The Bleeding Edge of TPS

Warning: Theorycraft inside.

Moderators: Fridmarr, Worldie, Aergis, theckhd

The Bleeding Edge of TPS

Postby Meloree » Wed Jul 14, 2010 3:30 pm

Farm content is boring, and one of the ways I tend to spice it up for myself and my healers is to push the bleeding edge of tank TPS and DPS. In doing so, it's often useful to think a little bit outside the box. I've been working with tiltp on maximizing output - I am not a mathgod of the calibre of tiltp or Theck - and he's been running MATLAB sims based on the rather extreme values of DPS stats available even to a tank, and some of the conclusions are quite remarkable, although I'll leave him to fill in the numbers.

The first step, for me, in setting up to chase parses is to set up my constraints.

Constraint the first: Uncrittable is not negotiable. I'm still a tank.
Constraint the second: 30k armor, 35k health, minimum unbuffed (for ICC25H) as an effective health floor. Given healing output, this may be higher than is strictly necessary with a 25 or 30% buff, but given some of the sacrifices we're about to make in the spec, it's reasonable.
Constraint the third: Hysteria, Tricks (after the first), and Power Infusion are off limits. They're just better used on other raid members, and sacrificing RDPS for PDPS is against my moral code. The possible exception is at Festergut with 9 stacks - at that point, I might make best-use of tricks.

Beyond those constraints, everything is in play: gear, spec, glyphs, rotation, everything.

When tiltp kindly volunteered to help me out with pushing maximum tank dps, those were the constraints I asked him to adhere to. Prior to tiltp's work, the spec that I'd been using was this one which was already very aggressively oriented towards threat. The gearset was this, which was certainly survivable enough. Before some prot/ret dps guys started taking some of the number 1 parses, I was pretty satisfied, and elected not to regem for further DPS.

The MATLAB sim confirmed some things, and opened my eyes to a few others that I hadn't thought about. The first is that Reckoning (even the final, worst, 5th point) was the best DPS talent available from the optional points - and not just by a little bit. Better than Seals of the Pure or Crusade. This is, of course, only true at such extreme levels of hit/expertise/crit and AP, Theck's simmed values are much more accurate with standard tanking gear. But it opened my eyes as to how much values would be shifting with such abnormal offensive stats. The next, somewhat intuitive, point, is that anticipation is a dps/tps loss. Speccing 5/5 divinity instead of 5/5 anticipation increases expected reckoning uptime - and when you're white-swinging for an expected damage of nearly 6k (seal/white hit, crit included), every reckoning proc is worth a LOT of damage. Similarily, 1/2 JotP increases boss attack speed, which increases expected reckoning uptime.

The next thing that was worth noting was that certain rotational substitutions were significant DPS upgrades. First, that Avenger's Shield is powerful enough to be substituted for any 9 or 6 ability in the rotation EXCEPT Hammer of the Righteous, and should not be delayed for anything except HotR. This lead to somewhat random substitutions which were a little difficult to get a handle on, but I'll execute better next week. The second change was that HoW should replace Holy shield on 100% of the HS opportunities below 20% boss health. HoW is not strong enough to displace Judgement or Consecration.

Another important point was DPS enchants and gems. Given the values of Strength, AP, Crit, Agility, and Haste with the gearset I had, it turned out that several enchants were better DPS trades than trading stamina gems into strength gems. Acanum of Torment, Enchant Chest - Powerful Stats, and Enchant Gloves - Crusher all turned out to be worth more "DPS/STAM" than gem swaps. After that, it was a matter of trading gems down to my stamina floor of 35k. This lead me into this spec and gearset.

The results were reasonable but can certainly be improved upon. In part, using a new "decoupled queue" rotation, rather than 969, caused me to fumble and lose a tenth of a second here and there, I could certainly increase my output without changing anything else. Also, focussing on running a different rotation lead me to make some mistakes that I wouldn't normally, leading to a death on Saurfang in a situation where I should have had a cooldown up.

On the plus side, I managed to pick up Bloodvenom Blade (277) this week, which is significantly better for DPS than Havoc's Call (271) - again, not expected by the normal weapon rankings, but at these high levels of crit and expected attack speed (high reckoning uptime) the weapon DPS becomes much more significant. It also allowed me to wear a Winding Sheet 277 which I had in the bank, while still maintaining the hitcap - and because of the human racial it netted me another 3 expertise, placing me very nearly at hardcap. Next weeks gearset. I expect that next weeks results will be significantly more impressive, even normalizing out the new buff level that we expect to hit on Tuesday.

