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

[Jonesy]

SoR always hits when the carrier attack hits.

[theckhd]

What do the results on S/JoB look like if you use Broken Promise, compared to SoV with Last Laugh? The slower speed should lead to higher judgements, if I understand the mechanics correctly, and the larger SoB hits might make the relative contribution of the HotR swings higher. It also tends to make slightly better use of reckoning procs.

I can run the numbers, but I'd like to wait a little bit until we're sure we've ironed out all the errors in the code first.

Next time you post graphs about DPS of specs could you please make diferent graphs for different boss attack speeds and weapon speeds?

Is your GCD modeled for a fixed 1.5s for everything or fixed only for HotR and ShotR if it's modeled accounting haste already could you post the data on rotations for BL range assuming every spell?(avengers, hammer of wrath, exo, holy wrath).

Boss attack speed isn't likely to make a huge difference. It will increase Holy Shield dps, which will be constant across configurations, and increase reckoning uptime somewhat. I suspect weapon speed will make a bigger difference because of SoB's scaling. Again though, I'll address this once we're more certain the code is correct.

The GCD is modeled for a fixed 1.5s for all of the cooldown-based abilities, since haste doesn't buy you anything in the 969 rotation. The only place I can see this making a difference is the FCFS simulation, which I can modify to incorporate the effects of haste.

I seem to remember reading that once you have Last Laugh, HoW and AS are always less DPS than HotR, so they shouldn't be substituted into the rotation. I'll take a look at it though.

[theckhd]

SoR always hits when the carrier attack hits.

OK, I'll move the boss-avoidance factor into the proc and DPS terms then, for consistency.

[majiben]

Judgement GCD is not affected by haste either.

[theckhd]

Judgement GCD is not affected by haste either.

GCD? or cooldown? If Judgement triggers a full 1.5s GCD regardless of haste, that's extremely strange. Why does it act differently from every other spell?

[majiben]

It's a melee attack. I has a parry and dodge suppression built in. I hope you're not modeling it on a 17% miss.

[theckhd]

Also, I finally got around to adding in the bonus effect of Libram of Obstruction. I haven't uploaded the files just yet, since I'm still tinkering with my local copy, but they'll be updated a little later today.

Here's the approach I took:
The formula for ShoR damage can be simplified to
(1.3*X + 520)*A,
where A is a factor that handles everything else (crits, buffs, resistance, etc), and X is block value after Redoubt and the Meta gem are factored in.

Since (1.3*(X+Y)+520)*A = (1.3*X + 520)*A + 1.3*Y*A, and block value is additive, it makes sense to separate out the Libram effect as an independent "bonus" ability. In other words, the code now looks like:

ShoR = (1.3*X + 520)*A
ShoR_LoO_Bonus = (1.3*Y)*A

The actual code is

Code: Select all

ShoR_LoO_Bonus = (1.3.*352.*(Redoubt + BV_Meta)).*
    (100+(melee_crit + HotC).*Crit_Meta-boss_miss-boss_mcrit_supp)./100 .*
    (OHWS.*Crusade.*GoSU).*
    (CoE.*SanctRet).*
    boss_resist_reduce.*
    LoO_equipped;  %0 if LoO not equipped, 1 if it is

Approaching it this way allows maximum flexibility in testing rotations. For example, in 969, you always get the bonus for 2 out of every 3 ShoR, so I can simply add the following term to that rotation file:

Code: Select all

    ShoR_LoO_Bonus./9 + ...             %Libram bonus

I can do similarly for the more complicated rotations. For example, Judgement replacement drops LoO to 1 bonus every 3 ShoRs, so it only gets ShoR_LoO_Bonus./18.

We'll see if this narrows the gap between 969 and the Exorcism-substitution rotations.

[theckhd]

It's a melee attack. I has a parry and dodge suppression built in. I hope you're not modeling it on a 17% miss.

No, it's modeled as a melee attack. I just never realized that special melee attacks trigger a full 1.5s GCD.

Good thing you mentioned that before I started messing around with the FCFS simulations, that will significantly change the way I have to model it.

[Jonesy]

I'm not wholly sure that Judgement *doesn'* trigger a hasted GCD. Plenty of our other abilities do, at least.

[Adanel]

Boss attack speed isn't likely to make a huge difference. It will increase Holy Shield dps, which will be constant across configurations, and increase reckoning uptime somewhat. I suspect weapon speed will make a bigger difference because of SoB's scaling. Again though, I'll address this once we're more certain the code is correct.

