much any brainfart that does or does not involve other targets). ie. 25% proc on 10 applications opposed to 10000 would make quite a dent, but 25% vs. 40% on 10000 events should still net a high probability to reach 2500 and 4000 procced events, respectively. The high chance that there's a consistent difference of more than an order of magnitude strikes me as very suspicious.
Image

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Thu May 02, 2013 10:17 pm

Alpheus wrote:Now I'm in no way too math-savvy but I do not believe that higher RPPM does in fact attribute a statistically closer result in practice.

Well then. I don't feel like writing out a full proof, but if you can believe that Poisson statistics apply, then it's pretty straightforward to see that the higher proc rate will see less relative deviation from the expected proc rate. Consider a 400-second fight, so let's say you expect 4 juju procs and 16 dancing steel procs. Since the number of procs is Poisson-distributed, the variance (mean square deviation) is equal to the mean and so the standard deviation of the juju proc count is sqrt(4) = 2 while that of the dancing steel proc count is sqrt(16) = 4. Immediately you see that while the dancing steel proc count has a larger absolute deviation, its deviation is smaller in a relative sense. That is, the standard deviation of the juju proc count from its expected value is 1/2 the expected value, while the standard deviation of the dancing steel proc count from its expected value is only 1/4 the expected value. In uptime terms, we could say we expect juju uptime 20%±10% and dancing steel uptime 48%±12%. In this illustration, you can see that it's completely normal for juju uptime to swing by a factor of 3 between attempts (betweem 10% and 30%), while dancing steel uptime scarcely varies by a factor of 5/3 (from 36% to 60%).

Just some quick caveats. I used easy sample numbers in this calculation, rather than actual values from the game data, so don't take this as a prediction for in-game uptimes. Also I didn't account for proc overlap; that level of care isn't necessary for this illustration. Finally the Poisson distribution isn't symmetrical about the mean, so saying something like "20%±10%" isn't quite accurate -- but it's close.

Actually this is most of a proof -- the only missing link is to show that Poisson statistics do apply. It's intuitively pretty clear that they should; the only complication is that the Poisson distribution is usually derived from a binomial distribution where the probability is the same for every event, whereas in the case of RPPM effects the probability varies for different events.

Exalted
Posts: 769
Joined: Tue Oct 23, 2012 7:15 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby raffy » Fri May 03, 2013 2:03 am

I agree with Stenhaldi's analysis: poisson distribution.

Additionally, from my Catus sims: (for 535 ilvl gear)
https://dl.dropboxusercontent.com/u/298 ... eroic2.png
(you can see the 3x up/down swings)
And the (slightly quantized) Uptime distributions:
https://dl.dropboxusercontent.com/u/298 ... -Histo.png
Image

Honored
Posts: 166
Joined: Mon Aug 01, 2011 6:20 pm

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Sorcerer » Fri May 03, 2013 8:34 am

I will ask from different angle.

How come when I last checked, assasination rogues have much higher dmg % overall from legendary meta, when last time I checked, their proc rate is lower than for feral.
Is this still the influence of dual wielding with RPPM? Wasnt it supposed to not have the influence or is it some other mechanic that I potentially miss atm.

I think the same still applies to trinkets but I guess I would need to deep into logs first.

I don't want to qq but current trinket / RPPM design is so stupid and game changing, that pardon my point of view, any WoL races lost their day. All the fun is gone. I think I have yet to rank on my level this tier, first becouse of lack or RoRo, than bug and than how stupid the RPPM behavior is for us.
Give back the consistency that rogues currently have. Having 4 raiding rogues, there is almost no %!@#!@#!@ dps difference between them, while I might deep 20k above or 20k below based on how RNG OP/Shit is RoRo based on luck.

Some might not agree, but comparing to previous tier, the consistency was really awsome thing that I miss in the day. Being able to maintain rank(dps) spots each week without stupid influx of 1 item. Skill, right...

Devolopers forgot that they wanted to make feral rotation harder, too bad they just gave it recently a luck factor.

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Fri May 03, 2013 11:10 am

I'm fairly certain rogues are seeing essentially the same trinket uptimes as ferals. As for the capacitive meta, Blizzard has given out a lot of different numbers and now that the threads are deleted, I'm not sure which ones are accurate. Just from looking at some logs it seems pretty clear that assassination rogues have a similar proc rate to ferals, however, but their proc does more damage.