So, I promised tiltp a post, and he certainly earned it for all his help - I'm looking forward to pushing the limits a little bit further. It would also be great if we could get a few more people involved with running extreme gearsets for extreme tank DPS - and TPS, but when you're doing 25k+ steady-state TPS, it's really not something you have to optimize for. If we've accomplished nothing else, we've established that the MATLAB TPS model remains valid at very extreme gear levels, but that the specific valuations based on a specific stat levels might need to be taken with a grain of salt sometimes. Either way, I'm looking forward to next week, now, and that's pretty rare for me outside of progression.
Meloree
Maintankadonor
 
Posts: 1420
Joined: Wed Mar 12, 2008 10:15 am

Re: The Bleeding Edge of TPS

Postby Rhiannon » Wed Jul 14, 2010 6:02 pm

One thing I'm curious about, how does Tiny Abom stack up in this scenario? It's hands down best trinket available for ret, I'd have thought perhaps following the same reasoning that reckoning increases in value it would as well.
Rhiannon
 
Posts: 1061
Joined: Mon Mar 03, 2008 8:17 am

Re: The Bleeding Edge of TPS

Postby Meloree » Wed Jul 14, 2010 6:33 pm

Rhiannon wrote:One thing I'm curious about, how does Tiny Abom stack up in this scenario? It's hands down best trinket available for ret, I'd have thought perhaps following the same reasoning that reckoning increases in value it would as well.


Without detailed math, I'm pretty certain it's better than DMC:G. I just don't have one.
Meloree
Maintankadonor
 
Posts: 1420
Joined: Wed Mar 12, 2008 10:15 am

Re: The Bleeding Edge of TPS

Postby inthedrops » Thu Jul 15, 2010 7:21 am

Nice thread Meloree. I would love to be able to participate. But since my guild hasn't killed heroic LK 25 yet, it would be a bad idea to do this on any 25 man content. Once we kill LK I'm sure the guild will be open to this kind of thing.

I will study this further and try it in some 10 mans and see what I learn. I haven't had the time lately to do anything but raid 25's....
inthedrops
Maintankadonor
 
Posts: 1281
Joined: Mon Oct 29, 2007 9:19 am

Re: The Bleeding Edge of TPS

Postby theckhd » Thu Jul 15, 2010 8:30 am

Reckoning's pretty linear per point with a single attacker, it's only when confronted with many attackers that you see significant diminishing returns on the last few points. I'm a little surprised that Reckoning overtook SotP and Crusade per point, I would have expected the break point to be higher than it apparently is.

I assume the "decoupled queue" you're talking about is just alternating between two priority queues? I.e. x6x6x6? I'd imagine that even with AS (glyphed?) being better than ShoR, you'd still end up with it in an x slot every time since it's a 30-second cooldown, leaving the 6's portion of 969 unchanged.
"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: The Bleeding Edge of TPS

Postby Meloree » Thu Jul 15, 2010 9:08 am

theckhd wrote:Reckoning's pretty linear per point with a single attacker, it's only when confronted with many attackers that you see significant diminishing returns on the last few points. I'm a little surprised that Reckoning overtook SotP and Crusade per point, I would have expected the break point to be higher than it apparently is.

I assume the "decoupled queue" you're talking about is just alternating between two priority queues? I.e. x6x6x6? I'd imagine that even with AS (glyphed?) being better than ShoR, you'd still end up with it in an x slot every time since it's a 30-second cooldown, leaving the 6's portion of 969 unchanged.


"decoupled queue" was tiltp's term. I think because if AS happens to come off cooldown where a ShoR should have gone, AS ends up replacing ShoR instead (and will end up replacing it for the rest of the fight). If AS manages to hit on a 9, it will replace a different 9s ability each time, but it'll always be a 9 for the rest of the fight. AS is so much damage per cast time that extending it's cooldown by 20% to always replace an HS cast is more damage lost than replacing anything but HotR.
Meloree
Maintankadonor
 
Posts: 1420
Joined: Wed Mar 12, 2008 10:15 am

Re: The Bleeding Edge of TPS

Postby Kelaan » Thu Jul 15, 2010 10:13 am

Meloree wrote:if AS happens to come off cooldown where a ShoR should have gone, AS ends up replacing ShoR instead (and will end up replacing it for the rest of the fight). If AS manages to hit on a 9, it will replace a different 9s ability each time, but it'll always be a 9 for the rest of the fight. AS is so much damage per cast time that extending it's cooldown by 20% to always replace an HS cast is more damage lost than replacing anything but HotR.


