Theck's MATLAB thread - Cataclysm/4.x

[theckhd]

I think I have the FSM implementation working now. Here's a sample output of the updated single-target queue comparisons. Keep in mind that this is still preliminary. I'm 99% sure the code is working properly, but I haven't had the time to evaluate it as thoroughly as usual.

Note the FSM code uses a new notation. An "I-" prefix means "iff Inq active," while an "i-" prefix means "iff Inq not active." "SD-" and "sd-" have similar connotations. AS+ still means "AS iff GC buff active."

The main reason I'm posting this is for people to look at the numbers and do some sanity checking - make sure there's nothing on here that seems nonsensical, like a queue that should have Inq uptime but doesn't, etc.

Code: Select all

SoT / 2% hit / 16 exp
                                         DPS            TPS            SHPS            E     I 
  Q#  Priority                           V=100%  V=30%  V=100%  V=30%  V=100%  V=30%   %     % 
   1  SotR>CS>AS>J                       14375    9249  43126   27747     0      0     7.3   0.0
   2  SotR>HotR>AS>J                     13085    8323  39256   24970     0      0     7.3   0.0
   3  SotR>CS>J>AS                       14271    9169  42813   27506     0      0     7.4   0.0
   4  SotR>AS>CS>J                       14198    9135  42595   27405     0      0     8.6   0.0
   5  AS>SotR>CS>J                       14067    9052  42202   27155     0      0     9.2   0.0
   6  SotR>CS>AS+>J>AS                   14346    9224  43039   27672     0      0     7.3   0.0
   7  SotR>AS+>CS>J>AS                   14320    9207  42959   27621     0      0     7.7   0.0
   8  sdAS>sdJ>SotR>CS>AS>J              13831    8888  41493   26663     0      0    11.6   0.0
   9  sdAS>SotR>CS>AS>J                  14229    9155  42686   27465     0      0     8.8   0.0
  10  SotR>CS>AS>J>HW                    14611    9485  43834   28455     0      0     2.8   0.0
  11  SotR>CS>AS>J>Cons>HW               14806    9590  44418   28770     0      0     1.6   0.0
  12  SotR>AS+>CS>AS>J>Cons>HW           14781    9576  44345   28728     0      0     1.8   0.0
  13  sdAS>sdJ>SotR>AS+>CS>AS>J>Cons>HW  14331    9291  42994   27873     0      0     4.8   0.0
  14  Inq>CS>AS>J                        12625    8114  37874   24342     0      0     7.7  89.0
  15  Inq>HotR>AS>J                      11961    7570  35882   22710     0      0     7.7  89.0
  16  iInq>SotR>CS>AS>J                  13955    8972  41866   26917     0      0     7.6  63.2
  17  ISotR>Inq>CS>AS>J                  13955    8972  41866   26917     0      0     7.6  63.2
  18  SDSotR>ISotR>Inq>CS>AS>J           14452    9294  43356   27881     0      0     7.5  41.5
  19  ISotR>SDSotR>Inq>CS>AS>J           14452    9294  43356   27881     0      0     7.5  41.5
  20  ISDSotR>Inq>CS>AS>J                13794    8867  41382   26602     0      0     7.6  73.5
  21  Inq>CS>AS>J>Cons>HW                13193    8566  39581   25699     0      0     1.8  89.0
  22  Inq>HotR>AS>J>Cons>HW              12530    8022  37590   24066     0      0     1.8  89.0
  23  ISotR>Inq>CS>AS>J>Cons>HW          14461    9373  43383   28119     0      0     1.7  63.2
  24  SDSotR>ISotR>Inq>CS>AS>J>Cons>HW   14935    9677  44806   29030     0      0     1.7  41.5
  25  SDSotR>ISotR>Inq>CS>AS>J>ICons>HW  14874    9644  44624   28933     0      0     2.1  41.5
  26  SDSotR>Inq>CS>AS>J>Cons>HW         14783    9581  44351   28745     0      0     1.7  51.2
  27  ISDSotR>Inq>CS>AS>J>Cons>HW        14324    9288  42973   27865     0      0     1.7  73.5
  28  WoG>SotR>CS>AS>J>Cons>HW           13618    8828  40854   26485  1266    902     1.7   0.0
  29  WoG>SotR>HotR>AS>J>Cons>HW         12307    7887  36922   23663  1266    902     1.7   0.0
  30  WoG>Inq>CS>AS>J>Cons>HW            12671    8226  38013   24680  1269    904     1.8  58.9
  31  WoG>Inq>HotR>AS>J>Cons>HW          11795    7551  35385   22654  1269    904     1.8  58.9
  32  WoG>Inq>AS+>CS>AS>J>Cons>HW        12680    8236  38042   24708  1251    892     2.1  58.1
  33  WoG>ISotR>Inq>CS>AS>J>Cons>HW      13246    8593  39740   25781  1267    903     1.7  44.0
  34  WoG>iInq>SotR>CS>AS>J>Cons>HW      13246    8593  39740   25781  1267    903     1.7  44.0
  35  SotR>CS>AS>J>HoW                   15569   10072  46706   30217     0      0     1.6   0.0
  36  SotR>CS>AS>HoW>J                   15906   10338  47718   31015     0      0     0.3   0.0
  37  SotR>CS>HoW>AS>J                   15864   10337  47592   31011     0      0     0.3   0.0
  38  SotR>HoW>CS>AS>J                   15722   10261  47165   30783     0      0     1.3   0.0
  39  HoW>SotR>CS>AS>J                   15608   10197  46823   30591     0      0     1.9   0.0
  40  SotR>CS>AS+>HoW>AS>J               15907   10349  47722   31047     0      0     0.2   0.0
  41  SotR>CS>HoW>AS>J>Cons>HW           15896   10358  47689   31075     0      0     0.0   0.0
  42  SotR>HoW>CS>AS>J>Cons>HW           15851   10349  47554   31048     0      0     0.0   0.0
  43  HoW>SotR>CS>AS>J>Cons>HW           15790   10325  47370   30975     0      0     0.0   0.0
  44  ISotR>Inq>CS>HoW>AS>J>Cons>HW      15856   10357  47567   31071     0      0     0.0  62.3
  45  HoW>ISotR>Inq>CS>AS>J>Cons>HW      15912   10439  47738   31317     0      0     0.0  63.3
  46  WoG>HoW>SotR>CS>AS>J>Cons>HW       14669    9601  44007   28802  1272    906     0.0   0.0
  47  WoG>SotR>HoW>CS>AS>J>Cons>HW       14764    9653  44294   28960  1225    873     0.0   0.0