Posts: 38
Joined: Thu Jan 03, 2013 11:01 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby feralminded » Fri May 03, 2013 12:09 pm

Sorcerer wrote:I think I have yet to rank on my level this tier, first becouse of lack or RoRo,


Trust me don't put all of your eggs in that basket. This trinket has made the rotation even more fiddly and even more proc-dependent and I am actually at the point I may just put down my druid. This week I was plague by poorly timed RoRo procs and I watched my DPS completely tank compared to pre-RoRo levels. I used to consistently rank (nothing top 25, just not there with gear or skill) but RoRo is killing me now. I hate the fiddliness and bad-luck-capacity of this style of play. I appreciate and laud skill which is why I still hang my hat on my top 25 parses in ICC but luck having such a massive factor upon my own personal success just turns me off completely. It is even more disheartening that even when played perfectly, outside of a couple of fights where we can optimize for AoE spam, we are still miles behind everyone else. I am hopeful the 10% buff will do a lot but it still won't change how blindly luck dependent we are post RoRo. I mean don't get me wrong I'm sure I will have a pull in the near future where lady luck swings the other way for me and the RoRos line up with my TF's and pots and JuJu and I get off those staggeringly nasty Rips but even still I'd take consistency that favors skill over chaos that favors luck.

I hate to be the QQ guy, in fact I REALLY hate being the QQ guy, but it's awfully hard to love my druid like I did back in ICC when I felt skill was all that mattered. Sure in plenty of ways the rotation got simplified and we got a nice amount of Quality of Life changes ... but it also got a helluva lot more fiddly with the procs of doom.
Nosis, GM of Neolutum, always recruiting

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Fri May 03, 2013 5:33 pm

I love Rune of Re-Origination. Shrug. Its proc rate is actually pretty high (about twice as high) compared to Renataki's and Juju, leading to a lot less randomness.

Honored
Posts: 58
Joined: Tue Dec 25, 2012 6:56 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby teddabear » Fri May 03, 2013 9:49 pm

I think the aggravating part is the difficulty of getting off a Rip during a RoR proc.

Honored
User avatar
Posts: 67
Joined: Sat Dec 08, 2012 3:35 pm

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Vami » Sat May 04, 2013 4:44 am

That feeling when you get Rip on the target 0.034 seconds after Re-Origination falls off.

Honored
Posts: 58
Joined: Tue Dec 25, 2012 6:56 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby teddabear » Sat May 04, 2013 7:10 pm

I don't think that is an issue that can be solved by better play. If you are at 0 or 1 combo points and your haste and crit go to 0 your extremely unlikely to get a Rip. if Tiger's Fury is down which happens quite a lot.

Honored
Posts: 166
Joined: Mon Aug 01, 2011 6:20 pm

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Sorcerer » Sat May 04, 2013 7:35 pm

Well at my gear level the rip issue is only existant whenever I should renew roar around the proc. Karma is a bitch and sometimes it just syncs almost always ;)

Revered
User avatar
Posts: 358
Joined: Wed Jun 29, 2011 4:49 am

Re: (RPPM) Trinket Proc Rate issues (Apr/May '13)

Postby Stenhaldi » Sat May 04, 2013 8:52 pm

teddabear wrote:I don't think that is an issue that can be solved by better play. If you are at 0 or 1 combo points and your haste and crit go to 0 your extremely unlikely to get a Rip. if Tiger's Fury is down which happens quite a lot.

You really just want, when possible, to avoid having simultaneously no combo points and no energy. If you have just one or the other you're generally fine.

For instance, consider this "nightmare" scenario: you cast a finisher, and then Re-Origination procs right after. Furthermore, Tiger's Fury won't come up inside its duration and you've already consumed all of the 4-piece set bonus charges from your last cast.

But suppose you had the foresight to cast that finisher at high energy, so you start with, say, 70 energy. You now have 170 energy to spend during the proc, which is enough for 4 casts of rake/mangle and 1 rip. Immediately you're guaranteed at least a 4-CP rip, but you can likely do better:
  • If the last finisher was cast at 5 CP, then you have a 75% chance to have a combo point due to the 2-piece bonus.
  • If you have a 30% crit chance under Re-Origination (33% before suppression -- that's slightly less than what I have), then you have a ~76% chance to get at least one crit on a CP builder.
  • You have a ~45% chance to get at least one clearcast during Re-Origination.
Put together, you have a 97% chance to pick up another combo point by at least one of these mechanics and thereby cast a 5-CP rip.

Return to Kitty DPS

Who is online

Users browsing this forum: MSNbot Media, Yahoo [Bot] and 3 guests