Remove Advertisements

Theck's MATLAB TPS analysis - WotLK/3.x

Warning: Theorycraft inside.

Moderators: Fridmarr, Worldie, Aergis, theckhd

Re: Theck's MATLAB TPS analysis

Postby tlitp » Fri Feb 26, 2010 3:11 am

We're currently working on several different issues, and "we need your dataz"(TM). I'll list them here, in order of decreasing priority :

(1) target_block events

Methodology :
- spec a regular 53/18 build (or a variation of it)
- equip this weapon, with no enchant on it (given that you have maximum weapon skill for it)
- equip a retribution-centric gear set (make sure to have 0 ArP rating)
- be sure to not have an active RED / 4pt7 / 2pt9 / 2pt10
- be sure not to have STR/AP/ArP parasite procs (libram/trinket)
- attack the boss-level dummy from the front, making sure that your attacks do not hit any other targets
- make sure that the dummy isn't hit by other players (i.e. avoid "overkill" reports)
- use autoattacks+HotR(+Divine Plea) for half an hour

Results :
- check the character panel and report your total AP, plus your current hit/expertise
- do not parse the CL, just upload the raw combat log

(2) Vindication [proc chance]

Methodology :
- spec a regular 53/18 build (or a variation of it), with Vindication
- equip this weapon, with no enchant on it
- equip a protection-centric gear set (stats are irrelevant)
- be sure not to have haste parasite procs
- attack a level 60 target dummy
- use only autoattacks for at least an hour (you can AFK freely)

Results :
- report your current (passive) haste
- parse (WMO/WoL) the corresponding CL

(3) Vindication [triggers]

Methodology :
- spec a regular 53/18 build (or a variation of it), with Vindication
- equip this weapon, with no enchant on it
- equip a protection-centric gear set (stats are irrelevant)
- be sure not to have haste parasite procs
- attack a level 60 dummy using only ranged abilities, and making sure that your attacks do not hit any other targets
- check J / ShoR / HotR / AS / HoW (10-20 casts, or until the first Vindication proc)

Results :
- report your findings

(4) SoComm's "cleave"

Methodology :
- spec a regular 53/18 build (or a variation of it)
- equip this weapon, with no enchant on it
- equip a protection-centric gear set (with very low hit/expertise)
- attack the boss-level dummy from the front, making sure that SoComm can cleave to another target
- use autoattacks+SoComm
- check if an avoided SoComm on the primary target (the boss-level dummy) can "cleave" or not

Results :
- report your findings

(5) accumulation of BW's stacks

Methodology :
- spec a regular 53/18 build (or a variation of it)
- equip this weapon, with BW on it
- equip a protection-centric gear set (stats are irrelevant)
- position yourself amidst the level 60 dummies
- use autoattacks+SoComm+HotR(+Divine Plea)
- check if BW's stacks accumulate sequentially (i.e. from i to i+1 stacks) or they can "jump" (from m to m+n, n>1)

Results :
- report your findings

(6) rechecking Aegis

Methodology :
Equip TBH (build/gear/stats are all irrelevant). Hit a dummy with the regular 6/9 combo. If TBH can still proc off paladin's own attacks you should see a proc in less than one minute. If you don't spot any procs in 5 mins, TBH was (stealthily) hotfixed.
User avatar
tlitp
 
Posts: 554
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB TPS analysis

Postby Yelena » Fri Feb 26, 2010 4:30 pm

tlitp wrote:(2) Vindication [proc chance]

Methodology :
- spec a regular 53/18 build (or a variation of it), with Vindication
- equip this weapon, with no enchant on it
- equip a protection-centric gear set (stats are irrelevant)
- be sure not to have haste parasite procs
- attack a level 60 target dummy
- use only autoattacks for at least an hour (you can AFK freely)

Results :
- report your current (passive) haste
- parse (WMO/WoL) the corresponding CL

http://www.worldoflogs.com/reports/ena9 ... details/0/

-64 minute log.
-Gear/spec are identical to what's in my Armory currently, except for using the required weapon for the test.
-No passive haste.
-1282 melee attacks (214 of which were critical hits).
-524 Vindication procs.
Valleri - Frostwolf
Yelena
 