[daiceman]

Edit: I read the text on SD incorrectly, I assumed you would only get a SD proc if the AS was a GC proc'ed one.

Probably not very important, but rotation #13 should probably be changed to:

sdAS+>sdJ>SotR>AS+>CS>AS>J>Cons>HW

I'm assuming that its supposed to be a SD fishing rotation, proving that we shouldn't fish for procs, but atm you currently have it casting AS regardless of GC status meaning its casting ASs without the chance to gain SD.

On a side note yay for my rotation looking like its doing the best atm (epeen++)

[Iminmmnni]

I'm assuming that its supposed to be a SD fishing rotation, proving that we shouldn't fish for procs, but atm you currently have it casting AS regardless of GC status meaning its casting ASs without the chance to gain SD.

The wording of PTR Sacred Duty doesn't indicate that you need a GrCr proc for AS to proc SD. Unless I'm missing something, you can still fish for an SD proc even if you're not going to get any HP from it.

[theckhd]

Single-target rotation simulations are updated. I probably won't get to the rest of them until this afternoon/evening.

Interestingly, there's a slight DPS gain to be had over the default 939. (SotR if SD or Inq active)>Inq>CS>AS>J beats out SotR>CS>AS>J by a hair (~75-100 DPS). Still less than a 1% increase though, so it's not clear if that's worth the extra mental bandwidth. On the other hand, if you're using a rotation manager like clcinfo, it should be easy enough to code the conditionals.

[Awyndel]

Thnx once more Theck. I can prepare for wednesdays raid .

I like the spec from the sim, I will prolly use it. Also replacing wog glyph with CS.

Finally we are allowed to care abit more about dps .

[Chicken]

• It's very likely that WoG>SotR2>CS>AS>J is a slight DPS boost over WoG>CS>AS>J for little to no SHPS cost. This will be added to the sims in the next pass.

