Theck's MATLAB thread - Cataclysm/4.x

Warning: Theorycraft inside.

Moderators: Fridmarr, Worldie, Aergis, theckhd

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby vexryn » Mon Dec 06, 2010 11:29 pm

theckhd wrote:
vexryn wrote:-I wonder if/whether/how much they can buff the proc rate on Grand Crusader to make it a reasonable single target talent without completely distorting the multi-target threat mechanics. Or perhaps it's still their fundamental goal to make Grand Crusader primarily a threat-AoE talent? On the other hand, there's still no primary threat talent you would necessarily replace it with. On single target and AoE, it's still better than Hallowed Ground. And our holy-tree threat talents seem negligible across the board (JotP and Arbiter of the Light). So in the end, it would seem that if you don't want Grand Crusader to support your AoE threat, you're replacing it with utility talents (Imp HoJ, Guarded by the Light, or Eternal Glory).


The problem isn't really the proc rate, it's the confining nature of 939. To fit extra casts in, you need to push back SotR, which is a big chunk of damage. The only way to make GC a useful proc is for the proc to make AS objectively better than anything else you were going to cast. So for example:
-If it gave 1 Holy Power, it could be objectively better to use AS immediately (SotR>AS>CS>J>Cons>HW)
-If it triggered SD, it would be better than Judgement, and would be the optimum choice for any filler (SotR>CS>AS>J>Cons>HW)
-If it gave AS +50% damage, so that it out-damaged a crit SotR, it would probably be better to use it immediately

That's the sort of change it will take to "fix" Grand Crusader. That or nerfs to other abilities.[/quote][/quote]


Well, I would assume that any buff to the Grand Crusader proc rate still increases the likely hood that our single target rotation becomes a "stable" CS-J-CS-AS-CS-SotR-repeat. As is, it seems that you "only" have to sub out AS for Consecrate if GC doesn't proc. You only ever have to use HW if your AS fails to proc for an extended streak. For example, you open with CS-J-CS-AS-CS-SotR; the second round is CS-J-CS-Cons-CS-SotR; by the third round, you're back to CS-J-CS-AS-CS-SotR because AS will be up; if you fail to proc AGAIN, now you'll see CS-J-CS-AS-CS-SotR.

So buffs to Grand Crusader reduce the likelihood you ever have to filter in HW, and to a lesser extent allow you to keep AS over Consecrate (especially if it's glyphed). Any buff to the GC rate would increase the likelihood that HW pushes off the table in the fourth round of the rotation.

Erm, which is a long-winded way of realizing that I guess that's your point. Buffing the GC rate primary just allows you to sub in an AS for a HW every four cycles through the rotation, which is a pretty mediocre gain for 2 talent points. :) Slightly more value because you can also sub in an AS for a Consecrate as well, although that gain is fairly small unless AS is glyphed (or even a tiny loss if Consecrate is glyphed and AS is not)?

theckhd wrote:
vexryn wrote:- I'm still wrapping my head around the interesting expertise/hit tradeoff. Expertise gives me more dps/tps, but hit keeps my taunts from missing. Kudos to Blizzard for giving me an interesting decision. :)

Taunts don't miss anymore, since patch 4.0.1.


Wow. How on earth did I fail to read that in patch notes?


theckhd wrote:
vexryn wrote:- I can't believe just how low HW really is. I'm very curious to see the AoE rotation numbers. If it's low enough, I have to wonder if a Grand Crusader spec can push HW below AS and Cons even on an AoE rotation to the point that with a decent proc rate on Grand Crusader, you barely push the button?

That's basically how the 4.0.1 sims played out. Inq>HotR>AS>Cons>HW>J is my bet for the new AoE rotation. Maybe Cons>HW>AS for lots of targets, just to maintain roughly equal aggro on everything. AoE is quite freestyle so as it is right now, which is a lot of fun. It's really more of an art form than anything, reacting to threat fluctuations on each mob with whatever tools are off of cooldown.


