Theck's MATLAB thread - Cataclysm/4.x

[M.C.]

Time for some collaborative fun with the AoE module (which, by the way, isn't updated in the repository yet, I'm still fooling with priority queues).

Has anybody tried using these AOE rotations?

I found most of them to be rather hard to sustain. JotW is a major source of mana regen. When its priority is so low, you risk running OOM rather quickly, especially if you have a "lucky" streak of GC procs and you decide to use them.

I was running HotR>Cons>HW>AS>ShoR>J and it definitely suffers from this. I assume other rotations have the same problem. Sanctuary helps, naturally, but I had issues with 2 or 3 mobs every once in a while (tanking 4+ melee mobs makes things easier). Needless to say, practicing these rotations on a training dummy is out of question, and Hallowed Ground is mandatory.

@Theck: Do you think you could add mana regen/consumption tracking to your models? This will give us an idea how many mobs you need to tank without running OOM.

Edit: by the way, why is the DPS difference so small between rotations with and without Inquisition? Almost all of our AOE damage is Holy. Shouldn't this mean a 30% DPS increase against non-primary targets?

[tlitp]

Do you think you could add mana regen/consumption tracking to your models?

by the way, why is the DPS difference so small between rotations with and without Inquisition?

I. Doable, but we need some things sorted out beforehand. This feature should be available till Cataclysm goes live.
II. There's a blunt answer (i.e. usage/priority of SotR), and there's a more subtle one (right now Vengeance is naively modeled, so there is a fairly significant overshoot for any ability that scales massively with AP).

[theckhd]

It also could be a bug in the Inq implementation. I'll take a closer look tomorrow.

[Awyndel]

Remember the scaling on sotr is nerfed in beta. Also, while aoeing, like mentioned before, single target dps might matter less then aoe threat. It's a choice.

Yeah I have noticed going oom when not judging too often. Specially when not speced into consecrate. For simplicity's sake I just keep judging and using holy power every cycle, except the pull.

[kirsty]

Guaranteed:

Code: Select all

                                             # Mobs
   Q#   Queue                        1     2     3     4     5    Empty   E%
    1   ShoR>HotR>J>AS>Cons>HW      1838  1829  1662  1173  1171     0   0.0
    2   ShoR>HotR>AS>J>Cons>HW      1949  1938  1745  1181  1178    19   0.2
    3   ShoR>HotR>AS>Cons>J>HW      2003  1986  1790  1224  1220    15   0.1
    4   ShoR>HotR>AS>Cons>HW>J      2045  2019  1818  1244  1239   101   1.0
    5   ShoR>HotR>Cons>AS>HW>J      2018  1992  1800  1253  1248   118   1.2
    6   ShoR>HotR>Cons>HW>AS>J      1945  1916  1744  1259  1253   110   1.1
    7   ShoR>HotR>HW>Cons>AS>J      1954  1926  1751  1259  1253    96   1.0
    8   ShoR>AS>HotR>Cons>HW>J      2043  2016  1801  1185  1180   378   3.0
    9   AS>ShoR>HotR>Cons>HW>J      2061  2035  1809  1165  1159   387   2.9
   10   AS>HotR>ShoR>Cons>HW>J      2188  2162  1928  1254  1249   121   1.2
   11   AS>HotR>Cons>ShoR>HW>J      2208  2182  1946  1268  1263   122   1.2
   12   HotR>AS>Cons>ShoR>HW>J      2201  2175  1941  1269  1264   106   1.1
   13   HotR>Cons>AS>ShoR>HW>J      2179  2153  1925  1275  1270   109   1.1
   14   HotR>Cons>HW>AS>ShoR>J      2152  2119  1902  1292  1285   267   2.7
   15   Cons>HotR>AS>ShoR>HW>J      2192  2166  1935  1275  1270   115   1.1
   16   Cons>AS>HotR>ShoR>HW>J      2144  2117  1888  1234  1229   226   2.0
   17   Cons>HW>AS>HotR>ShoR>J      2073  2039  1811  1166  1159   559   5.2
   18   HotR>Cons>HW>AS>Inq>J       2322  2283  2035  1337  1330   513   5.1
   19   HotR>AS>Cons>HW>Inq>J       2416  2381  2103  1312  1305   514   5.1
   20   Inq>HotR>AS>Cons>HW>J>Inq   2480  2443  2158  1349  1342   177   1.8
   21   Inq>HotR>Cons>AS>HW>J       2455  2419  2143  1359  1351   413   4.1
   22   Inq>HotR>AS>Cons>HW>J       2471  2437  2153  1343  1336   398   4.0
   23   Inq>AS>HotR>Cons>HW>J       2412  2375  2086  1262  1255   734   5.7
   24   AS>Inq>HotR>Cons>HW>J       2376  2340  2061  1267  1260   739   6.1