I have no idea how feasible this is with your Sim coding, but wouldn't a cooldown based check work better for this? As in use a SotR if you have 3 Holy Power (Unlikely but can potentially happen if WoG>CS>AS+>CS ends up happening) or if a certain amount of time/ability uses is left until WoG is ready again. Time variations would likely be in the 9.5-15.5 seconds range depending on how save you want to play it: 15.5 seconds usually leading to an effective "use SotR2 once" in most cycles if I'm looking at things correctly, whereas the 9.5 seconds end of the range would result in trading more HPS for DPS; varying with your expertise/hit obviously.

To summarize: Is it a good idea to burn your currently available Holy Power on a SotR when there's a certain amount of time left before WoG is ready again?

[theckhd]

• It's very likely that WoG>SotR2>CS>AS>J is a slight DPS boost over WoG>CS>AS>J for little to no SHPS cost. This will be added to the sims in the next pass.

I have no idea how feasible this is with your Sim coding, but wouldn't a cooldown based check work better for this? As in use a SotR if you have 3 Holy Power (Unlikely but can potentially happen if WoG>CS>AS+>CS ends up happening) or if a certain amount of time/ability uses is left until WoG is ready again. Time variations would likely be in the 9.5-15.5 seconds range depending on how save you want to play it: 15.5 seconds usually leading to an effective "use SotR2 once" in most cycles if I'm looking at things correctly, whereas the 9.5 seconds end of the range would result in trading more HPS for DPS; varying with your expertise/hit obviously.

To summarize: Is it a good idea to burn your currently available Holy Power on a SotR when there's a certain amount of time left before WoG is ready again?

Yes, it's entirely feasible for the FSM code to handle that, we just have to define a parameter for it. A generic parameter for checking the cooldown of another spell is probably the best way to implement it - i.e. "SotR(cdWoG>3)" or something.


Side note: all of the sims on the front page are updated to r290. I'll probably do another pass in a day or two with any bugfixes and such.

[Iminmmnni]

Yes, it's entirely feasible for the FSM code to handle that, we just have to define a parameter for it. A generic parameter for checking the cooldown of another spell is probably the best way to implement it - i.e. "SotR(cdWoG>3)" or something.

fsm code has been updated to handle conditionals of the form [cd|buff<Ability/Buff><op><value>]
. We can now run rotations such as WoG>SotR[cdWoG>15]>CS>AS>J. I put the logic in for both cooldowns and buffs so we can do stuff like Inq[buffInq<3] as well. iInq/ISotR et al (and the sd/SD variants) are now just shorthands for Inq[buffInq=0]/SotR[buffInq>0]. (The rotation parsing code is much more complicated now as > is now being used as both the separator for the abilities CS>AS and a conditional operator).

Aside: the code is now using an analytical model instead of simulations. It's *significantly* more computationally expensive than the previous model but a whole lot faster than running simulations. As we're no longer performing simulations, 50dps between rotations is actually meaningful again.

[raistlin212]

Depends on how you define 'meaningful'. The analysis might be perfectly accurate, but it model is too ideal to be real world accurate to that degree. .3% changes in DPS tend to smooth out once mental bandwith, game lag, movement, miss streaks, using Hand spells on raid members, etc happen.

[theckhd]

Single-Target Rotation Simulations
The simulation has extremely tight error tolerances, giving us accuracy down to ~1 DPS. So any difference in the output of two queues can be considered statistically significant from a purely mathematical point of view. However, keep in mind that the results are only as good as the model, and the model is Patchwerk with no latency. So talking about differences smaller than 10-25 DPS is probably still meaningless, even if our numerical accuracy is better than that.

[raistlin212]

Yeah, what he said (twice).

[theckhd]

Weapon Calculation updated again to include the new ZG/ZA weapons. Renataki's is a very nice addition, essentially replacing Soul Blade as the pre-raid tanking weapon of choice.

[theckhd]

I have updated results for the single-target rotation sims, but rather than updating the first post 3 or 4 times, let's spend a little time discussing which queues should be added/removed.