It still really gives me the feeling that HW is "broken" somehow, though. I just have trouble believing that the grand purpose of HW was just to be filler for when Consecrate is down. I guess we give it a slight edge for the utility of the glyphed stun given the number of elementals in Cata?
vexryn
Maintankadonor
 
Posts: 157
Joined: Tue Oct 28, 2008 3:12 pm

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby Extermi » Wed Dec 08, 2010 12:47 am

Hi Theck,

first of all, a big thank you for all the hard work you put here to help us getting better - I cannot count how often I have been making gear and talent choices after reading you nice work. Especially big thanks for including Inq/HotR simulations. ANd, now I have a question.

Going through the priority simulations, I noticed that inquisition rotations were not that far behind as I thought, and some alternating with ShoR were actually quite good. I was wondering if this could be put "to the extreme":

- During rotation, if Inquisition is down, cast CS else HotR to build up 3 HoPo
- Once 3 HoPo are reached, check if Sacred Duty has procc´ed from judgement. If so cast ShoR, else cast Inq

That way, you would cast only critical ShoR, while at the same time getting good use of Holy Power for increasing holy damage over the board if we have no proc luck. That would also often generate situations where a critical ShoR is cast while Inq is active. Would that be significantly better than standard 939, or not ?

Thanks,

Extermi
PS: Just a side note on the Grand Crusader talent: In my eyes, it is mostly a (strong !) PvP talent so seeing only marginal DPS increase in PvE is what I would expect.
Extermi
 
Posts: 225
Joined: Mon Oct 22, 2007 1:23 pm

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby KysenMurrin » Wed Dec 08, 2010 7:52 am

theckhd wrote:
vexryn wrote:- Did the 1.4x Stamina conversion actually hit? Are we scaling up to it at level 85? Without it, I'm assuming that Str > Stamina once again for dps per point?

No idea, all of the sims are performed assuming it has. If we find otherwise, I'll revise.

I believe it has, and scales up with each level above 80. At level 85 I have 5029 Stamina, and the tooltip says that this grants 70146 health.
I don't play WoW any more.
Donnan - Nangun - Kysen - Kysen - Mardun - Timkins

Mostly-Book Blog.
KysenMurrin
 
Posts: 6843
Joined: Thu Jun 26, 2008 6:37 am
Location: UK

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby tlitp » Thu Dec 09, 2010 2:58 am

Extermi wrote:Going through the priority simulations, I noticed that inquisition rotations were not that far behind as I thought, and some alternating with ShoR were actually quite good. I was wondering if this could be put "to the extreme":

- During rotation, if Inquisition is down, cast CS else HotR to build up 3 HoPo
- Once 3 HoPo are reached, check if Sacred Duty has procc´ed from judgement. If so cast ShoR, else cast Inq

Implemented analytically in r199. The CPG weaving can be performed in two different ways :
  • cast SotR independently of SD
    Code: Select all
    inq -> (hotr.filler.hotr.j2.hotr.sotr) . (cs.filler.cs.j2.cs.inq) ->
  • cast SotR only on SD, else refresh Inq
    Code: Select all
    inq -> (hotr.filler.hotr.j2.hotr.sotr) . (cs.filler.cs.j2.cs.inq) ->
    inq -> (hotr.filler.hotr.j2.hotr.inq) ->
Across the parameter field (weapon ilvl 346-372, hit/exp to caps, Vengeance), the second model is consistently better.


For the following, I'll denote that weaving model as W2. I guess most people are interested in comparing ISH9/W2 to the basic 9C9.
  • weapon ilvl scaling : 9C9 >> W2 > ISH9
  • hit scaling : 9C9 > W2 > ISH9
  • exp scaling : 9C9 >> W2 > ISH9
  • Vengeance scaling : ISH9 > W2 >> 9C9
Cutting to the chase, stick to 9C9 if the weapon isn't terribly outdated and the hit/exp values are decent. If not, start considering the alternatives - they can increase the total output by up to 6%.
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby theckhd » Thu Dec 09, 2010 3:20 pm

I updated the single-target priority simulations today. I'm now using 4 Prime glyphs, SoT/SotR/CS/HotR. This way both CS and HotR get fair representation. Since none of the queues use both CS and HotR, this shouldn't have any other serious side effects.


