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 Awyndel » Sat Dec 19, 2009 7:40 am

Wrathy wrote:
Kihra wrote:
Awyndel wrote:I also swapped conviction for socomm. For the adds on certain bosses it will be nice.

Bit off topic, doing 6K dps in heroics is great for showing off and all but i'd rather save downtime by using seal of wisdom.


This really isn't an issue. You can SoComm tank the heroics by just chain pulling fast enough to ensure Divine Plea stays up. If you do ever find yourself in situations where Divine Plea is forced to fall off, just cut out Consecrate and you'll be fine.


Yeah I rarely run out of mana. It all depends on your gear level and your healers ability to maintain their mana pool. As long as you healer has enough mana, Sanctuary should provide you with more than enough mana to keep you topped off. If you keep dropping too low on your mana pool, remove consecrate from your rotation. I know this is a bit counter intuitive for an AoE trash pack, but the fact that you have SoComm and HotR is more than enough to get 5 to 6 mobs stuck to you.


Well maybe it's coz I take no melee damage at all in my str gemmed block set. I guess I can try swapping concecrate for the seal. You sure that's more dps though? ( hey look, back on topic lol :) ) I really like concecrate for picking up groups though.

T10 2set 110 str, omgwtf I guess that's worth swapping the gunship shoulders for the tier ones. That's like a whole extra item worth of threat.
User avatar
Awyndel
 
Posts: 672
Joined: Sat Feb 14, 2009 8:49 am
Location: The Netherlands

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

Postby Grehn|Skipjack » Sat Dec 19, 2009 7:49 am

Depending on the size of the groups you're pulling, it really shouldn't matter if you're taking damage or not. After a certain point, Divine Plea and BoSanc are enough to keep your mana pool at max no matter what you do. I'm inclined to say it's around 4 melee mobs swinging at you.
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 » Sat Dec 19, 2009 10:22 am

tlitp wrote:On the topic at hand : we have some numbers. And they show that 1 point in Redoubt improves the overall threat output more than 1 point in Conviction. But this is true for a very particular set of variables and conditions, i.e. a given gear set, a given raid environment (buffs/debuffs), a given encounter (the PC attacks precisely one boss-level NPC, and it is attacked in return by the same boss-level NPC; the "time on target" is precisely 100%) and so on. Where I'm getting at : what holds true for a specific set of input vars/params may not for another set, so one should really think "out of the box". Following blindly (sorry if it's a bit harsh) some numbers posted by Theck or anyone else, without understanding how they are obtained and what's their relevance, is not at all very productive.

Please keep in mind that, by default, Theck's generic analysis is based on a Patchwerk-style 25man encounter (unless other situation is stated explicitly).

To be fair though, the sims are going to be accurate for the vast majority of boss fights. As long as you can keep up a consistent 969 rotation, the sims will fairly accurately reflect the damage you do to a single target.

For AoE tanking, you could argue that Conviction would give you more overall damage than Redoubt. It's something I'll be able to calculate explicitly soon with this new AoE rotation module I'm writing. But then you're losing the mitigation value of Redoubt, which is very strong in AoE situations.

So either way, the choice there seems clear - You're trading a weak single-target threat talent (Conviction) for a strong AoE talent (Seal of Command). Redoubt is strong single-target and strong multi-target (still solid damage and extra mitigation in the form of block). By dropping Conviction instead of Redoubt, you make the smallest impact on your single-target threat and gain a considerable amount of AoE threat from SoCom.

However, as you gain more Strength, Conviction improves in value while Redoubt does not (in fact, it gets weaker due to DR). I could imagine that by the end of the expansion, we'll reach a point where Conviction ends up being stronger point-for-point than Redoubt for threat on a single target. We're just not there yet.
"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: 7849
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

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

Postby Iselian » Sat Dec 19, 2009 10:40 am

tlitp wrote:Redoubt has one static and one dynamic component. While the static one is time-invariant, the dynamic one is not. In other words : the blocking value is a "consistent" metric, but the blocking chance is stochastic. Blocking itself should never be considered "consistent" mitigation, and it's not related to the traditional EH metric.

