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

[jere]

I think the fast weapons would be the ones we would want to typically model for tanking since we have so few slow ones. Only broken promise comes to mind, and it is becoming quite a few item levels behind in defensive stats to the point where we might as well use a slow DPS weapon for the comparison rather than it.

I think the difference from back in the brutallis days was it was modeled completely via a simulation results. I think Theck is using an equation for modeling Reckoning in the matlab sims. That would be my first guess at the differences.

[theckhd]

For the simulation, I used a 1.6 speed weapon (Shiver). Reckoning does get stronger with slower weapons, which is why it was actually half decent with Broken Promise for SoB tanking. But since all of the newer tanking weapons are fast, and we're about to lose SoB anyway, I don't expect that a slow-speed simulation will have much relevance in 3.2.

The code for reckoning is in stats_recalc.m:

Code: Select all

%% Reckoning
%note that the way to weave this in to DPS calculations is to multiply
%player_swing_speed by Reck in the ability_recalc module, since Reckoning's
%proc doesn't help all skills equally
Reck=1-(1-0.02.*Reck_points.*(1-player_avoid)).^ ...
        (min([8.*ones(size(player_swing_speed)); 4.*player_swing_speed]) ...
        ./boss_swing_speed);
%% end


It gets factored into abilities like this:

Code: Select all

SoV_Melee_DPS = Seal_of_Vengeance.*(100-boss_avoid)./100./(player_swing_speed./(1+Reck));

Ignore the (1-boss_avoid) term, which represents lost procs due to missing with your melee swing. Effectively, you're reducing your melee swing speed by a factor of (1+Reck).

I'm fairly certain that Dorvan was the one who came up with the formula, much earlier in this thread or the one that spawned it.

[knaughty]

Ah well. I was hopeful Reckoning would be viable again - it was great fun back in Sunwell, especially for things like Brute.

Hang on...

Part of the reason it was so OP at Brute was his ridiculous swing speed. I assume the sim uses a 2.0 debuffed to 2.4 speed boss?

How would reckoning be versus, say, Algalon:

Dual-wield boss, 1.0 -> 1.2 swing speed. IE: Four times the base attack rate. Of course, there's usually a fairly substantial increase in miss rate, but even so, you'd be looking at double. That about doubles Reck's effectiveness? Late era TBC was full of DW-demon bosses.

This was probably why we were seeing much heavier non-linearity on Reckoning - we were modelling a boss that provokes a much higher uptime, and proc overlap becomes a serious issue.

[Amaranthea]

AFAIK, the 3/5 number came from a simulation that modeled an increasing number of attackers on 2.0 attack speed. Even in BC it showed up as fairly linear for the case of one attacker.

viewtopic.php?f=6&t=5124 is the old thread.

[knaughty]

Ah, might have been from Hyjal AE tanking perspective. Right "era".

[Jasari]

Theck, I was going through your code to add to the web app I'm developing to pull armory data and give customized threat information and I can't find the spell power bonus of Totem of Wrath added in anywhere. Not that it'd make much of a difference to the calculations, I'm just wondering if it was an oversight of if it's in there and I'm not just seeing it.

[theckhd]

Theck, I was going through your code to add to the web app I'm developing to pull armory data and give customized threat information and I can't find the spell power bonus of Totem of Wrath added in anywhere. Not that it'd make much of a difference to the calculations, I'm just wondering if it was an oversight of if it's in there and I'm not just seeing it.

Looks like it was overlooked. I just uploaded updated buffs_debuffs and stats_recalc modules that include it. The new variables are ToW_up (boolean) and ToW (0 or 280).

[majiben]

I read reports off EJ about SotP affecting the melee strike of SoV and I can confirm that it not only updates the tool tip from 33% to 37-38% and the damage done as well on the current PTR. This will shift things around some.

[theckhd]

I read reports off EJ about SotP affecting the melee strike of SoV and I can confirm that it not only updates the tool tip from 33% to 37-38% and the damage done as well on the current PTR. This will shift things around som.

It will likely bring SotP back up to the levels seen in this post. I'll try and find some time to update the code today, though no promises since this week's been very busy.

[Jasari]

I think I found a typo in the boss armor calculation

Code: Select all

boss_armor=boss_base_armor.*(1-0.2.*Sund_up-0.5.*CoR_up).*(1-ArPen./ArPen_rating_to_ArPen./100);

The 0.5 for Curse of Recklessness (now Curse of Weakness) should probably be 0.05

[Saeverud]

Probably a stupid question as I just found this thread today, but in regards to the weapon enchant post, does that refer to berserking as the highest dps/tps if you are hit capped?

Edit: Nevermind. I completely blanked on the armor drop from it, making it completely useless for tanks. Then I assume Superior Potency would be the best if you're hit capped.

[Dread]

Probably a stupid question as I just found this thread today, but in regards to the weapon enchant post, does that refer to berserking as the highest dps/tps if you are hit capped?