Note that simulation time is basically irrelevant now, as the whole thing takes under a minute, so in theory we could just add any queue we want. The counter-argument is that from an aesthetic point of view, the list is far easier to read if we trim it down to only the queues that we care about.

The counter-counter-argument is that sifting through the raw data and distilling it into a concise commentary section is my job, and that people not interested in the raw data probably skip the table and just read the bullet points I type up anyway, so I should just dump all the raw data in the table for completeness.

Anyhow, here's the raw data for 939/SoT/2%/10:

Code: Select all

                                                  DPS            TPS            SHPS            E     I 
  Q#  Priority                                    V=100%  V=30%  V=100%  V=30%  V=100%  V=30%   %     % 
   1  SotR>CS>AS>J                                14242    9173  42727   27520     0       0    7.3   0.0
   2  SotR>HotR>AS>J                              13125    8360  39374   25079     0       0    7.3   0.0
   3  SotR>CS>J>AS                                14135    9091  42405   27273     0       0    7.4   0.0
   4  SotR>AS>CS>J                                14077    9067  42231   27201     0       0    8.6   0.0
   5  AS>SotR>CS>J                                13947    8985  41842   26954     0       0    9.2   0.0
   6  SotR>CS>AS+>J>AS                            14211    9147  42633   27441     0       0    7.3   0.0
   7  SotR>AS+>CS>J>AS                            14190    9134  42570   27401     0       0    7.7   0.0
   8  sdAS>sdJ>SotR>CS>AS>J                       13726    8830  41178   26490     0       0   11.6   0.0
   9  sdAS>SotR>CS>AS>J                           14107    9087  42322   27261     0       0    8.8   0.0
  10  SotR>CS>AS>J>HW                             14479    9410  43437   28230     0       0    2.8   0.0
  11  SotR>CS>AS>J>Cons>HW                        14673    9515  44021   28546     0       0    1.6   0.0
  12  SotR>AS+>CS>AS>J>Cons>HW                    14655    9505  43964   28515     0       0    1.8   0.0
  13  sdAS>sdJ>SotR>CS>AS>J>Cons>HW               14227    9234  42681   27703     0       0    4.8   0.0
  14  Inq>CS>AS>J                                 12518    8053  37553   24160     0       0    7.7  94.1
  15  Inq>HotR>AS>J                               12034    7627  36101   22880     0       0    7.7  94.1
  16  iInq>SotR>CS>AS>J                           13870    8926  41611   26777     0       0    7.6  72.3
  17  iInq>SotR2>CS>AS>J                          12676    8163  38029   24488     0       0    5.2   0.0
  18  ISotR>Inq>CS>AS>J                           13870    8926  41611   26777     0       0    7.6  72.3
  19  SDSotR>ISotR>Inq>CS>AS>J                    14351    9237  43052   27712     0       0    7.5  47.5
  20  ISotR>SDSotR>Inq>CS>AS>J                    14351    9237  43052   27712     0       0    7.5  47.5
  21  ISDSotR>Inq>CS>AS>J                         13701    8816  41102   26447     0       0    7.6  81.0
  22  Inq>CS>AS>J>Cons>HW                         13088    8506  39264   25520     0       0    1.8  94.1
  23  Inq>HotR>AS>J>Cons>HW                       12604    8079  37812   24240     0       0    1.8  94.1
  24  ISotR>Inq>CS>AS>J>Cons>HW                   14376    9327  43130   27982     0       0    1.7  72.3
  25  SDSotR>ISotR>Inq>CS>AS>J>Cons>HW            14834    9621  44504   28865     0       0    1.7  47.5
  26  SDSotR>ISotR>Inq>CS>AS>J>ICons>HW           14774    9589  44322   28767     0       0    2.1  47.5
  27  SDSotR>Inq>CS>AS>J>Cons>HW                  14679    9524  44038   28572     0       0    1.7  56.4
  28  ISDSotR>Inq>CS>AS>J>Cons>HW                 14232    9237  42696   27713     0       0    1.7  81.0
