Theck's MATLAB thread - Cataclysm/4.x

[d503]

I wasn't suggesting re-running the sims, but just updating the front page of this post to remove the note about 22 expertise being a bug on 4.0.1 live.

As a dwarf using a mace and receiving 3 expertise from that, I should only need 96 - 3*7.68 = 73 rating with the SoT glyph?

It's not a bug, it's an intended rescaling of expertise to maintain the 26 number at level 85.

[Arianne]

Yes, that was my point. Theck has it noted on the front page of this thread as a likely bug in the 4.0.1 client.

[theckhd]

Yes, that was my point. Theck has it noted on the front page of this thread as a likely bug in the 4.0.1 client.

FixT.

[d503]

Yes, that was my point. Theck has it noted on the front page of this thread as a likely bug in the 4.0.1 client.

My bad. Soz m8.

[theckhd]

Open question for discussion:

I've got the basics of the AoE simulation worked out, and it's spitting out mostly-reasonable numbers. Now I need some feedback on what people think would be useful.

It's currently spitting out two different calculations based on a model proposed by tlitp. The first set of calculations is the total damage done to N mobs. Note that this isn't the average damage per mob, because you could be wailing away on one thing and just getting incidental damage on everything else from HotR, Holy Wrath, Consecration, and AS bounces. It's juts a sum of every point of outgoing damage you produce, though if you're tab-targetting you could make a reasonable argument for dividing the output by N to get an average damage per mob. Here's an example output:

Code: Select all

                                            # Mobs
   Q#   Queue                      2      3      4      5      6     Empty   E%
    1   ShoR>HotR>Cons>AS>HW>J   10845  12988  14201  15414  16628    312   1.0
    2   ShoR>HotR>AS>Cons>HW>J   10882  13052  14251  15450  16648    285   0.9
    3   ShoR>HotR>AS>Cons>J>HW   11162  13328  14516  15704  16892     21   0.1
    4   ShoR>HotR>AS>J>Cons>HW   11301  13434  14581  15727  16874     13   0.0
    5   ShoR>AS>HotR>Cons>HW>J   10914  13135  14283  15431  16579    991   2.5
    6   AS>ShoR>HotR>Cons>HW>J   10947  13186  14322  15459  16595   1103   2.8
    7   AS>HotR>ShoR>Cons>HW>J   10914  13276  14501  15726  16952    263   0.9
    8   AS>HotR>Cons>ShoR>HW>J   10855  13229  14465  15701  16938    302   1.0
    9   HotR>AS>Cons>ShoR>HW>J   10861  13243  14480  15717  16954    301   1.0
   10   HotR>Cons>AS>ShoR>HW>J   10848  13203  14446  15688  16930    294   1.0
   11   Cons>HotR>AS>ShoR>HW>J   10880  13233  14476  15718  16960    270   0.9
   12   Cons>AS>HotR>ShoR>HW>J   10895  13216  14419  15621  16824    580   1.7

It's also calculating a "split" output based on Holy Wrath, Consecration, and HotNova. This can be thought of as "guaranteed" damage on every mob, because each of these three abilities hits everything around you. The output is calculated as a per-mob value rather than a sum. It looks something like this:

Code: Select all

                                          # Mobs
   Q#   Queue                      1     2     3     4     5    Empty   E%
    1   ShoR>HotR>Cons>AS>HW>J   1287  1262  1250  1243  1238    312   1.0
    2   ShoR>HotR>AS>Cons>HW>J   1275  1249  1237  1229  1224    285   0.9
    3   ShoR>HotR>AS>Cons>J>HW   1237  1221  1213  1208  1204     21   0.1
    4   ShoR>HotR>AS>J>Cons>HW   1179  1168  1163  1160  1157     13   0.0
    5   ShoR>AS>HotR>Cons>HW>J   1226  1200  1187  1179  1174    991   2.5
    6   AS>ShoR>HotR>Cons>HW>J   1214  1188  1175  1168  1162   1103   2.8
    7   AS>HotR>ShoR>Cons>HW>J   1303  1277  1264  1256  1251    263   0.9
    8   AS>HotR>Cons>ShoR>HW>J   1314  1288  1275  1268  1262    302   1.0
    9   HotR>AS>Cons>ShoR>HW>J   1315  1289  1276  1268  1263    301   1.0
   10   HotR>Cons>AS>ShoR>HW>J   1320  1294  1281  1274  1268    294   1.0
   11   Cons>HotR>AS>ShoR>HW>J   1320  1294  1281  1274  1268    270   0.9
   12   Cons>AS>HotR>ShoR>HW>J   1282  1255  1242  1234  1229    580   1.7

Don't take these values too seriously yet, there are still some tweaks to be made (proper glyphs in particular, but a few other things too).