The GCD is modeled for a fixed 1.5s for all of the cooldown-based abilities, since haste doesn't buy you anything in the 969 rotation. The only place I can see this making a difference is the FCFS simulation, which I can modify to incorporate the effects of haste.

I seem to remember reading that once you have Last Laugh, HoW and AS are always less DPS than HotR, so they shouldn't be substituted into the rotation. I'll take a look at it though.

Yeah the boss speed i don't believe it makes a big difference between specs, holy shield doesn't crit so the DPS we get from reckoning and maybe SoB from crits would almost make up for that when thinking SotP vs Conviction, the weapon speed to get an idea about reckoning would be the most interesting.

My main interest on the GCD is at the BL range that is sometimes the most important part of a fight, there the GCD for spells will be at ~1.1s so i believe the impact for a FCFS would be noticiable. You said HotR is more DPS... it hits harder for each move but at BL clearly not more than it costs in time for sure, my main concern with this is the impact of trying to always use the skills with more dmg/gcd time leading me to run out of fillers in my rotation.

And judgement is 1.5, forgot that , you can test with /run hooksecurefunc("CooldownFrame_SetTimer", function(_, start, duration, enable) if start > 0 and enable > 0 then DEFAULT_CHAT_FRAME:AddMessage(duration) end end) from EJ, AS and HoW too, HS Cons HW Exo aren't fixed.

[theckhd]

While we discuss hasted GCDs, here's a little diversion. With the updated SoV formula and the LoO bonuses added, the rotation analysis has gotten a lot more interesting. I checked the values of the abilities, and ShoR is hitting for 4393 with the default talent/glyph setup, and while the LoO bonus is 892. Which seems about in line with what I'd expect.

First, the DPS bar plot (threat plot here):


Judgement substitution gets hit hard by loss of LoO uptime. I was actually surprised by how much it suffers.

Now, lets see how spec affects things (threat):

The gap is far smaller now. 969 makes some gains across the board from LoO uptime, and now I'd be hard-pressed to say the Consecration replacement is even preferable, since it requires an extra glyph and doesn't outperform 969 by much with the 0/5x/18c spec. Judgement substitution has completely fallen behind, as well.

Exorcism on cooldown has lost some ground as well, while HS replacement is still king for DPS.

And finally, the spec graph where we only allow for 2 DPS glyphs:

969 has basically caught up in this plot. The only specs that prop up the Exorcism-on-cooldown rotation are ones that take 2+ points in Sanctity of Battle. But for anything else, 969 is within 10 DPS, and has the benefit of 100% HS uptime. Nothing else catches up.

So, provided this last graph is correct, the TLDR summary is:
1) 969 is still going to be the best tanking rotation for bosses that melee.
2) Exorcism is for Ret paladins, off-DPSing, and replacing Holy Shield on bosses that don't melee
3) Never drop Judgement from your rotation, as long as you're using LoO.

[theckhd]

Just to make sure I haven't missed anything, I'm making a quick list of "requests," or things people would like to see modeled.

• Last Laugh vs. Broken Promise, SoV/SoB comparison
• Variation of DPS with weapon speed, possibly across specs
• Boss attack speed / Reckoning uptime effects
• Revisit FCFS with haste/bloodlust effects
• 13/52/6 dropping Redoubt or CE
• 13/53/5 with new FCFS code

Did I miss anything?

[majiben]

Lose of 1 point in TbtL, Redoubt, and CE. This is to see the full effects of 13/52/6 specs. I suppose a 13/53/5 is also a possiblity but the lose of LoO uptime and CD conflicts would likely shoot that one before it ever left the gate.

[theckhd]

Lose of 1 point in TbtL, Redoubt, and CE. This is to see the full effects of 13/52/6 specs. I suppose a 13/53/5 is also a possiblity but the lose of LoO uptime and CD conflicts would likely shoot that one before it ever left the gate.

I've already tested 13/52/6 with 2/3 TbtL here. I can repeat it for the other two versions easily enough, though.


13/53/5 would have to switch to a FCFS system to even have a chance at being competitive, since it breaks 969. I already took a look at that here. I can revisit it again to incorporate the new formulas once I get around to revamping the FCFS code to incorporate haste.

[Conaan!]

one request i have is for a acronym page, way to many acronyms makes it hard to understand alot of this

like LoO, i kept thinking you were talking about a toilet, or a talent, until i remembered my libram