why is the guaranteed damage so different if you look at the number of mobs? i mean, it SHOULD be exactly the same, the only numbers that should matter is 3 (because of AS) and 11+ (because of ae-dmg cap). i mean, all of these enemies are hit by the same attacks, HW, cons and hotr (+AS if 3 mobs or less). for example, rotation #20, 2 mobs get around 2.4k dps, 3 mobs get 2.1k dps. at first i thought this only happened from 2 to 3 mobs, but there is also a slightly difference between 4 to 5 mobs. i thought at first that was only random bad luck, but as ALL rotations do more guaranteed damage to 4 mobs, and less guaranteed damage to 5 mobs (only slightly though) im not so sure anymore.

[theckhd]

Holy Wrath splits its damage evenly amongst targets, so the net damage per target will inevitably go down as you add targets.

[theckhd]

@Theck: Do you think you could add mana regen/consumption tracking to your models? This will give us an idea how many mobs you need to tank without running OOM.

Edit: by the way, why is the DPS difference so small between rotations with and without Inquisition? Almost all of our AOE damage is Holy. Shouldn't this mean a 30% DPS increase against non-primary targets?

1) I can, but I doubt it's necessary. In my experience (at 80 anyhow), even with 0/2 HG mana's never been a serious issue regardless of number of mobs. With less than 3, I stick to something like ShoR>HotR>AS>J>Cons>HW, so Cons is pretty low in the queue. For more than 3, the incoming attack rate tends to be self-sustaining even with Cons high in the queue.

2) I found the source of this discrepancy: HammerNova wasn't getting the Inq modifier properly applied to it. Also, keep in mind that some of the queues don't have 100% Inq uptime.

[kirsty]

ok, i missed that change...thanks a lot.

but, even though holy wrath damage decreases with the number of mobs (per mob at least) there shouldnt be that steep decrease in damage from 2 mobs to 3 mobs. i mean, thats several hundret dps, and looking on your previous calculation where you didnt factor in avangers shield even for 1-3 mobs i cant see this decrease in damage.

also i wonder how you calculated wotl. currently, cs and hotr deal about equal damage, with cs pulling slightly ahead single target (at least fully buffed and everything). i assume that if you wont specc into wotl, and thus wont get 60% damage bonus on cs, this means hammer of the righteous suddenly deals even single target way more damage than cs. wotl thus provides only the difference between hotr and cs plus increased holy wrath, hammer of wrath and judgement damage. did you factor in that you could use a hotr rotation (even single target) without wotl?
personally i guess that if you didnt factor that in, wotl would still be a must-have talent, but "guess" is not as good as "know".

also, it would be nice to know how good which talent would be for ae-tanking, this is of course only possible if we know the ideal rotation, but still. hallowed ground for example really stinks single target, but it definitly outperforms other talents in ae situations.

[theckhd]

but, even though holy wrath damage decreases with the number of mobs (per mob at least) there shouldnt be that steep decrease in damage from 2 mobs to 3 mobs. i mean, thats several hundret dps, and looking on your previous calculation where you didnt factor in avangers shield even for 1-3 mobs i cant see this decrease in damage.