Posts: 314
Joined: Sat Jun 27, 2009 4:54 am
Location: Behind the Eight Ball

Re: Theck's MATLAB TPS analysis

Postby Yelena » Fri Feb 26, 2010 5:25 pm

tlitp wrote:(3) Vindication [triggers]

Methodology :
- spec a regular 53/18 build (or a variation of it), with Vindication
- equip this weapon, with no enchant on it
- equip a protection-centric gear set (stats are irrelevant)
- be sure not to have haste parasite procs
- attack a level 60 dummy using only ranged abilities, and making sure that your attacks do not hit any other targets
- check J / ShoR / HotR / AS / HoW (10-20 casts, or until the first Vindication proc)

Results :
- report your findings

I'll do more extensive testing on this one over the weekend (when I have more time), as well as provide a log. There's a few anomalies that we probably need to iron out, which I'll detail below.

I noticed, as I was testing Judgement (using Seal of Light), that Vindication (when it procced) was consistently being applied after the Judgements of the Just debuff. It has already been documented enough to show that Judgements of the Just acts funny in that it does things that regular Judgements do not; like refresh Divine Plea from outside of melee range for example, when the absence of the talent (as I confirmed during my Last Word testing) will not cause Judgements to refresh DP. Other ranged attacks, like Avenger's Shield, also do not refresh DP. I did look for double procs with Judgement to see if there were two potential triggers, but my brief testing didn't show any. I'll tweak with one of my specs over the weekend to isolate JotJ.

Hammer of Wrath is impossible to test in the given environment, as the target has 1 hit point and is always at 100% health. Zul'Gurub resets tomorrow; I'll try to remember to take a combat log of Bloodlord Mandokir as I solo him, switching weapons, turning auto-attack off, and removing applicable buffs when he reaches execute health.

Edit1: Bloodlord Mandokir
Edit2: Expert's Training Dummy

-20 Judgements, without Judgements of the Just, Vindication procs.
-19 Judgements (I lost count), with Judgements of the Just, Vindication procs. Still didn't notice any double procs, so it may well not be a trigger, I'll chalk my initial suspicion up to the delayed application of the debuff.
-20 Shield of Righteousness, Vindication procs.
-20 Hammer of the Righteous, Vindication procs.
-10 Avenger's Shield, Vindication procs.
-Bloodlord Mandokir - Hammer of Wrath, no buffs/procs/melee (with the required weapon swapped in for the test), Vindication procs.

Interestingly enough, Avenger's Shield seemed to proc Vindication every time. The sample size in the log is small, but further testing is showing the same behavior.
Last edited by Yelena on Sat Feb 27, 2010 11:01 pm, edited 2 times in total.
Valleri - Frostwolf
Yelena
 
Posts: 314
Joined: Sat Jun 27, 2009 4:54 am
Location: Behind the Eight Ball

Re: Theck's MATLAB TPS analysis

Postby Ragingsoul » Sat Feb 27, 2010 7:20 pm

post raw combat log m8, you can't do anything with those logs on your own.
While tanking new content, one rule is important -- it's not about taking less damage; it's about being able to take more.
Image
Ragingsoul, Palatank, Tsunadee, Resto Druid, Bisoul, Mutilate Rogue
User avatar
Ragingsoul
 
Posts: 346
Joined: Sat Jul 26, 2008 10:01 pm

Re: Theck's MATLAB TPS analysis

Postby Yelena » Sat Feb 27, 2010 10:04 pm

I'll upload a log of the trigger tests when I have time to take one, as I stated in the first sentence of my previous post.

Edit: Previous post updated with WoL parses.
Valleri - Frostwolf
Yelena
 
Posts: 314
Joined: Sat Jun 27, 2009 4:54 am
Location: Behind the Eight Ball

Re: Theck's MATLAB TPS analysis

Postby xstratax » Sun Feb 28, 2010 2:20 am

Yelena wrote:Interestingly enough, Avenger's Shield seemed to proc Vindication every time. The sample size in the log is small, but further testing is showing the same behavior.