It looks like 939 and IS39 retained their values. Fixing the AS/Cons glyphs shuffled some of the other rotations around. The only noteworthy changes were:

  • ISC9 > ISH9 now, because both glyphs are fairly represented. We can start referring to this as a generic "IS39" where you use whichever 3 you have glyphed.
  • In the "sub-20%" category, we have a few new entries. Alternating Inq and SotR seems to be a reasonably significant net gain (Inq>SotR*>CS>HoW>J). However, you can get almost identical performance from HoW>(939), or in other words, making HoW your highest priority and filling the rest of your GCDs with the usual 939 rotation. This gives a stable pattern:

    HoW-SotR-CS-J-HoW-CS-X-CS-

    where X is filled with AS>Cons>HW

After messing with this module, it's clear to me that I need to do a pretty heavy re-write of it. There's a lot of redundancy in it, which makes working with it sort of painful. I have an idea in mind, which I'll start implementing once I get the weapon and enchant sims posted.

I want to do some quick consistency checking on the already-posted results with tlitp's modifications to rotation_model, after that I'll start working on weapons and enchants. Unfortunately, I've got some prep work to do on Theck before this weekend's raids, so I won't have a lot of time to mess with things until the weekend.
"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: 7950
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby theckhd » Fri Dec 10, 2010 9:19 am

Weapon calculation up. I'll try and get to enchants this weekend some time, the rest of today is almost totally booked.
"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: 7950
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Agility as a survivability stat

Postby Digren » Mon Dec 13, 2010 10:44 am

Is agility still 75%ish of the dodge of dodge rating at level 85? Is strength still 25% of the parry of parry rating? I ask because of choices between the stats when selecting survivability gear. For example, consider the discussion below.
Digren wrote:
econ21 wrote:Trinket #1: [Harrison's Insignia of Panache] : Uldum (Quest)

So here's an interesting thought. I finished this quest yesterday, and rather than the above item, I chose [Schnotzz's Medallion of Command].

Strength is a second-tier threat stat, which provides 25% of the parry of parry rating. In other words, a strength/mastery trinket is dual threat/survivability, with poor threat and most of the survivability random. it's the equivalent of:
59 parry rating
some amount of attack power
mastery proc

On the other hand, Agility is a worst threat stat but provides 75% of the dodge of dodge rating. Thus, the agility/mastery trinket is the equivalent of:
175 dodge rating
some amount of crit rating
mastery proc

...and in my opinion that makes it better than the strength trinket for a survivability set.

Am I way off on this? Is agility not this good at 85? I think my ratios are based on level 80 maths.
User avatar
Digren
Moderator
 
Posts: 2139
Joined: Sat Jul 26, 2008 2:41 pm

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby theckhd » Mon Dec 13, 2010 11:24 am

The conversion factors for dodge rating and agility are:
Code: Select all
cnv.dodge_dodge=176.71890258.*(base.lvl==85);
cnv.agi_dodge=304.50769251.*(base.lvl==85);


In other words, Every point of Agility effectively gives you 0.5803 dodge rating before Kings is applied, or 0.6094 rating post-Kings.

Strength converts to parry rating at exactly 4:1, or 0.25 rating per point of Strength. It's floor()-ed in a number of places though, as far as we can tell, so you won't get any benefit from adding 1 point of Strength unless it ticks you over to a new integer. Since agility is a direct Agi->dodge% conversion, it doesn't share this behavior.
"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: 7950
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby tlitp » Mon Dec 13, 2010 6:45 pm