One might argue that things change a bit when being block-capped (and they do), but that merely makes blocking a conditional mitigation. Damned purists... :mrgreen:


I was speaking about the threat aspect of Redoubt's additional block value versus the threat of 1% more critical hits (ie, 1 point in Conviction). All of Theck's math aside, I meant to say that Conviction's crit either will or will not work, much like how 1% dodge from anticipation will or will not work.

Redoubt on the other hand adds a baseline amount of block value that you will see in every ShoR that lands. In the 20-30 seconds on a set of Deathwhisper adds, you might not see that 1% crit play out and help your threat. You will see Redoubt's increase on ShoR (theoretically, at least 3 times).


All that said, <3 SoCleave.
Iselian
 
Posts: 116
Joined: Fri Jan 23, 2009 11:12 pm

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

Postby tlitp » Sun Dec 20, 2009 5:03 am

In order to expand the MATLAB code to properly assess combat situations versus trash mobs, we need the base armor values of 80-82 elites. Anyone that has a hunter is asked to provide the said data by using Beast Lore on some easily accessible elites :
Thanks in advance.
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

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

Postby Kihra » Sun Dec 20, 2009 3:15 pm

Awyndel wrote:Well maybe it's coz I take no melee damage at all in my str gemmed block set. I guess I can try swapping concecrate for the seal. You sure that's more dps though? ( hey look, back on topic lol :) ) I really like concecrate for picking up groups though.


Yeah I don't use a block set when tanking heroics anymore. Block value just buffs the threat you do with one ability, but Strength buffs everything. My focus has shifted to maximizing Strength when running heroics. Now admittedly a block value threat set could gem Strength and maybe surpass the higher ilvl gear in terms of Strength, but I've never really cared enough to do that. :)

I usually run heroics with ridiculously geared DPS, but with Seal of Command I basically throw my Avenger's Shield, run in, HotR, SoR, maybe judge something once, HotR again and the mobs are all dead by then. :)

I rarely even bother with Consecrate on the packs of 2-4, since stuff dies so fast that I don't really even need it. Also, since I'm chain pulling, I rarely stand still long enough to even be in a Consecrate for enough ticks to make it worthwhile.
Kihra
 
Posts: 554
Joined: Mon Dec 31, 2007 12:01 pm

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

Postby Epimer » Mon Dec 21, 2009 2:36 am

Kihra wrote:Yeah I don't use a block set when tanking heroics anymore.


Same here, I just do then in my MT gear set these days. My block set might have 4k (with libram) BV, but my MT set is expertise soft-capped, has a healthy dose of hit and a ton more strength (because the pieces are higher iLvl) than the block stuff.

And in that set I'm don't ever oom, either, even doing VH with a disc priest healing. So I'm not sure how people are running out of mana in heroics these days, because I'm not doing anything special other than being conscious of how much mana I have from pull to pull and adjusting accordingly.
User avatar
Epimer
 
Posts: 1496
Joined: Wed Jun 18, 2008 6:54 am

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

Postby Racolus » Mon Dec 21, 2009 3:09 am

Gee, the SoComm is doing nutty damage and I can't find any reason to go without it.

It is just too good to be true, what I fear is that it will go as how it came in.

At present time, enjoy to be the king of AOE tank again!
Racolus
 
Posts: 91
Joined: Sat May 31, 2008 5:09 pm

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

Postby Jaesthenis » Mon Dec 21, 2009 3:30 am

theckhd wrote:Yes, though Reckoning uptime will still be pretty abysmal.


Even with chill of the throne (who basically increase chance of reckoning proc by 20%) and 1.2 swing timer ?

On 2+ adds, reckoning coupled with SoComm is really amazing :
each reckoning swing proc 2 SoComm
Chill of the throne + 2+ adds ensure a really good uptime

As 2 of 4 bosses available have adds, and the two other have 1.2 swing timer, I was thinking at picking reckoning.
Is the reckoning uptime so low, even with 1.2 swing timer & chill of the throne ?
Mande-étoile Jaesthenis, FR-La Croisade écarlate
Image
Jaesthenis
 
Posts: 18
Joined: Fri Sep 11, 2009 7:51 am

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

Postby theckhd » Mon Dec 21, 2009 7:12 am

Jaesthenis wrote:
theckhd wrote:Yes, though Reckoning uptime will still be pretty abysmal.