Edit: Nevermind. I completely blanked on the armor drop from it, making it completely useless for tanks. Then I assume Superior Potency would be the best if you're hit capped.

You're correct, but if you're at that point, I'd suggest using Potency over Superior Potency. You lose 1.0 DPS/1.2 TPS, but you gain the additional block value from the Strength (for EH benefits).

[theckhd]

<edit>Update: Be wary of the conclusions of this analysis, as it is only valid for a "regular" mob. For Demons, Undead, Humanoids, and Elementals, Crusade is twice as effective, which tips the balance. See this post for more careful analysis.

Now that Seals of the Pure is properly affecting our 5-stack SoV proc, we need to revisit the talent calculation once more. The code is now also correctly handling the DR on ShoR, but the gear set is an avoidance/EH set that doesn't quite hit the DR range, so it won't really manifest itself in these calculations. Also, I've fixed the boss armor typo that Jasari noticed, resulting in slightly lower numbers across the board.

First, let's look at the TPS per talent point graph. I've slightly changed the specs uses, which should more accurately reflect each talent.

Here are the possible specs we're comparing:

1. no threat talents
2. 3/3 One-Handed Weapon Specialization (OHWS)
3. OHWS + 3/3 Touched By The Light (TBTL)
4. OHWS+TBTL + 3/3 Crusade
5. OHWS+TBTL + 5/5 Conviction
6. OHWS+TBTL + 5/5 Seals of the Pure (SotP)
7. OHWS+TBTL + 5/5 Reckoning
8. OHWS+TBTL+SotP + 5/5 Reckoning

We simulate the DPS and TPS for each of these talent configs, and calculate the value of each talent point through simple subtraction and division. For example, to evaluate Crusade, we subtract spec 2 from spec 3, and divide the result by 3 (the number of points in Crusade). Since OHWS and TBTL are so far ahead, we use those as the base spec with which to compare the other talents. There are really only two differences from previous analyses here:

• Previously, we compared Crusade using a OHWS+TBTL+3/5Conviction base to try and account for the synergy with Conviction. In 3.2, we have Vindication and PoJ in the previous tier, making it much more likely that someone will be taking Crusade with as few as 1 point in Conviction. As such, I've decided to just evaluate Crusade on it's own.
• I was curious how SotP and Reckoning synergize, so I calculated Reckoning's value for two bases, one with no points in SotP and one with 5 points in SotP

Setup:
Rotation: 969
Seals: SoV
Glyphs: V only
Talents: Various, in combinations to isolate the benefit of each talent individually.
Gear: Theck's "WTF Double Vanq again?" set

Code: Select all

Talent       DPS    TPS   (both per point)
Reck        14.5   28.4
Reck (5S)   15.3   30.6
Conv        17.7   43.6
SotP        26.9   74.7
Crus        29.9   76.8
TbtL        50.0  138.6
OHWS        86.2  220.1

The change to SotP makes a world of difference in its effectiveness. If it didn't affect the 5-stack proc, each point would be worth only 45.8 TPS (16.5 DPS), slightly ahead of Conviction but far behind Crusade. In the current form though, each point is worth almost as much as a point in Crusade, without having any pre-requisites. This will have a noticeable effect on the spec comparison, as we'll see shortly.

Interestingly, despite all the hype, the Reckoning/SotP synergy is quite weak. It does make Reckoning stronger, but nowhere near enough to make it worth as much as Conviction. On the other hand, it's much more accessible, which will also have a noticeable effect on the spec comparison.

Let's take a look at the plot of viable talent combinations to see how this changes the threat output of popular specs. I've made a lot of changes to the specs represented here as a result of the SotP change. Hopefully the reasoning will be self-evident later on when I post example talent specs.

Setup:
Rotation: 969
Seals: SoV
Glyphs: V only
Talents: Various combinations of SotP (S), Reckoning (R), Conviction (V), and Crusade (C)
Gear: Theck's "WTF Double Vanq again?" set

Code: Select all

Spec       DPS         TPS  Utility Points
4S        3101        7977   11   (PoJ, Vin, HotC, DS, DG, 2/2 Imp Jud)
5S        3128        8051   10
5S+3R     3174        8144    7           
5S+5R     3205        8205    5
5S+3R+1V  3193        8191    7   (Pick one: Vin/PoJ, can't get DS/DG)
5S+3R+3V  3231        8284    5   (can't get Vin/PoJ/DS/DG)
5S+1V     3146        8097    9
5S+3V     3183        8188    7
5V        3082        7896   10
1V+3C     3101        7953   11   (Same as 4S)
3V+3C     3143        8054    9   (Pick one: Vin/PoJ)
5V+3C     3174        8133    7   (can't get Vin/PoJ)

Lots of interesting things going on in this graph. First of all, we have a new winner for the top DPS spec. Several of the 5S+ specs have eclipsed the 5V+3C 53/18 spec, with 5S+3R+3V being the absolute highest. However, anything with 5S and either 2+ points in Conviction or 3+ points in Reckoning will out-threat the standard 5V+3C. Note that even 5S+5R is beating out the Crusade specs now; this isn't really due to any virtue of Reckoning other than being easily available though. Reckoning's still a very weak talent, it's just a very weak talent that costs nothing to access. Seals of the Pure is what really pulls the weight in the spec.