Some code snippets to assess Windwalk (regular 3/9 with CS/SotR, raid set, full time-on-target) :
Code: Select all
ww.pc=gear.swing./60;ww.pd=10;
ww.q=(1-ww.pc.*mdf.mehit).^(ww.pd./player.wswing) ...            %AA
    .*(1-ww.pc.*mdf.mehit).^(ww.pd.*3./(9+1.5.*rot.xtragcd)) ... %CS
    .*(1-ww.pc).^(ww.pd./(9+1.5.*rot.xtragcd)) ...               %SotR
    .*(1-ww.pc.*mdf.rahit).^(ww.pd./(9+1.5.*rot.xtragcd)) ...    %J
    .*(1-ww.pc.*mdf.rahit).^(ww.pd.*P.CS./(9+1.5.*rot.xtragcd)); %AS
ww.p=1-ww.q;
dtrack=[];
for dloop=0:600:1800
tmpavd=avoid_dr(base,gear.dodge+consum.dodge+dloop,gear.parry+consum.parry ...
    +floor((player.str-base.stats.str)./4),player.agi-base.stats.agi);
dtrack=[dtrack tmpavd.dodgedr];
end
ww.tm=[ww.q ww.p    0    0;
       ww.q    0 ww.p    0;
       ww.q    0    0 ww.p;
       ww.q    0    0 ww.p];
ww.dd=ww.tm^100;
ww.upt=ww.dd(1,:);
ww.gupt=1-ww.upt(1,1);
ww.stacks=(ww.upt(2)+2.*ww.upt(3)+3.*ww.upt(4))./ww.gupt;
ww.dodge=0;
for dloop=1:4
    ww.dodge=ww.dodge+ww.upt(dloop).*dtrack(dloop);
end
ww.dodge(1)=ww.dodge-dtrack(1);
ww.dodge(2)=ww.p.*(dtrack(2)-dtrack(1));
[ww.gupt ww.stacks ww.dodge(1) ww.dodge(2)]

Results :
Code: Select all
ph=physical hit (%)
exp=expertise (skill)

              global.uptime(%)  avg.stacks  avg.dodge(%)  avg.dodge(%)
ph=0,exp=0            32.51         1.4308     1.0938        0.7886
ph=8,exp=0            35.39         1.4792     1.2275        0.8585
ph=8,exp=26           39.20         1.5456     1.4153        0.9508


EDIT : Apparently, WW doesn't stack (to three) anymore. The second column of dodge entries account for this behaviour. The uptime (well, the global one) remains unchanged.
Last edited by tlitp on Tue Dec 14, 2010 12:25 pm, edited 2 times in total.
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby tlitp » Tue Dec 14, 2010 2:40 am

And some code snippets to assess Landslide/Avalanche (regular 3/9 with CS/SotR, full time-on-target) :
Code: Select all
ls.pc=gear.swing./60;ls.pd=12;
ls.q=(1-ls.pc.*mdf.mehit).^(ls.pd./player.wswing) ...            %AA
    .*(1-ls.pc.*mdf.mehit).^(ls.pd.*3./(9+1.5.*rot.xtragcd)) ... %CS
    .*(1-ls.pc).^(ls.pd./(9+1.5.*rot.xtragcd)) ...               %SotR
    .*(1-ls.pc.*mdf.rahit).^(ls.pd./(9+1.5.*rot.xtragcd)) ...    %J
    .*(1-ls.pc.*mdf.rahit).^(ls.pd.*P.CS./(9+1.5.*rot.xtragcd)); %AS
ls.p=1-ls.q;
ls.base=rot.totdps;
gear.ap=gear.ap+1000;stat_model;ability_model;rotation_model;
ls.proc=(rot.totdps-ls.base).*ls.p;
[ls.p ls.proc]

av.ppc=5.*gear.swing./60;av.spc=0.1;
av.dpp=500.*mdf.spdmg.*mdf.spcrit.*target.resrdx;
av.ptps=mdf.mehit./player.wswing ...        %AA
    +3.*mdf.mehit./(9+1.5.*rot.xtragcd) ... %CS
    +1./(9+1.5.*rot.xtragcd) ...            %SotR
    +mdf.rahit./(9+1.5.*rot.xtragcd) ...    %J
    +P.CS.*mdf.rahit./(9+1.5.*rot.xtragcd); %AS
av.stps=1./cens.NetTick ...                                           %Cens
    +max([0.75-P.CS;rot.val.zeros]).*mdf.sphit./(9+1.5.*rot.xtragcd); %HW
av.pps=av.ptps.*av.ppc+av.stps.*av.spc;
av.dps=av.dpp.*av.pps;
[av.dpp av.pps av.dps]