You're right, there was another bug with the way I was calculating AS damage. Those columns are actually supposed to be for 2-6 mobs, not 1-5 mobs (i.e. the headers I put in by hand are wrong). It's actually the third column that's wrong, since it's getting additional AS damage that it shouldn't.

did you factor in that you could use a hotr rotation (even single target) without wotl?

No. I don't think anyone would be daft enough to run a spec without WotL. Even if trying to optimize for AoE, WotL boosts Holy Wrath, so there would be better places to drop points in order to pick up Hallowed Ground (which will probably still be a weak talent for AoE DPS, by the way). Once we're done with the AoE simulations, I can go back and use the coefficients from the simulation to re-evaluate the talents and see how they stack up for AoE.

[kirsty]

I don't think anyone would be daft enough to run a spec without WotL.

that was not my point. my point was, that the calculation is not really correct. most of the dps you get from WotL is because of the 60% dmg buff for crusader strike. as i said, i really dont believe that its a good thing to go without WotL, but it would still be nice to know how much tps WotL EXACTLY gets you.


i tried to calculate this on my own, but got some fishy results. i took the damage values from your ability damage chart and divided them by the time i would use the ability, and then subtracted the amount the talent is responsible for.

this way, judgement with 7178 net dmg unglyphed divided through 1.6 for the talent got me 4486 for untalented dmg. the difference is 2691, divided through 9 for dps, because you use jot every 9 secs, and also divided through 2 for 2 talent points resulted in 149.5 dps per point.

the same for cs would result in 271.75 dps, which is together with jot bonus already WAY higher than your numbers for each talent point. (which are around 300 dps)
for holy wrath i took the raw dmg of hw and multiplied by 0,075 (0,3 crit chance, only 50% bonus dmg) and divided through 15 it should result in 18,645 dps. how is a bit complicated, i just assumed the time below 20% would be 15% of the bosstime which seemed lower than the actual bosstime would be. this way, how got me 20.3 dps. until now we are already at 460 dps per point, which is 50% higher than your value. if i add the difference between cs dps and hotr dps instead of the dps bonus i get from cs, i would get 48 dps bonus instead of 271, which would add up to 236.5 dps for 1 point of wotl.

you see, either way i get completly wrong results, somehow not even near your value. maybe you can show me where i made a mistake? or perhaps something weird is going on in the code...i simply cant explain the high discrepancy between your values and mine.

[theckhd]

Hm... apparently the talent spec sim was only applying self-buffs, which explains why the values don't match up. Fixing that gives me about 450 DPS per point for WotL. I'll upload updated data and plots shortly.

[theckhd]

AoE Rotation Simulations

Setup:
Race: Human (hardly matters)
Talents: 0/31/5 with all prot damage talents + Crusade
Glyphs: SoT/ShoR/HotR for Prime, AS for Major
Seal: SoT
Gear: T10 sample set
Rotation: N/A
Buffs: Full Raid, Full Vengeance
Code: calc_prio_aoe.m
Revision: r169
Date: 6 November 2010

These are very similar to the single-target rotation simulations. The primary difference is the damage metrics we calculate.

For AoE, we care about two things: the damage to our primary target (in case we're burning things down one by one) and the "guaranteed" damage to all secondary targets. Guaranteed damage thus only includes HammerNova, Holy Wrath, Consecration, and for 3 or fewer targets, unglyphed Avenger's Shield.