  29  WoG>CS>AS>J                                 11522    7425  34566   22275  1537    1096   10.5   0.0
  30  WoG>SotR>CS>AS>J                            13046    8404  39137   25212  1268     904    7.5   0.0
  31  WoG>SotR[cdWoG>10]>CS>AS>J                  12220    7874  36660   23623  1519    1083    9.0   0.0
  32  WoG>SotR2[cdWoG>10]>CS>AS>J                 11729    7556  35187   22669  1518    1082    8.7   0.0
  33  WoG>SotR[cdWoG>10]>SotR2[cdWoG>5]>CS>AS>J   11921    7678  35764   23035  1441    1028    7.2   0.0
  34  WoG>SotR[cdWoG>15]>SotR2[cdWoG>10]>CS>AS>J  11729    7556  35187   22669  1518    1082    8.7   0.0
  35  WoG>Inq>CS>AS>J                             12041    7749  36124   23248  1270     906    7.7  65.1
  36  WoG>CS>AS>J>Cons>HW                         12026    7834  36080   23502  1537    1096    3.6   0.0
  37  WoG>SotR>CS>AS>J>Cons>HW                    13482    8750  40446   26252  1268     904    1.7   0.0
  38  WoG>SotR2>CS>AS>J>Cons>HW                   13041    8444  39123   25332     0       0    0.6   0.0
  39  WoG>Inq>CS>AS>J>Cons>HW                     12567    8167  37702   24503  1270     906    1.8  65.1
  40  WoG>Inq>HotR>AS>J>Cons>HW                   11871    7609  35613   22828  1270     906    1.8  65.1
  41  WoG>Inq>AS+>CS>AS>J>Cons>HW                 12582    8181  37748   24543  1253     893    2.1  64.2
  42  WoG>ISotR>Inq>CS>AS>J>Cons>HW               13144    8535  39432   25607  1269     905    1.7  50.3
  43  WoG>iInq>SotR>CS>AS>J>Cons>HW               13144    8535  39432   25607  1269     905    1.7  50.3
  44  SotR>CS>AS>J>HoW                            15437    9998  46310   29995     0       0    1.6   0.0
  45  SotR>CS>AS>HoW>J                            15774   10264  47323   30793     0       0    0.3   0.0
  46  SotR>CS>HoW>AS>J                            15729   10261  47186   30782     0       0    0.3   0.0
  47  SotR>HoW>CS>AS>J                            15597   10192  46791   30575     0       0    1.3   0.0
  48  HoW>SotR>CS>AS>J                            15493   10134  46478   30402     0       0    1.9   0.0
  49  SotR>CS>AS+>HoW>AS>J                        15773   10273  47318   30820     0       0    0.2   0.0
  50  SotR>CS>AS>J>HoW>Cons>HW                    15585   10102  46756   30307     0       0    0.0   0.0
  51  SotR>CS>AS>HoW>J>Cons>HW                    15806   10285  47418   30855     0       0    0.0   0.0
  52  SotR>CS>HoW>AS>J>Cons>HW                    15761   10282  47283   30846     0       0    0.0   0.0
  53  SotR>HoW>CS>AS>J>Cons>HW                    15727   10280  47181   30841     0       0    0.0   0.0
  54  HoW>SotR>CS>AS>J>Cons>HW                    15675   10262  47026   30786     0       0    0.0   0.0
  55  ISotR>SDSotR>Inq>CS>AS>J>HoW>Cons>HW        15869   10288  47609   30865     0       0    0.0  47.5
  56  ISotR>SDSotR>Inq>CS>AS>HoW>J>Cons>HW        16144   10513  48432   31539     0       0    0.0  49.7
  57  ISotR>SDSotR>Inq>CS>HoW>AS>J>Cons>HW        16109   10522  48326   31565     0       0    0.0  50.3
  58  ISotR>SDSotR>Inq>HoW>CS>AS>J>Cons>HW        16161   10580  48484   31741     0       0    0.0  48.8
  59  ISotR>SDSotR>HoW>Inq>CS>AS>J>Cons>HW        16138   10569  48414   31707     0       0    0.0  48.0
  60  HoW>ISotR>SDSotR>Inq>CS>AS>J>Cons>HW        16111   10565  48333   31695     0       0    0.0  51.8
  61  WoG>SotR>CS>AS>J>HoW>Cons>HW                14415    9351  43245   28054  1268     904    0.0   0.0
  62  WoG>SotR>CS>AS>HoW>J>Cons>HW                14675    9560  44027   28679  1268     904    0.0   0.0
  63  WoG>SotR>CS>HoW>AS>J>Cons>HW                14656    9574  43969   28721  1257     896    0.0   0.0
  64  WoG>SotR>HoW>CS>AS>J>Cons>HW                14636    9581  43909   28742  1227     875    0.0   0.0
  65  WoG>HoW>SotR>CS>AS>J>Cons>HW                14546    9532  43639   28596  1273     908    0.0   0.0