That seems like an excessive dichotomy. If you replace a '6' skill (ShoR), doesn't that screw up your 6's timing? If you delay AS not until a holy shield but just until your next '9' slot (so replacing judge/cons/HS), wouldn't that still be good? How does that compare?
User avatar
Kelaan
 
Posts: 4037
Joined: Thu Jan 03, 2008 12:01 pm

Re: The Bleeding Edge of TPS

Postby Meloree » Thu Jul 15, 2010 10:38 am

Kelaan wrote:
Meloree wrote:if AS happens to come off cooldown where a ShoR should have gone, AS ends up replacing ShoR instead (and will end up replacing it for the rest of the fight). If AS manages to hit on a 9, it will replace a different 9s ability each time, but it'll always be a 9 for the rest of the fight. AS is so much damage per cast time that extending it's cooldown by 20% to always replace an HS cast is more damage lost than replacing anything but HotR.


That seems like an excessive dichotomy. If you replace a '6' skill (ShoR), doesn't that screw up your 6's timing? If you delay AS not until a holy shield but just until your next '9' slot (so replacing judge/cons/HS), wouldn't that still be good? How does that compare?


If you replace a 6 skill, it doesn't affect 6s timing at all. HotR, Judge, AS, HS, HotR (replacing 6) works just fine for that segment of the rotation. At this gear level, ShoR has roughly the same expected damage as Judge or Consecration, and is only barely ahead of Holy Shield - largely because we're in DR range for BV (or overcapped with procs) all the time, where Judge and Cons just keep on scaling, so there's no reason to replace a 9 instead of that particular 6.
Meloree
Maintankadonor
 
Posts: 1420
Joined: Wed Mar 12, 2008 10:15 am

Re: The Bleeding Edge of TPS

Postby theckhd » Thu Jul 15, 2010 6:08 pm

Meloree wrote:
Theck wrote:I assume the "decoupled queue" you're talking about is just alternating between two priority queues? I.e. x6x6x6? I'd imagine that even with AS (glyphed?) being better than ShoR, you'd still end up with it in an x slot every time since it's a 30-second cooldown, leaving the 6's portion of 969 unchanged.


"decoupled queue" was tiltp's term. I think because if AS happens to come off cooldown where a ShoR should have gone, AS ends up replacing ShoR instead (and will end up replacing it for the rest of the fight). If AS manages to hit on a 9, it will replace a different 9s ability each time, but it'll always be a 9 for the rest of the fight. AS is so much damage per cast time that extending it's cooldown by 20% to always replace an HS cast is more damage lost than replacing anything but HotR.


Well, yeah, but if you push back ShoR it will come back to haunt you later, because you'll run into clashes. So while AS is more DPCT than ShoR, I'm pretty sure it would be a net TPS gain to push it back into an X slot the first time, which would ensure that it stays in that slot.

"Decoupled queue" is a term I've used before to describe 969. It's basically two different priority queues which you ping-pong between. One queue is just HotR/ShoR, the other is "everything else", which in your case is AS>Judge>Cons>HoW>(HolyWrath?)>HS. It's a natural outgrowth of the fact that HotR/ShoR are both 6-second cooldowns (and both high-priority casts). Another way to describe it is that the 6-second queue is orthogonal to the 9-second one in a traditional 969, so the two never interfere with one another (you never have the option to cast a 6 in a 9 slot, or vice versa).

That's why I wanted to clarify - AS replacing ShoR isn't a decoupled queue, but AS in a 9's slot is.
"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: The Bleeding Edge of TPS

Postby Meloree » Thu Jul 15, 2010 7:31 pm

theckhd wrote:Well, yeah, but if you push back ShoR it will come back to haunt you later, because you'll run into clashes. So while AS is more DPCT than ShoR, I'm pretty sure it would be a net TPS gain to push it back into an X slot the first time, which would ensure that it stays in that slot.


I don't understand. If AS comes off cooldown such that it should replace a ShoR, then I replace the ShoR with AS, and hit whatever 9 is off cooldown. Then HotR, and rotation continues unaffected. When AS comes off cooldown again, it's going to be replacing a ShoR again, because of the way the cooldowns line up, but there's no clash that I can see.

But I've probably missed something fundamental.
Meloree
Maintankadonor
 
Posts: 1420
Joined: Wed Mar 12, 2008 10:15 am

Re: The Bleeding Edge of TPS

Postby theckhd » Thu Jul 15, 2010 8:49 pm

Ah, I see. That's using the following decoupled queues:

1) HotR>AS>ShoR
2) Everything Else

My point is that HotR, AS, and ShoR are all much higher DPCT than "Everything Else." So if you stick to these queues, you're substituting AS (high damage) for ShoR (high damage). If you used the queues I posted:

1) HotR>ShoR
2) AS>Judge>Else

Then you'll push AS back one GCD, but you'll see larger gains from the substitution over the course of the rotation, because now you're substituting AS (high damage) for something else (low damage).

The other option would be to push off ShoR into a 9's slot. That couples the queues, which leads to clashes.
"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: The Bleeding Edge of TPS

Postby tlitp » Thu Jul 15, 2010 10:10 pm

Meloree wrote:Similarly, 1/2 JotP increases boss attack speed, which increases expected reckoning uptime.