Even with chill of the throne (who basically increase chance of reckoning proc by 20%) and 1.2 swing timer ?

On 2+ adds, reckoning coupled with SoComm is really amazing :
each reckoning swing proc 2 SoComm
Chill of the throne + 2+ adds ensure a really good uptime

As 2 of 4 bosses available have adds, and the two other have 1.2 swing timer, I was thinking at picking reckoning.
Is the reckoning uptime so low, even with 1.2 swing timer & chill of the throne ?

Yeah. The equations of interest are in this post.

theckhd wrote:Reckoning uptime is most easily calculated by figuring out it's downtime, i.e. the chance that Reckoning is not active. That chance is:

Code: Select all
reck_down = (1 - 0.02*Reck_Points*(1-player_avoidance))^(4*weapon_swing*incoming_attack_rate)

In other words, there's a 0.02*(# of points in Reckoning)*(chance you don't avoid an attack) chance that reckoning will proc on any given attack. There's a 1-(that stuff) chance it won't proc, so if you're taking N attacks per second, the chance of not having a proc to refresh it in the time it takes for the effect to expire is (1-stuff)^(N * expiration_time)

The expiration time is 4x your weapon swing rate, since you probably won't be using a 2.0+ speed weapon for this.

In that post, I plugged in numbers for anub adds, which is 8 or 16 attacks per second. For this, let's plot uptime for 1 through 5 points in reckoning, and for N=1 to 16. I'll use the following static values:
weapon_swing = 1.3 seconds
player_avoid = 40

Image
You can see that you need to be taking at least 2+ attacks per second and 5/5 Reckoning to even get 50% uptime. It only starts to shine when you're taking 4+ attacks per second. This is possible while tanking 4-5 adds, but not any currently-available bosses.

Most bosses will still be under 1 attack per second, which means even with 5/5 Reckoning you're looking at an abysmal 20% uptime. Note that this will get worse as your avoidance increases as well.

Here's the m-file in case anyone wants to play with it.
"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: 7849
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

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

Postby theckhd » Mon Dec 21, 2009 7:47 am

Also, I've gone back and updated the first 3 analyses to reflect the change from a base boss swing speed of 2.0 to 1.0. The Ability Damage and Stat calculations haven't really changed, but the Talent one has a noticeable difference - Reckoning got enough of a boost to out-threat Conviction.

So Conviction is now our weakest threat talent, even on single targets (i.e. bosses), and has a stronger benefit in AoE situations.

This finally makes it an interesting choice for the 3 discretionary prot points - you can take utility with DS/DG, or more threat (roughly 190 TPS, or 63 per point) with 3/5 Reckoning.

I will probably be busy for the rest of the morning, but will try and get the AoE post up around lunchtime. If I get time, I'll get the rest of the updated analyses posted as well.
"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: 7849
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

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

Postby æ » Mon Dec 21, 2009 9:24 am

I did an Excel version a few months back for fun if anyone is interested as well with reck.

http://sites.google.com/site/questpsdps/excel

By the way Theck what do you think of FreeMat? Its giving me alot of errors and Im inclined to think its compatibility issues or something Im doing wrong. Some files work others dont.
10 SIN
20 GOTO HELL
User avatar
æ
 
Posts: 213
Joined: Sun Oct 12, 2008 8:02 pm
Location: Santa Barbara

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

Postby theckhd » Mon Dec 21, 2009 10:56 am

æ wrote:By the way Theck what do you think of FreeMat? Its giving me alot of errors and Im inclined to think its compatibility issues or something Im doing wrong. Some files work others dont.

I've used FreeMat once or twice at most, and not for anything complicated, so I don't really have much of an opinion either way. I'd like to be better about maintaining compatibility with Octave, Freemat, and the like, but it's enough of a job just to keep these files updated and working in MATLAB itself.
"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: 7849
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

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

Postby theckhd » Mon Dec 21, 2009 2:07 pm

3.3 AoE Round-Up

I've written some code to several of the calculations to give me some AoE DPS and TPS numbers. Rather than go back and stitch them in, I'll just put them together in one post. I haven't decided how I'll handle these in the future; it's actually a bit easier to keep these calculations separate from the main Stat and Talent analysis files. I may integrate these into the individual calculation posts rather than make a separate AoE post, though.