For further commentary on 29-34:
-29 doesn't cast SotR at all, and has a mean time between WoGs (MTBW) of 21.0 seconds
-30 has a MTBW of 25.5s and a mean time between successful SotRs (MTBSS) of 17.1 seconds.
-31 has a MTBW of 21.3s and a MTBSS of 39.3s.
-32 has a MTBW of 21.3s and a MTBSS of 24.2s (all SotR2).
-33 has a MTBW of 22.4s and a MTBSS of 15.0s (all SotR2).
-34 has a MTBW of 21.3s and a MTBSS of 24.2s (all SotR2).

It seems obvious to me now that 33 & 34 don't do what I wanted them to - the conditionals should really be SotR2[10>cdWoG>5] and SotR2[15>cdWoG>10] respectively, but the FSM won't parse those as inputs. I haven't had a chance to look at the conditional parsing code yet to see if there's a way to handle a double conditional (cdWoG>5 && cdWoG<10). That said, I'm not sure it's entirely necessary to pursue those options - most of what we want to know can be inferred from 30-32.

30 delays WoG by ~4.5 seconds (compared to 29) to fit in a 3 HP SotR, trading about 270 HPS for 2.5k DPS.
31 delays WoG by a negligible amount, but only manages to fire off about 43% of the SotRs that 30 does. It recovers almost all of the HPS of 29 but beats it in DPS by about 700.
32 is more or less identical to 31 in terms of HPS, but at a significant DPS loss. It's only 200 DPS ahead of 29 for a minimal HPS cost. The poor scaling of SotR2 is probably what makes this a terrible option - even with more frequent SD utilization, you're hitting for far less thanks to the nonlinear scaling of SotR.

My expectation for 33/34 (if I had coded them with the right conditionals, anyhow) was to fall somewhere between 30 and 31. In essence, you'd get 3-point SotRs off whenever you could, but if it looked grim you'd settle for a 2-pointer before starting your next WoG build-up. It's probably safe to say that the DPS increase from this would be very small though, and perhaps more importantly would be more difficult to execute.

In any event, I'm open to suggestions about what queues should be added/removed/etc. For the sims on the first page, I've been using "WoG>SotR>AS>J>Cons>HW" as the default for W39.

[Chicken]

Suggestion 1:
Only host a show of the best queues in the post here, provide a link to a fully detailed summary on a website elsewhere; optimally allow results on said website to be filtered or sorted in various ways. That means all the informations stays available for those interested, but the list of specs to check through for someone wanting quick optimal results isn't as big as it is now.

Suggestion 2:
Split up the "optimal queue" process. First do a bunch of cases with just SotR as finisher, which tests out which is the optimal filler queue to use. Once you've established the optimal filler queue, start examining the various different finisher queues. I think this would cut out a reasonable amount of queues, and as far as I can see browsing through the current large list, the optimal filler isn't dependent on the finishers used.

[inthedrops]

My suggestion:

* List the "practical" queues that make the most sense to stick to. For me, this means a non-Hammer of Wrath, and a Hammer of Wrath queue (for when it becomes usable)

* Provide information, *relative* to the optimal queues as a simple TPS loss value. It would describe how much of a threat loss it is to use SoI and/or WoG instead of SoT and ShoR. This way people can easily know "40k optimal, with SoI I'd lose 8k, cool I can expect about 32k, or whatever). They don't need to see rotations with WoG and SoI in it. They just need to know what kind of loss they can expect.

* Provide an scenario or two that talk about the rotations that are practical, and which might provide a bit more TPS when executed perfectly. But don't recommend them when we're talking 20 extra TPS!

In other words, people want a summary with some tips. I don't think most of us want a list of even 10 rotations. I want the one best rotation, and some sense of how it relates to other options.