Remove Advertisements

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

Warning: Theorycraft inside.

Moderators: Fridmarr, Worldie, Aergis, theckhd

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby honorshammer » Wed Sep 23, 2009 1:42 pm

Regarding Mongoose

"HV applications can proc the buff"

Can you suggest a good methodology to test this?
User avatar
honorshammer
Moderator
 
Posts: 2147
Joined: Tue Apr 10, 2007 9:35 am
Location: Charleston, SC

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby majiben » Wed Sep 23, 2009 1:53 pm

honorshammer wrote:Regarding Mongoose

"HV applications can proc the buff"

Can you suggest a good methodology to test this?

Test proc rate with no seal and test with SoV. Be sure not to have the SoV glyph when testing SoV
Amirya wrote:some bizarre lovechild of Hawking, Einstein, and Theck
User avatar
majiben
Moderator
 
Posts: 6999
Joined: Fri Aug 22, 2008 4:37 pm
Location: Retired

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby Petrus » Wed Sep 23, 2009 1:58 pm

In response to the DUH-E thing mentioned earlier: Elementals haven't mattered. It's nice, but we have had zero elemental bosses in Wrath so far. None. So the D-U-H part is what counts, especially considering 4/5 ToC encounters fit the bill for making Crusade best for threat.
User avatar
Petrus
Maintankadonor
 
Posts: 940
Joined: Wed Feb 06, 2008 3:45 am
Location: St. Paul, Minnesota

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby Grehn|Skipjack » Wed Sep 23, 2009 2:02 pm

Petrus wrote:In response to the DUH-E thing mentioned earlier: Elementals haven't mattered. It's nice, but we have had zero elemental bosses in Wrath so far. None. So the D-U-H part is what counts, especially considering 4/5 ToC encounters fit the bill for making Crusade best for threat.

Algalon is an Elemental.
Image
The Seeker.
User avatar
Grehn|Skipjack
Maintankadonor
 
Posts: 4529
Joined: Tue May 27, 2008 8:53 am
Location: Malorne

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby theckhd » Wed Sep 23, 2009 2:23 pm

I think you guys are over-thinking the DUH vs. DUHE thing. I chose to ignore elementals because it made a funnier acronym. That's really all there was to it. :P
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7717
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby tlitp » Wed Sep 23, 2009 2:46 pm

majiben wrote:
honorshammer wrote:Regarding Mongoose

"HV applications can proc the buff"

Can you suggest a good methodology to test this?

Test proc rate with no seal and test with SoV.

tlitp wrote:1. grab a gear set heavy on hit/expertise (aiming for both softcaps) and haste
2. grab a fast weapon (easy picks : sword / mace) for which you have maximum weapon skill, enchant it with Berserking (it's easier to analyze than Mongoose, as it doesn't generate additional haste)
3. maintain a constant haste (eliminate parasite haste sources, buffs/procs)
4. whack a dummy for at least an hour, using only regular autoattacks (+seals)
5. parse the CL for the average uptime, or upload the said CL


one test per seal is enough; the number of proc opportunities is computed from the average uptime

PS. I'll restate : this behavior is not a 3.2.2 change - first experimental data concerning the change of ppm-based procs were gathered in 3.1 PTR; it is true, however, that the matlab code never included the effect of haste on ppm-based procs... up to now, that is
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby Cylan » Sat Sep 26, 2009 8:33 pm

theckhd wrote:I think you guys are over-thinking the DUH vs. DUHE thing. I chose to ignore elementals because it made a funnier acronym. That's really all there was to it. :P

I thought you were all about the details, theck. :P

I'd like to propose HUDE as the new acronym, as it is accurate and doesn't sound like DUH. :D
Image
Cylan
 
Posts: 72
Joined: Sun Nov 25, 2007 9:31 am

Re: 3.2.2 Talent/Spec Analysis

Postby Apollya » Mon Sep 28, 2009 5:50 am

theckhd wrote:3.2.2 Talent/Spec Analysis

First, we'll look at the breakdown of TPS per talent point spent.

Setup:
Rotation: 969
Seals: SoV
Glyphs: V only
Talents: Various, in combinations to isolate the benefit of each talent individually.
Gear: Theck's 9.245 Progression Set
File: calc_talent_analysis.m


Obviously I've just quoted a small part of the text to show where it was from. The main point for me here is the set used for comparison.