I've included Inquisition in the queues below, even though it's irrelevant at 80. It shows up three different ways (Inq, Inq*, and Inq**) that indicate different conditionals:
Inq: 3 Holy Power and duration of Inq < 1.5s
Inq*: 3 Holy Power
Inq**: 1+ Holy Power

Code: Select all

                                    Primary       DPS/mob for # Mobs   
   Q#                                  DPS      2     3     4     5     6    Inq   Empty   E%
    1   ShoR>HotR>J>AS>Cons>HW        8294   1833  1824  1175  1173  1171     0%      0   0.0
    2   ShoR>HotR>AS>J>Cons>HW        8274   1955  1944  1178  1175  1173     0%     34   0.1
    3   ShoR>HotR>AS>Cons>J>HW        8201   2010  1993  1232  1226  1223     0%     37   0.1
    4   ShoR>HotR>AS>Cons>HW>J        8042   2040  2015  1252  1245  1240     0%    349   1.2
    5   ShoR>HotR>Cons>AS>HW>J        8029   2034  2008  1260  1252  1247     0%    361   1.2
    6   ShoR>HotR>Cons>HW>AS>J        7987   1949  1921  1267  1259  1253     0%    357   1.2
    7   ShoR>HotR>HW>Cons>AS>J        8001   1961  1933  1269  1261  1255     0%    333   1.1
    8   ShoR>AS>HotR>Cons>HW>J        7971   2064  2038  1189  1181  1176     0%   1042   2.7
    9   AS>ShoR>HotR>Cons>HW>J        7968   2064  2038  1172  1165  1159     0%   1122   2.8
   10   AS>HotR>ShoR>Cons>HW>J        7872   2179  2153  1261  1253  1248     0%    351   1.2
   11   AS>HotR>Cons>ShoR>HW>J        7802   2203  2176  1275  1268  1262     0%    360   1.2
   12   HotR>AS>Cons>ShoR>HW>J        7806   2194  2168  1277  1269  1264     0%    343   1.1
   13   HotR>Cons>AS>ShoR>HW>J        7816   2190  2163  1283  1275  1270     0%    297   1.0
   14   HotR>Cons>HW>AS>ShoR>J        7442   2169  2135  1302  1292  1285     0%    820   2.7
   15   Cons>HotR>AS>ShoR>HW>J        7818   2192  2166  1283  1275  1270     0%    324   1.1
   16   Cons>AS>HotR>ShoR>HW>J        7825   2145  2118  1242  1234  1228     0%    653   1.9
   17   Cons>HW>AS>HotR>ShoR>J        7541   2077  2043  1182  1172  1165     0%   1565   4.9

   18   Inq*>HotR>J>AS>Cons>HW        7598   2418  2406  1551  1547  1545   100%      0   0.0
   19   Inq>HotR>J>AS>Cons>HW         7477   2600  2572  1626  1618  1612   100%    950   3.2
   20   Inq>HotR>AS>Cons>HW>J         7357   2756  2721  1658  1648  1641   100%   1187   4.0
   21   Inq>HotR>Cons>AS>HW>J         7324   2742  2706  1672  1662  1654   100%   1244   4.1
   22   Inq>HotR>Cons>HW>AS>J         7248   2656  2619  1675  1664  1656   100%   1477   4.9
   23   Inq>HotR>HW>Cons>AS>J         7278   2678  2636  1679  1666  1658   100%   1099   3.7
   24   Inq>HotR>Cons>HW>AS>J>Inq*    7338   2712  2672  1681  1669  1661   100%    795   2.6
   25   Inq>HotR>Cons>HW>AS>J>Inq**   7392   2735  2692  1687  1674  1666   100%      0   0.0
   26   HotR>Inq>Cons>HW>AS>J         7254   2661  2624  1675  1664  1656   100%   1453   4.8
   27   HotR>Inq>Cons>HW>AS>J>Inq*    7346   2713  2673  1681  1669  1661   100%    773   2.6
   28   HotR>Inq>Cons>HW>AS>J>Inq**   7392   2741  2698  1687  1675  1666   100%      0   0.0
   29   HotR>Inq>AS>Cons>HW>J>Inq*    7440   2788  2751  1662  1651  1643   100%    551   1.8
   30   HotR>Inq>AS>Cons>HW>J>Inq**   7471   2813  2775  1665  1653  1646   100%      0   0.0
   31   HotR>Cons>HW>AS>Inq>J         7018   2564  2525  1575  1563  1555    75%   1490   5.0
   32   HotR>Cons>HW>AS>Inq>J>Inq**   7298   2704  2662  1640  1627  1619    89%      0   0.0
   33   HotR>AS>Cons>HW>Inq>J         7071   2631  2596  1545  1535  1528    72%   1489   5.0
   34   Inq>AS>HotR>Cons>HW>J         7354   2687  2650  1547  1536  1529   100%   2230   5.7
   35   AS>Inq>HotR>Cons>HW>J         7317   2683  2648  1560  1549  1542    97%   2101   5.7