To run these calculations, I use the following m-files:
trash_stats.m
rotation_AoE.m
calc_aoe.m

trash_stats is a modified boss_stats file with the level set to 82 (and thus lower dodge/parry/miss chance) as well as a modified swing timer (2/n, where n is the number of mobs).

rotation_AoE takes the output of rotation_969, and calculates two AoE DPS values for each seal. The first integrates all abilities based on the number of targets, so HotR hitting 3 targets instead of one, Consecration hitting n targets, Holy Vengeance spread amongst n targets, and so on.

The second just uses multi-target abilities, so it only includes Hammer of the Righteous (and its associated seal procs), Consecration, and Holy Shield. Seal of Command procs from Judgement, melee attacks, and Shield of Righteousness are also included.

calc_aoe runs the talent and threat stat algorithms using rotation_AoE's numbers.

I'm going to incorporate weapon choice into the updated version of the Weapon Analysis rather than put it here. The weapon algorithm is a lot of code, and trying to copy/paste that into the aoe file would have been a pain in the ass. I should have that done later this afternoon, but probably won't get to post it until tomorrow morning.

First lets look at talents, as we did in the Talent Analysis.

Setup:
Rotation: 969
Seals: SoC
Glyphs: V only (effectively increasing the expertise of the default set by 10)
Talents: Various, in combinations to isolate the benefit of each talent individually.
Gear: T10 Progression Set
File: calc_talent_analysis.m
Number of targets: 5

Image
Image
If you look back at the single-target plot in the talent analysis, you'll see a few noticeable differences. DPS has increased for most skills since we're hitting 5 targets now, and each talent has grown stronger or weaker because certain abilities are no longer being considered.

Redoubt is much less influential in the AoE calculations, and drops out completely in the limited skill set that drops ShoR. Seals of the Pure drops out of both because it doesn't affect Seal of Command.

Reckoning gets a big boost due to higher uptime. With n=5, our swing timer is effectively 0.4 seconds, or 2.5 attacks per second. You can see from the Reckoning Uptime Plot that this is enough to get a decent amount of uptime. This just reinforces what we already knew - that Reckoning was strong in situations where the incoming attack rate is very high.


Next let's look at stats.

Setup:
Rotation: 969
Seals: SoC
Glyphs: none
Talents: 1V+3C
Gear: T10 Progression set
File: calc_stat_tps.m
Number of targets: 1-10

Image
DPS plot

Just looking at the pure multi-target skill set, we see that expertise and hit dominate the chart. Strength is in a distant third place, with AP a slightly more distant fourth, and everything else far behind.

The full skill-set plot looks very similar, but with BV hovering below Agi and Crit, and ArPen in a distant last place. You can see the plot here if you're interested.

This means that as far as gearing goes, for AoE trash clears you should aim to hit the expertise soft cap, and then the melee hit cap. Keep in mind that these caps are lower for trash - it's 6% melee hit and 6% dodge/parry, which is 197 hit rating and 24 expertise.

After that, just stack STR.


TLDR Summary:
  • AoE tanking makes Reckoning good again
  • For trash gearing, Expertise (to 24) > Hit (to 6%) > STR (to the ground sky, baby!)
"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: 7849
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

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

Postby æ » Mon Dec 21, 2009 4:25 pm

theckhd wrote:
æ wrote:By the way Theck what do you think of FreeMat? Its giving me alot of errors and Im inclined to think its compatibility issues or something Im doing wrong. Some files work others dont.

I've used FreeMat once or twice at most, and not for anything complicated, so I don't really have much of an opinion either way. I'd like to be better about maintaining compatibility with Octave, Freemat, and the like, but it's enough of a job just to keep these files updated and working in MATLAB itself.


No worries about keeping compatibility with them, they should be keeping compatibility with you anyway!

/runs off to edit some m files and stack exp and hit
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

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