The issue is this. Retribution uses Libram of Valiance (incidentally as with many others i use this in my tanking gear too). This libram changes the effectiveness of Seal of Vengeance. Meaning for ret it's actually better to use SoV all the time pretty much, rather than SoComm/ SoR on trash due to the almost 100% uptime and dmg increase.

Perhaps it would be better to test Socomm with libram of Defiance (since not using SoV doesn't matter which libram is used as you will not be getting the 200 Str buff) Vs SoV with Libram of Valiance? to see clearer if we should be using SoV on trash if we use Libram of Valiance.
User avatar
Apollya
 
Posts: 58
Joined: Wed Oct 22, 2008 4:27 am

Re: 3.2.2 Talent/Spec Analysis

Postby æ » Mon Sep 28, 2009 9:46 am

Apollya wrote:This libram changes the effectiveness of Seal of Vengeance. Meaning for ret (Did you mean 'prot' here? Id agree.) it's actually better to use SoV all the time pretty much, rather than SoComm/ SoR on trash due to the almost 100% uptime and dmg increase.


I dont know where to start with this, but this is not true in the least for Retribution. SoC is used for trash, where trash >1. SoR is used for trash where trash = 1. SoV for everything else. The libram you mention would need several times the strength on it to come close to how SoC preforms on multiple trash right now. Im not sure how you cant see that. Im not talking about a couple % here and there, I literally do double the dps with SoC on trash as ret. Maybe Im reading you wrong?

Anyway...


Apollya wrote:Perhaps it would be better to test Socomm with libram of Defiance (since not using SoV doesn't matter which libram is used as you will not be getting the 200 Str buff) Vs SoV with Libram of Valiance? to see clearer if we should be using SoV on trash if we use Libram of Valiance.


This is a more logical post, and also outlined here in this last post by "Quest". At least I think this is what youre implying :P
http://www.wowhead.com/?forums&topic=122555&p=1440361

Snip:
In Thecks' analysis, it basically (in the gearset) comes down to; 3 stacks SoV vs SoC, SoC will come out 350TPS ahead. Now as I outlined in my first paragraph, in an easy encounter, youre going to slap on your 200str libram, and its going to turn into 253 etc str just from our own buffs. (DivStr+Sanc)

On the same page you will see he also did the new analysis of Strength. You gain roughly 2.7 (depends on spec +/-0.2) TPS per strength. For 253, that comes out to 683 TPS. Or a gain of about 333(non repeating)TPS even at only 3 stacks. At two stacks? It would be closer but I doubt SoC would be taking over, also, you can apply 2 stacks on 3-4 targets without tab targeting in a matter of 6 seconds, while your main target will have a full 5 stack by that time.
Last edited by æ on Mon Sep 28, 2009 10:54 am, edited 2 times in total.
10 SIN
20 GOTO HELL
User avatar
æ
 
Posts: 213
Joined: Sun Oct 12, 2008 8:02 pm
Location: Santa Barbara

Re: 3.2.2 Talent/Spec Analysis

Postby Apollya » Mon Sep 28, 2009 10:29 am

æ wrote:I dont know where to start with this, but this is not true in the least for Retribution. SoC is used for trash, where trash >1. SoR is used for trash where trash = 1. SoV for everything else. The libram you mention would need several times the strength on it to come close to how SoC preforms on multiple trash right now. Im not sure how you cant see that. Im not talking about a couple % here and there, I literally do double the dps with SoC on trash as ret. Maybe Im reading you wrong?


Not sure you are reading me wrong on this but in advise I read elsewhere and in talking to others, it was suggested that SoV with libram > SoComm (proviidng the mob(s) do not die in less than 8secs). I perhaps should've added the timeline, but didn't think to since I only tank Hardmodes or raid instances where trash rarely dies in under 8secs.

æ wrote:
Apollya wrote:Perhaps it would be better to test Socomm with libram of Defiance (since not using SoV doesn't matter which libram is used as you will not be getting the 200 Str buff) Vs SoV with Libram of Valiance? to see clearer if we should be using SoV on trash if we use Libram of Valiance.