I can trivially extend these tables out to 10 or more mobs if I want to, I'm just not sure it's terribly useful to do so.

The real question is, "Do these tables tell you something useful about how to AoE tank?" AoE tanking has a lot of variables involved, so it's impossible to come up with one model that covers every situation. What I'm aiming for is to come up with general trends that we can turn into rules of thumb ("Use GC procs immediately" or "ignore Holy Power unless you can't cast an AoE spell" for example).

I think both of these are interesting metrics, but I want to make sure that there aren't other (possibly more interesting) metrics I should be looking at. In particular, I'd like to hear suggestions; what would you like to see this simulation spit out?


(Aside: It's probably worth noting that given how similar all of the outputs are, AoE tanking is basically going to be a giant clusterfuck. As long as you're casting something every GCD, and as many of those somethings as possible are AoE spells, you'll be within a few percent of whatever result wins. In all likelihood, it'll just be easiest to 9H9 while using AS procs on cooldown and filling any gaps with Cons/HW/J/HoW/ShoR/WoG. I'm sort of hoping that the advice will boil down to "make sure you're casting something every GCD," which would make AoE tanking very malleable so tanks can develop their own style without worrying too much about what's best.)

[polonadis]

For me aoe tanking generally has 2 phases, with phase 1 being the more important one.

Phase one is the first 4-5 gcds that let you grab all the mobs you are going to tank and the priority here is not maximizing total damage done over time, but making sure everything will stay in range of our aoe abilities (meaning we want to prioritize maximum instant damage above superior damage over time). What I personaly do is I pop DP for hp stacks, then I AS the most distant mob (or a caster if there is one), pop inquisition (and righteous defence the person who has aggro, commonly the healer) while running in melee range, hotr, hw, cons, hotr, judgement before slipping into phase2.

Phase 2 is about maximizing dmg output over time and taunting the mobs dpsers pull off. In this phase I prioritize as follows : hotr > hw > cons > as > judgement saving holy power for the mobs that I'm losing aggro on (sotr) or inquisition if there are a lot of mobs/everything sticks just fine. The reason why I prioritize hw above consecration is mana efficiency, as I put 1 or 0 points in HG. I also run with AS glyphed for single target, unglyphed would be higher, of course.


To answer your question, the simulation you did kinda covers the second phase (excluding my preference to save sotr for the slipping mob, which i only do in up to 5 mobs really), but doesn't really say much about the first phase.

At 80 lvl I do basically the same thing, but without inquisition, obviously.

[knaughty]

(Aside: It's probably worth noting that given how similar all of the outputs are, AoE tanking is basically going to be a giant clusterfuck. As long as you're casting something every GCD, and as many of those somethings as possible are AoE spells, you'll be within a few percent of whatever result wins.

Q: What's our AoE rotation?
A: HotR, and whatever AE is off CD

Looks like I guessed right...

My view is that there's two situations for AE tanking. Given there's no significant difference between the rotations you're modelling it's somewhat moot, but when "multi-mob-tanking" you're generally doing one of two things:

(1) Cleave-tanking.
You have a large trash pack that needs tanking or a small number of adds + a boss. There's a primary DPS target, you need to hold aggro on secondary targets that are being cleaved by DPS. Some mobs are more dangerous than others and you'll be focus-firing those mobs down. Depending on danger levels, you may be blowing some CDs to avoid a death while the scary mobs get killed. Good historical examples: TBC trash packs from raid content in Tier 5, Black Temple and especially Sunwell.

(2) Pure AoE tanking.
There's a boatload of small mobs - at least 6, possible a dozen or more. They're pretty much identical, and hit gently enough that one loose mob isn’t going to kill anyone. The Magelocks are dumping pure AE on your head as hard as they can. Examples include basically all of WotLK raid-trash and hopefully very little of Cataclysm.

From a modelling perspective, I think that doing level 80 mechanics is close to worthless – either people have already worked it out or they’re not raiding. Concentrate on level 85 mechanics. A key criteria that I’m interested in is when we switch from ShoR to Inquisition – but this has more to do with what your DPS are doing and what the pack is like. Cleave tanking you’ll probably ShoR (or WoG). Pure-AE you’d pop Inquisition.

[theckhd]

From a modelling perspective, I think that doing level 80 mechanics is close to worthless – either people have already worked it out or they’re not raiding. Concentrate on level 85 mechanics.

Well, as far as modeling goes, there's little to no difference. The code I write is basically the same, the only thing that changes at 85 is what damage values go into the matrices. So whatever code I write will be the same either way. I've even included Inquisition rotations (though I didn't post them) in the list of priority queues to sim out - the simulation doesn't know (or care) that we don't get it until 81.