Observations:

• There's a pretty clear trade-off between Primary and Secondary DPS. If we just replace CS with HotR in the standard 939, we get our maximum DPS on the primary target, but a near-minimum on secondary targets. By shifting the queue around to favor AoE abilities, we lose Primary DPS but gain Secondary DPS.
• The differences we're talking about here aren't very large. Just using a 9H9 clone of the single target rotation nets you 90% of the secondary DPS that the optimum rotation (#14) does.
• HotR is the bread-and-butter of our AoE dps. The queues with high HotR priority tend to outperform anything else in secondary damage.
• High AS priority is a boost to secondary DPS on 2-3 adds, as expected.
• For general purpose AoE, HotR>Cons>HW>AS>(everything else) seems to be the optimal queue.
  
• There's a lot of interesting insight to be had from the Inq rotations, even if it's irrelevant at 80.
  
  • Even though it creates empty GCDs, it's better to wait until Inq is about to run out to refresh it.
  • It's actually a bit better to prioritize HotR over Inq
  • Refreshing Inq early, even with less than 3 Holy Power, is a DPS increase over an empty GCD.
  • Dropping Inq in the queue sacrifices uptime, which is a flat-out dps loss
  • Again, AS priorities tend to do well on 2-3 mobs, though it's still a DPS loss to prioritize it above HotR or Inq.

TLDR Summary:

• At 80, your best pure AoE queue is generally HotR>Cons>HW>AS>(ShoR or WoG)>J. Bump J up if you have mana troubles.
• For cleave tanking, move ShoR back up to the top of the queue for extra single-target threat.
• Mixing the rotation up doesn't make a huge difference in threat. As long as you're casting something every GCD, you're going to be pretty close to optimal.

[M.C.]

1) I can, but I doubt it's necessary. In my experience (at 80 anyhow), even with 0/2 HG mana's never been a serious issue regardless of number of mobs. With less than 3, I stick to something like ShoR>HotR>AS>J>Cons>HW, so Cons is pretty low in the queue. For more than 3, the incoming attack rate tends to be self-sustaining even with Cons high in the queue.

I am not sure where the differences are coming from, but it has been an issue for me (which is why I posted it in the first place). In any case, it would be nice if cleave rotations that aren't sustainable mana-wise are excluded from the simulation.

[theckhd]

I am not sure where the differences are coming from, but it has been an issue for me (which is why I posted it in the first place). In any case, it would be nice if cleave rotations that aren't sustainable mana-wise are excluded from the simulation.

It's not easy to draw a line between sustainable and non-sustainable though. That will depend on a lot of environmental factors:

• How many mobs are you fighting?
• What sort of attack speed do they have?
• Are they caster mobs or melee mobs?
• What sort of block % do you have? (i.e. Have you reforged heavily for mastery or not?)
• Are you utilizing GC procs or not? (if you're even specced for it)
• Do you have Consecration glyphed?

AoE is going to be a lot more subjective from now on, I'm afraid. Even if one rotation is a few DPS higher than another, there will be situations where you'll be better off breaking that rotation to front-load threat or hold a spell back to utilize it more efficiently (HW as a stun or Cons placement on multiple targets, for example). A good paladin's just going to have to be practiced and recognize those situations and react accordingly, which includes watching his resources and slipping Judgement in when necessary.

[theckhd]

Also, I posted the Enchant/Food analysis on the first page of the thread.