This is a more logical post, and also outlined here in this last post by "Quest". At least I think this is what youre implying :P
http://www.wowhead.com/?forums&topic=122555&p=1440361

Snip:
In Thecks' analysis, it basically (in the gearset) comes down to; 3 stacks SoV vs SoC, SoC will come out 350TPS ahead. Now as I outlined in my first paragraph, in an easy encounter, youre going to slap on your 200str libram, and its going to turn into 253 etc str just from our own buffs. (DivStr+Sanc)

On the same page you will see he also did the new analysis of Strength. You gain roughly 2.7 (depends on spec +/-0.2) TPS per strength. For 253, that comes out to 683 TPS. Or a gain of about 333(non repeating)TPS even at only 3 stacks. At two stacks? It would be closer but I doubt SoC would be taking over, also, you can apply 2 stacks on 3-4 targets without tab targeting in a matter of 6 seconds, while your main target will have a full 5 stack by that time.

TL;DR
If you use this libram on trash, SoV is going to be better. Go beg Blizzard to allow SoR and SoC to proc your libram.
If you dont use that libram, youre doing encounters where trash does not die within 8-10 seconds....or you still need to get yours


And yes this is more to the point. Basically I was suggesting that the gear set used for SoV tps should include Libram of Valiance not libram of defiance in the modelling. Or perhaps even added as an extra scenario (with / without libram of Valiance)
User avatar
Apollya
 
Posts: 58
Joined: Wed Oct 22, 2008 4:27 am

Re: 3.2.2 Talent/Spec Analysis

Postby æ » Mon Sep 28, 2009 10:44 am

Apollya wrote:Not sure you are reading me wrong on this but in advise I read elsewhere and in talking to others, it was suggested that SoV with libram > SoComm (proviidng the mob(s) do not die in less than 8secs). I perhaps should've added the timeline, but didn't think to since I only tank Hardmodes or raid instances where trash rarely dies in under 8secs.


Ret DPS Code of Justice:
Code: Select all
If mobs > 1
then SoC
else if mobs == 1
then SoR
else if mobs == 1 & mobtimetodeath > ~14 seconds
then SoV


Now back to your regularly scheduled program; The TheckHD show! (always in HD)
10 SIN
20 GOTO HELL
User avatar
æ
 
Posts: 213
Joined: Sun Oct 12, 2008 8:02 pm
Location: Santa Barbara

Re: 3.2.2 Talent/Spec Analysis

Postby Jasari » Mon Sep 28, 2009 11:16 am

æ wrote:Ret DPS Code of Justice:
Code: Select all
If mobs > 1
then SoC
else if mobs == 1 & mobtimetodeath < ~14 seconds
then SoR
else if mobs == 1 & mobtimetodeath > ~14 seconds
then SoV


fixt
User avatar
Jasari
 
Posts: 2195
Joined: Sun Apr 27, 2008 12:32 am
Location: Boston, MA

Re: 3.2.2 Talent/Spec Analysis

Postby Thels » Tue Sep 29, 2009 1:42 am

Ret DPS Code of Justice:
Code: Select all
If mobs > 1
then SoC
else if mobtimetodeath < ~14 seconds
then SoR
else SoV


Is all you need.
ImageImage
User avatar
Thels
 
Posts: 1236
Joined: Tue May 15, 2007 7:30 am
Location: The Netherlands

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby theckhd » Tue Sep 29, 2009 8:10 am

I've updated the SoC vs SoV calculation slightly. For one thing, I wasn't scaling Judgement properly. For another, the damage numbers all seem too low (compare them to the Ability Damage post), which probably means the wrong gear set was loaded. I did that whole thing by hand though, which is partly to blame.

I've written a quick m-file to handle the situation: calc_seals.m

It loads the default gear set in 3 configurations: no glyphs with LoD, V glyph with LoD, and V glyph with LoV. Here's the updated table of ability damage for the 3 configurations, along with "diff" columns for comparing V and V+LoV to SoC ("none").

Code: Select all
Abil    none   V   V+LoV   2-1   3-1
ShoR    3502  3502  3756     0   254
HotR    3386  3478  3723    92   337
Cons    3413  3413  3671     0   258
Exor    2634  2634  2768     0   134
JoV        0  2550  2824  2550  2824
JoC     1148     0     0 -1148 -1148
AS      2512  2512  2586     0    74
HoW     2476  2476  2550     0    74
SoV        0   464   497   464   497
SoC      464     0     0  -464  -464
VDoT       0  4605  5101  4605  5101
HS       927   927   994     0    67
Melee    759   776   831    17    72


It then calculates the net damage differences as follows. Assume N mobs, h stacks of HV per mob.