Unlike LS, Av doesn't scale with weapon ilvl. As such, let's see if it can best its counterpart using the pre-raid gear set :
Code: Select all
ph=physical hit (%)
exp=expertise (skill)

              LS.uptime(%)  LS.dps  Av.procspersecond  Av.dps
ph=0,exp=0        37.88        176             0.2313     135
ph=8,exp=0        41.08        209             0.2526     148
ph=8,exp=26       45.26        255             0.2820     165
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby econ21 » Tue Dec 14, 2010 8:20 am

With the weapon comparison, it would be nice to have the Dragonmaw mace/Mace of the Gullet added - as the 333 weapon you get while questing in TH, it is the default starter weapon for new tanks.

The Tol Barad weapon looks pretty good on your numbers and is a nice "failsafe": not too hard to get at revered with 12 dailies to farm.

I'd be interested to see how the new PvP weapons shape up. The weapon is the one piece of PVP gear I might want to farm to use in PvE.
econ21
 
Posts: 1378
Joined: Tue Apr 01, 2008 8:53 am

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby Rinion » Tue Dec 14, 2010 10:11 am

Is this the correct place to post this even? I hope so.

Theck, using the two gear lists you have, what plans are there to get around the low hit/expertise that you end up with? Or is it a problem even?
Rinion
 
Posts: 3
Joined: Tue Jan 19, 2010 2:11 am

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby theckhd » Tue Dec 14, 2010 11:11 am

tlitp wrote:Some code snippets to assess Windwalk (regular 3/9 with CS/SotR, raid set, full time-on-target)

I was under the impression that Wind Walk does not stack. Towellie's parses have expired, but my two haven't, and both show Windwalk "refreshing" rather than stacking.

I was planning on writing an "enchant_model" module this week. If you've already started it, let me know so we can avoid duplicating work. I'll plan on using a lot of the code you posted, since it's essentially exactly what I was planning on writing.
"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: 7950
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby theckhd » Tue Dec 14, 2010 11:18 am

econ21 wrote:With the weapon comparison, it would be nice to have the Dragonmaw mace/Mace of the Gullet added - as the 333 weapon you get while questing in TH, it is the default starter weapon for new tanks.

True, but it should be pretty obvious that as a 2.6-speed weapon it's going to fall at the bottom of the list next to Ravening Slicer. Without fail, 13 ilevels makes a bigger difference than itemization. Plus, once the next tier of content is available, we'll probably be dropping all of these blue items anyway.

Rinion wrote:Theck, using the two gear lists you have, what plans are there to get around the low hit/expertise that you end up with? Or is it a problem even?

Well, the awesome thing about being an imaginary gear set on an imaginary tank soloing imaginary bosses in MATLAB is that you never lose aggro.

There's really no issue with the gear set being low on hit/exp from a computational standpoint. In my experience in raids so far, there's not much wrong with it in practice either. Hit and expertise don't matter so much when you're dumping all of your Holy Power into Word of Glory anyway.
"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: 7950
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: Theck's MATLAB thread - Cataclysm/4.x

Postby tlitp » Tue Dec 14, 2010 12:33 pm

theckhd wrote:I was under the impression that Wind Walk does not stack. Towellie's parses have expired, but my two haven't, and both show Windwalk "refreshing" rather than stacking.

I was planning on writing an "enchant_model" module this week. If you've already started it, let me know so we can avoid duplicating work. I'll plan on using a lot of the code you posted, since it's essentially exactly what I was planning on writing.

I. Uh, WW doesn't stack anymore ? On beta, the stacking applications were also true aura refresh events (e.g. : one stack (5 sec left)->proc->two stacks (10 sec left)). Either way, the code alterations are minimal (refer to the initial post); the uptime is the same, all that's changed is the effect - yes, it makes WW even crappier than before. :P
II. Yeah, go right ahead if you're interested in "pikturz". I merely wanted to sketch the code framework.
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

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