As far as I knew this was widely known, AS does proc Vindication on all 3 targets everytime you cast it. If you just stand back and AS only it will show.
xstratax
Maintankadonor
 
Posts: 108
Joined: Fri Aug 14, 2009 1:18 pm

Re: Theck's MATLAB TPS analysis

Postby Yelena » Sun Feb 28, 2010 6:43 pm

I probably glazed over that when Vindication changed to what it currently is. One of those things you read & see, without fully processing it. I noticed the behavior a long time ago, in retrospect, but generally speaking in a raid I'm usually more focused on other things than noticing a proc off of every single application of a single ability.
Valleri - Frostwolf
Yelena
 
Posts: 314
Joined: Sat Jun 27, 2009 4:54 am
Location: Behind the Eight Ball

Re: Theck's MATLAB TPS analysis

Postby xstratax » Mon Mar 01, 2010 12:18 am

I wouldnt have noticed if not for testing TankadinTPS on a Dummy a while back.
xstratax
Maintankadonor
 
Posts: 108
Joined: Fri Aug 14, 2009 1:18 pm

Re: Theck's MATLAB TPS analysis

Postby tlitp » Mon Mar 01, 2010 1:51 am

Yelena wrote:
TreeofMana wrote:Alright, I got done testing the uptime of BW so here are my results. This is the spec I ran with.

Code: Select all
hit rating = 103
haste rating = 0
base swing timer = 3.0

http://www.worldoflogs.com/reports/ena9 ... details/0/

Code: Select all
haste rating = 0
base swing timer = 3.0

  1. Vindication is on spell hit table, so Tree's chance to miss with Vindication was 0.0735% (and 1.8588% to miss with autoattacks), whereas Yelena's was 0%
  2. Tree's averaged swing timer is computed to ~2.73, after finding the value of the parryhaste factor

Possible mechanics for Vindication :
1. flat % on hit
2. ppm-based, not haste-normalized
3. ppm-based, haste-normalized
4. icd-based, ICD > buff duration
5. icd-based, ICD < buff duration
#4 and #5 are busted right of the bat, after a quick review of the logs.

Assuming hypothesis #1 :
Code: Select all
proc_chance=(1-nthroot(1-0.842,10./2.73))./(0.981412.*0.999265); %Tree
proc_chance=0.4035

proc_chance=(1-nthroot(1-0.83,10./3))./(1.*1); %Yelena
proc_chance=0.4123

Assuming hypothesis #2 :
Code: Select all
ppm=(1-nthroot(1-0.842,10./2.73))./(0.981412.*0.999265.*3./60); %Tree
ppm=8.0704

ppm=(1-nthroot(1-0.83,10./3))./(1.*1.*3./60); %Yelena
ppm=8.2466

Assuming hypothesis #3 :
Code: Select all
ppm=(1-nthroot(1-0.842,10./2.73))./(0.981412.*0.999265.*2.73./60); %Tree
ppm=8.8685

ppm=(1-nthroot(1-0.83,10./3))./(1.*1.*3./60); %Yelena
ppm=8.2466

#3 is also busted. A (parry)hasted weapon should have a lower proc chance/buff uptime, given that #3 is valid.


At this point it's either #1 or #2 (and the two tests above are inconclusive). But after reviewing some more testing sessions, #1 is busted as well. As far as autoattacks are concerned, Vindication is 8 ppm, not haste-normalized.
Last edited by tlitp on Mon Mar 01, 2010 9:57 am, edited 1 time in total.
User avatar
tlitp
 
Posts: 554
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB TPS analysis

Postby jere » Mon Mar 01, 2010 6:08 am

Just a quick question, what does "not haste normalized" mean? The PPM changes with haste or that the PPM is independent of haste?
Image
User avatar
jere
 
Posts: 2951
Joined: Fri May 11, 2007 5:12 pm

Re: Theck's MATLAB TPS analysis

Postby tlitp » Mon Mar 01, 2010 10:07 am

jere wrote:what does "not haste normalized" mean? The PPM changes with haste or that the PPM is independent of haste?