Judgement damage: You do (JoV-JoC) damage every 9 seconds, divided evenly among the mobs. JoV is scaled by (1+h/10)/1.5, since the simulation generates damage assuming a 5-stack. Thus, the dps is
(JoV*(1+h/10)/1.5-JoC)/9/N


Seal Damage: Instead of the complicated way I calculated it last time, we can approach this more simply. We get SoV*h/5 damage per SoV proc, and we get one every melee swing, 3 every HotR, and 1 every judgement, for a dps value of
SoV*h/5 * (1/player_swing + nprocs/6 + 1/9)
I've replaced "3" with "nprocs," because for N<3, you don't get 3 procs. The code correctly compensates for this (2 procs for 2mobs, etc).

Similarly, we get SoC damage per SoC proc, and we get 3 every melee swing, 3 every HotR, and 3 every Judgement, for
SoC*nprocs*(1/player_swing+1/6+1/9)
Subtracting those two numbers gives you the net benefit of SoV over SoC. Divide by N to get DPS per mob.

Holy Vengeance: This one's fairly easy, it's HV*h/5 damage for an h-stack, over 15 seconds, on every mob. Thus it's a dps of
HV*h/5/15

Finally, we have to include everything else:
ShoR every 6 seconds, divided evenly amongst N mobs (ShoR/6/N)
HotR every 6 seconds, nprocs hits divided evenly amongst N mobs (HotR*3/6/N)
Cons every 9 seconds, hits every mob (Cons/9)
HS procs (at about 0.18 pps) divided evenly amongst N mobs (HS*0.18/N)
melee attacks, divided evenly amonst N mobs (melee/player_swing/N)

Assuming N=3 mobs, we get this for h=0 to 5:
Code: Select all
h        0    1    2    3    4    5
LoD   -456 -345 -234 -123  -12  100
LoV   -347 -226 -104   17  139  260

Or in convenient plot format:
Image
Note that this time around, I've accidentally switched directions - now positive means "SoV is better" and negative means "SoC is better." Regardless, we see reasonably good qualitative agreement - for fewer stacks, SoC wins out. With LoValiance equipped, it only takes 3 stacks per mob for SoV to catch up, whereas without you need a full 5-stack on every mob to make it worth it.

Let's look at the scaling with number of mobs instead. Assuming h=3, we get for N=1 to 10:
Code: Select all
N       1    2    3    4    5    6    7    8    9   10
LoD   133  -59 -123  -46    0   31   53   69   82   92
LoV   367  105   17   71  104  125  140  152  161  168

This table is a little misleading - this simulation doesn't take into account the build-up time of the HV stack, so for many mobs, it will generally appear to favor SoV, because it's assuming you get "free" stacks up. To try and compensate, I've plotted both h=3 and h=1, since one way to compensate is just assume a lower average number of stacks per mob:
Image
As you can see, even with LoValiance, you're better off with SoC if you can't average more than one stack per mob. On the other hand, if you can keep a 3-stack up on everything, you'd be better off using SoV. Keep in mind that most of this is somewhat academic, because Consecration will account for the majority of your AoE threat once you're tanking more than 6 mobs anyway.

The most accurate guess as to the real-world effects would be to use the higher-stack curve for fewer mobs, and lower-stack curve for more mobs. It isn't really fair to model this as a 5/N type decay, since HotR spreads stacks on 3 targets. If anything, it would be a slight decay from N=1 to 3, and then a steeper decay from N=4 on.
We can make a pretty mesh plot of the results, using the LoValiance numbers, to help us with that:
Image
It seems that for less than 3 mobs, where 3+ stacks is reasonable, you'll want SoV. For 4+ mobs, where you're more likely to have 2 or fewer stacks per mob, you'll probably be better off with SoC.

TLDR version:
  • If you're tanking things long enough to get a 3-stack or higher on each mob, use SoV.
  • If you're tanking 3 or fewer mobs, use SoV.
  • If you're tanking 4+ mobs, SoC will likely give you slightly higher average threat generation on each mob.
  • The differences here are all on the order of +/-300TPS except for single-mob scenarios. So SoC will not make or break your aoe threat generation.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7717
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB TPS analysis (A Jonesy derivative work)

Postby æ » Tue Sep 29, 2009 11:09 am

Those are some sexy plots.
10 SIN
20 GOTO HELL
User avatar
æ
 
Posts: 213
Joined: Sun Oct 12, 2008 8:02 pm
Location: Santa Barbara

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