Although the intention is transparent, I'll add a clarification : lower swing timer => higher attack rate => inflation of all reactive damage sources (parryhaste/reckoning/HS/auras).

theckhd wrote:I'm a little surprised that Reckoning overtook SotP and Crusade per point, I would have expected the break point to be higher than it apparently is.

Overtaking SotP isn't an extraordinary feat; coming within 80% of 6%C is. Also, Mel's setup pushes Reckoning neck-to-neck with DivStr/TbtL.

theckhd wrote:My point is that HotR, AS, and ShoR are all much higher DPCT than "Everything Else".

ShoR isn't, as far as Mel is concerned. There are almost no delayed queues, only straight substitutions - the exception being HotR. You'll probably follow the logic more easily with a color scheme :
696 => 6 | 6 (AS activates before a 9, or before HotR)
696 => | 96 (AS activates before ShoR)
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

Re: The Bleeding Edge of TPS

Postby Meloree » Thu Jul 15, 2010 11:10 pm

theckhd wrote:My point is that HotR, AS, and ShoR are all much higher DPCT than "Everything Else." So if you stick to these queues, you're substituting AS (high damage) for ShoR (high damage). If you used the queues I posted:


From the parse on Marrowgar, average ability hit damage:
HotR Damage 7435.
ShoR Damage 5086
Seal Proc Damage: 1386
Judgement 5472
Consecration 1021 (per tick)
Holy Shield 2252 (per proc)
Avenger's Shield 8545
Hammer of Wrath 5016

Expected crit rate ~38%. ShoR gives an expected damage per cast (including seal proc) of ~9005, ahead of Consecration, and behind Judgement.
As far as expected damage per cast goes, the list is HotR > AS > Judgement > ShoR > Consecration > HoW > HS. I'm not even entirely convinced that HS won't beat HoW for expected damage per cast on, say, Frenzied Saurfang.

ShoR ain't no big thang no more. So, the queues in question boil down to (mostly):
1) HotR > AS > ShoR
2) AS > Judge > Cons > HoW > HS
Meloree
Maintankadonor
 
Posts: 1420
Joined: Wed Mar 12, 2008 10:15 am

Re: The Bleeding Edge of TPS

Postby tlitp » Fri Jul 16, 2010 6:31 am

Meloree wrote:From the parse on Marrowgar, average ability hit damage: (...)

I'm not even entirely convinced that HS won't beat HoW for expected damage per cast on, say, Frenzied Saurfang.

I. Speaking with others mostly, as this part Mel and I already covered up : some abilities are more stochastic than others. AS/HoW have a base damage range, not a fixed value; J depends on the number of stacks (5-stacks doesn't have 100% uptime); HS is as stochastic as blocking itself.
Moreover, the expected output varies as a function of the dynamic variables - both raid environment (again, inconstant uptimes) and gear-based effects (trinkets/weapon enchant).

II. Agreed, there are circumstances in which HoW probably shouldn't substitute HS. But HS is too much of a gamble, I'd say - the theoretical output lead should be high enough as to back up the cases when RNG doesn't work for you.


EDIT : grammar oversight, du darfst mein Bestrafer sein.
Last edited by tlitp on Fri Jul 16, 2010 7:05 am, edited 1 time in total.
User avatar
tlitp
 
Posts: 556
Joined: Mon Jul 27, 2009 3:25 pm

Re: The Bleeding Edge of TPS

Postby theckhd » Fri Jul 16, 2010 6:52 am

Ah, I see my mistake. I had (in)conveniently overlooked the fact that ShoR stops scaling. I could be coy and claim that I thought that since you were using a lot of ret gear, you wouldn't be stacking BV and thus wouldn't have hit the serious DR region. But in truth, I simply forgot. :P

Once ShoR stops being a heavy-hitter, I'm not even sure it makes sense to stick to decoupled 6 and 9 queues. A single HotR>AS>Judge>Shor>Cons FCFS might actually compete with a decoupled queue. The decoupling comes out naturally from this sort of FCFS when HotR and ShoR are the top two abilities. But pushing ShoR that low likely removes the cute orthogonality.

In terms that tlitp would appreciate, ShoR dropping lower in the FCFS queue creates a different basis set of ability eigenvectors. :P

I assume that tlitp has already run the calculation to compare a single FCFS queue to decoupled queues though? If not, shame on him - the code is still written and on the repository from when I wrote the FCFS article!
"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

Next

Return to Advanced Theorycraft and Calculations

Who is online

Users browsing this forum: Google [Bot] and 1 guest

Who is online

In total there are 2 users online :: 1 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: Google [Bot] and 1 guest