Proc chance per hit and haste are inversely proportional for haste-normalized effects; for effects that aren't haste-normalized, the proc chance is static, haste-independent.

haste-normalized : uptime=1-(1-ppm*hasted_swing_timer/60)^(buff_duration/hasted_swing_timer)
not haste-normalized : uptime=1-(1-ppm*base_swing_timer/60)^(buff_duration/hasted_swing_timer)
User avatar
tlitp
 
Posts: 554
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB TPS analysis

Postby Offtrack » Thu Mar 04, 2010 4:31 am

(4) SoComm's "cleave"

Methodology :
- spec a regular 53/18 build (or a variation of it)
- equip this weapon, with no enchant on it
- equip a protection-centric gear set (with very low hit/expertise)
- attack the boss-level dummy from the front, making sure that SoComm can cleave to another target
- use autoattacks+SoComm
- check if an avoided SoComm on the primary target (the boss-level dummy) can "cleave" or not

Results :
- report your findings


Avid reader here, and I thought I'd help out in whatever small way I can with all the amazing work you all do here, so I decided to spend some time on the Target Dummies in SW to test out SoComm's cleave.

Stats: 0 Hit / 9 Expertise
Setup: Attacking heroic boss level target dummy with SoComm cleaving to level 70 target dummies, as well as attacking level 70 target dummy cleaving to another 70 and the boss level heroic dummy.

After watching my combat log, I found that if the heroic target dummy avoids a SoComm proc, it does not prevent it from cleaving to other targets. Additionally, the secondary cleave targets also have a chance to miss. I'm under the impression that the cleave affect is similar to how HotR and AS chain. Both HotR and AS will chain to it's next target even if the initial hit misses. Unlike HotR/AS however, SoComm appears to cleave 360 degrees around your character while HotR/AS only chains to targets that are in a 180 degree frontal cone.

I hope at least some of this is useful information. If you need me to combatlog anything, let me know. I've got free time today to kill.

Also, I plan on logging some time to collect data for target_block events, I'll post a combat log up when I'm done.
Offtrack
 
Posts: 1
Joined: Thu Mar 04, 2010 3:30 am

Re: Theck's MATLAB TPS analysis

Postby Metherlance » Fri Mar 05, 2010 7:50 am

Paging DR. Theck. Could you do those "stats from gear" graphs using SoC(ommand)? I'd guess crit has a lot more value there, prolly haste too.
http://eu.battle.net/wow/en/character/chromaggus/Metherlance/advanced

No question remains unanswered. No doubts linger. You are Azeroth's greatest champions!
Metherlance
 
Posts: 90
Joined: Thu Jan 14, 2010 2:30 am
Location: Ljusdal, Sweden

Re: Theck's MATLAB TPS analysis

Postby theckhd » Fri Mar 05, 2010 8:03 am

Metherlance wrote:Paging DR. Theck. Could you do those "stats from gear" graphs using SoC(ommand)? I'd guess crit has a lot more value there, prolly haste too.

I can, but I'm not sure what purpose it would serve. The AoE round-up already covers how to gear for AoE tanking with SoCom.

I could generate a plot of single-target stat contributions for SoCom, but you wouldn't want to be using it for anything that survives more than 10 seconds anyway, and anything with a shorter lifespan isn't something you typically gear for.

What question were you trying to answer? If it's something I've overlooked, I can run the numbers for it.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Call to Arms 5.x, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7658
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB TPS analysis

Postby Metherlance » Fri Mar 05, 2010 8:16 am

theckhd wrote:I can, but I'm not sure what purpose it would serve. The AoE round-up already covers how to gear for AoE tanking with SoCom.

I could generate a plot of single-target stat contributions for SoCom, but you wouldn't want to be using it for anything that survives more than 10 seconds anyway, and anything with a shorter lifespan isn't something you typically gear for.

What question were you trying to answer? If it's something I've overlooked, I can run the numbers for it.


Oh my bad. Totally missed that aoe round-up post. Thanks!
http://eu.battle.net/wow/en/character/chromaggus/Metherlance/advanced

No question remains unanswered. No doubts linger. You are Azeroth's greatest champions!
Metherlance
 
Posts: 90
Joined: Thu Jan 14, 2010 2:30 am
Location: Ljusdal, Sweden

PreviousNext

Return to Advanced Theorycraft and Calculations

Who is online

Users browsing this forum: No registered users and 1 guest


Remove Advertisements

Who is online

In total there is 1 user online :: 0 registered, 0 hidden and 1 guest (based on users active over the past 5 minutes)
Most users ever online was 380 on Tue Oct 14, 2008 6:28 pm

Users browsing this forum: No registered users and 1 guest