Another interesting thing to note is that 4 points in SotP is stronger than 1V+3C. This is interesting because I and many others expected to spec 53/18, with PoJ and Vindication, leaving 1V+3C to fill out the threat subcomponent of the ret spec. However, those four points would now be better spent in SotP, making a 4/53/14 spec like this one superior to 53/18 for threat while keeping both utility talents in Ret. Those who want more threat could drop the three points in DS/DG for 3 points in Reckoning. Unfortunately, we can't move those 3 points out of Prot since we need them to move down the tree. I've also included 5S+1V and 5S+3V specs on the plot for those who decide to pick up Conviction instead of either PoJ or Vindication.

One conclusion we can come to is that this patch signals the death of the Crusade spec. There's really no point to speccing Crusade anymore, because even with a full-on 5V+3C with no utility, you'll lag behind a 5S+3R which still leaves a few points to your discretion for fun stuff or utility. It's not that Crusade has gotten any worse, but it's location deep in the Ret tree and the massive improvement to SotP just make it too inefficient to keep up.

This means that for patch 3.2, 4/53/6 will be our base spec, leaving 8 points for whatever we want. Note that even with 4 points in SotP, you can still get every utility talent in Ret that we currently have access to, or most of the reasonable utility talents in Prot (all but Stoicism and Guardian's Favor, for example). The 3 points in DS/DG are the 3 "prot" points required to go deeper in the tree, and can go in any lower tier prot talent you want. That's a total of 11 points for utility or threat.

Out of our 8 "free" points, our options are as follows:

• For threat, fill in the last point of SotP. Then:
  
  • If you used your 3 "prot" points 3/5 Reckoning, then push into ret for 3/5 Conviction to max out threat.
  • Otherwise, just fill 5/5 Reckoning and spend the rest of your points in whatever utility talents you want
• For Utility builds, grab whatever utility talents you want. The remaining points should be spent to fill SotP, and then Conviction or Reckoning depending on what you have left. 5R > 3V, for example, but if you have a choice between 3R and 3V fill conviction first.

Here are some example builds that would make sense, along with the data point on the graph they correspond to:

• 4/53/14 Utility spec, has PoJ/Vindication/DS/DG/HotC/Imp Jud. DS/DG can be swapped for Divinity, Reckoning, Imp HoJ, etc. (4S)
• 5/53/13 Utility+Threat balance, drops PoJ or Vindication for slightly higher threat. (5S+1V) DS/DG can be moved to Reckoning for another 90 TPS if desired (5S+3R+1V).
• 5/53/13 All-out Threat, drops PoJ and Vindication for Conviction. DS/DG can be dropped for Reckoning to maximize threat. (5S+3V) or (5S+3R+3V)
• 5/60/6 Blast from the Past, ignores the utility in Ret to pick up utility goodies in prot. You can shift the 10 points in DS, DG, Imp HoJ, and Reckoning around as much as you like. Anything with 5/5 Reckoning will be a "High-Threat" build (5S+5R), anything with 0/5 Reckoning would be, well, weird, but a Utility build of some sort (5S). I doubt many people will go 5/60/6 though simply because the utility offered by Ret is far better than that in Prot.

Note that the difference between the highest-threat spec and the lowest-threat spec is only 300 TPS (7977 to 8284), so we're only talking a 3-4% variation even if you go for the highest-utility build. At the moment, I'm leaning towards the 4/53/14 utility build myself, because I definitely want Vindication and PoJ, despite PoJ costing around 120 TPS (1 point each in SotP and Conviction). Depending on the result of the DS/DG changes, those points may move to Reckoning for 90 more TPS.


<edit> Again, this is only roughly correct, and only for regular mobs. See this post for more accurate (and simpler) rules-of-thumb.
TLDR version

• Crusade specs are so last patch.
• SotP is the new black.
• To maximize threat, you'll generally want to fill points in the highest "threat per point" talent you have available. There's no hard-and-fast rule for how to maximize this since it will depend on how you allocate your utility talents. The most generic rule I can come up with is to fill points in the following order (after maxing out OHWS and TbtL): SotP > Conviction (only if already unlocked by utility talents) > Reckoning. Ignore Crusade.
• YMMV though, depending on how you spec utility (example: if you use DS/DG to go down prot and didn't take more than 8 points in Ret, then you're better off with 5/5 Reckoning than pushing to get 3/5 Conviction). Your best bet is to check the graph and see which combination of threat talents you can hit with different variations on your build.
• Picking up Utility talents will cost you up to 300 TPS at most, or around 3.5% of your total threat.

[NYRIN1]

welcome back reckoning!!!

you have been missed.

[Worldie]

Glad i can put those 3 points in Reckoning rather than Divinity then.