It sounds like I've got both of your situations covered - total damage is a pretty good measure of Cleave tanking, and split (or "guaranteed") damage sort of covers pure AoE. The questions end up being in the details - would it be useful to also consider AS damage for pure AoE (i.e. AS_DAMAGE*3/N as an average estimate), or do we throw out anything that isn't 100% reliable? Similarly, do we care about the breakdown between single-target and AoE in the Cleave tanking model? Should I show that explicitly, or just make general comments about it in the write-up?

[knaughty]

If Inquisition is in the model and it’s minimal extra work for 80 vs 85, then I revoke my objection

All the below is “IMnsHO”.

I wouldn’t include AS in the “Pure AoE” tanking model. In the circumstance where you have a dozen mobs to tank, all you care about is the minimum damage you’re doing to each mob, because what matters is the clothies not getting their faces ripped off by peeling half the pack. The fact that 3/10 are glued to you from AS doesn’t help when the other 7 just ran off and ate the magelock who popped his trinkets.

For the “Cleave-tank” model, what matters is the damage to the primary target and the minimum damage to any target, because you’re trying to avoid two things: Losing aggro on the current kill target that the rogues are going bonkers on while also not losing aggro on anything that is collecting cleave/incidental AE (the warrior that just popped whirlwind or the lock tab-spamming SoC on secondary targets in order to smash the primary).

”Total damage done by the tank” is not a compelling metric – we’re tanks, not DPS. Prioritising AS over Consecrate might be more damage, but if it leads to “Mob #5” running off and hitting a DPS or healer, you’re doing it wrong.

[Kelaan]

I wouldn’t include AS in the “Pure AoE” tanking model.... The fact that 3/10 are glued to you from AS doesn’t help when the other 7 just ran off and ate the magelock who popped his trinkets.

Prioritising AS over Consecrate might be more damage, but if it leads to “Mob #5” running off and hitting a DPS or healer, you’re doing it wrong.

I think you make some good points; I mostly agree with you. (Also, many laughs at "Magelocks".) However, Avenger's Shield is a large chunk of damage (threat) that we can put on a portion of the adds -- and with Grand Crusader, will at times get chances to cast relatively quickly in succession. We can't count on that, for obvious reasons, but it's still powerful.

I also think that, if the herd of adds is large enough that we worry about Magelocks pulling aggro on a subset of them, they won't be a gibbing hazard. And, if they are, they might be hitting hard enough that focusing down a few ("cleave" situation) is a better idea anyways. Besides, Magelocks have tools to deal with undesired aggro.

[Suzytincan]

Does anyone see any benefit to modeling different combat lengths? Before they started talking about having Divine Plea generate HP, I found myself wondering how many HP we should save up before hitting inquisition. Now that seems like a mostly moot point to me, but maybe someone can think of an interesting and sufficiently common scenario where knowing combat time may affect the outcome.

[Metherlance]

Does anyone see any benefit to modeling different combat lengths? Before they started talking about having Divine Plea generate HP, I found myself wondering how many HP we should save up before hitting inquisition. Now that seems like a mostly moot point to me, but maybe someone can think of an interesting and sufficiently common scenario where knowing combat time may affect the outcome.

The change on DP doesn't really change anything, it's just there so we can start with our hard hitting ability. And like Theck, already pointed on another topic, even with the change of ShoR our priority SHOULD NOT change at all.

Edit: if your point was on when to use INQ or when not, it should not change at all, so you should not use INQ on single target tanking.

[Thelmiance]

Is it safe to assume that, when DP comes back up part way through a long, single-target fight, it's optimal to do ShoR-DP-ShoR? If for some reason we were really concerned about aggro or total dps, that is.

[Metherlance]

Is it safe to assume that, when DP comes back up part way through a long, single-target fight, it's optimal to do ShoR-DP-ShoR? If for some reason we were really concerned about aggro or total dps, that is.

I can't see a boss fight it is an issue. But for some adds or so, it is really good for it. You have a great lead on threat on the boss, you need to pick up an add, judge, DP, SHOR, and he is all yours.

[d503]

Is it safe to assume that, when DP comes back up part way through a long, single-target fight, it's optimal to do ShoR-DP-ShoR? If for some reason we were really concerned about aggro or total dps, that is.

Maybe if you're struggling with threat or something, but how about using it for more of a survivability cooldown?

Case in point:

You're at 60% health, you have 5 seconds on the zzBigDragonBreath timer that will hit for up to 70k. At that point you can WoG-DP-WoG to better prep for the hit, and you might even have an EG proc to still do so more threat.

As the poster above me said, by the time you're two minutes in, you should have enough of a lead to not need 2 ShoRs in a row.

Certainly, if you're trying to maximize DPS/TPS, go for it, but I think I'll probably be saving mine for an "OH SHIT" moment.