Skip to content
Maintankadin

A Call to Arms - MoP Mechanics Testing

Klaudandus
Thu Jul 05, 2012 6:39 am

Re: A Call to Arms - MoP Mechanics Testing

I've only been able to find lvl 85 dummies in orgrimmar

but i just tested Judgement on a lvl 93 and a lvl 85 dummy and the difference is -/+ 1, as you said, so should I stick to lvl 93 dummies instead?
theckhd
Thu Jul 05, 2012 6:46 am

Re: A Call to Arms - MoP Mechanics Testing

Yeah, given that information, 93 dummies can be used for everything. In fact, you could use any level dummy for the spell damage ones, it's only the weapon/physical damage ones (melee, CS, HotR) that absolutely have to be level 93 (because armor values change with dummy level).

I'm trying to get what I can out of the data sets you provided as we speak, using tweaked spellpower values (modifying the third one to be exactly half of the AP value). I think I've convinced myself that your Judgement data is consistent with the tooltip values (though again, a low- or zero-sp data point would help that a lot).

<edit> after some more testing, I think that the AP values from sets 2-4 are correct, but the SP values from 2 & 3 are off (as indicated by the ratios). In addition, both values from set 1 are wrong (both too high). If I fit vs. AP, values from sets 2-5 all fit very, very well with the tooltip formula for Judgment. But the first data point is always very low, and drags the intercept down. So far, I'm concluding that I can use sets 2-5 under the assumption that the AP data is correct, but I'll have to toss set 1 unless further data suggests it's correct.
Klaudandus
Thu Jul 05, 2012 6:58 am

Re: A Call to Arms - MoP Mechanics Testing

don't worry, I'm doing new sets -- just to verify things

Here is the first one.

Naked
Str 176
Agi 107
Sta 332
Int 117
AP 602
SP 301
Haste 0
Hit 0
Crit 5.01
Exp 0

The problem I had with the first set 1 was that I used an epic shield -- that's why it's out of whack
Klaudandus
Thu Jul 05, 2012 9:32 am

Re: A Call to Arms - MoP Mechanics Testing

Code: Select all
         set #1   set #2  set #3  set #4
Str      1624     2953    4481    5931
Agi      107
Sta      3244     5675    8587    11365
Int      117   
AP       3498     6156    9212    12112
SP       1749     3078    4606    6056
Haste    0        0.76    3.07    3.07
Hit      0        0.62    0.62    1.97
Crit     5.36     5.36    6.30    7.31
Exp      2.55     3.89    3.89    3.89


I'll have more sets after the server reset.
Klaudandus
Thu Jul 05, 2012 10:43 am

Re: A Call to Arms - MoP Mechanics Testing

Code: Select all
      set #5  set #6  set #7
Str   7580    8690    10566
Sta   15355   17365   
AP    15410   17630   21382
SP    7705    8815    10691
Haste 4.40    4.40
Hit   1.97    3.37
Crit  8.53    8.53
Exp   3.89    3.89

Klaudandus
Thu Jul 05, 2012 10:44 am

Re: A Call to Arms - MoP Mechanics Testing

Set 6 and 7 data
theckhd
Thu Jul 05, 2012 2:11 pm

Re: A Call to Arms - MoP Mechanics Testing

Thanks Klaud, I've entered all of the data in and am starting to fit it. So far the static abilities are fitting beautifully. I haven't started on the ones that depend on weapon damage yet, probably tomorrow or this evening.

Weapon was still a club, right?
Klaudandus
Thu Jul 05, 2012 2:15 pm

Re: A Call to Arms - MoP Mechanics Testing

yes, club.

Let me know if you need granularity for the tests, the way I did it is that each set increments the pieces of gear by 2. So I start naked, then I add helm and neck, then next set is shoulders and cloak and so on -- let me know if you need odd number of pieces equipped just to verify and I'll get on that.
theckhd
Thu Jul 05, 2012 2:38 pm

Re: A Call to Arms - MoP Mechanics Testing

Are you sure you used a club for the first round of testing? Because something isn't adding up. For example, this time around, in set #7 with 21k AP, your CS is hitting for 4777-4780. In set #5 from a few weeks ago, with 19.8k AP, you were hitting for 10713-15790. Not only are the values much larger, but the range is much larger. I haven't had time yet to figure out what I expect from the formulas, but just looking at the data, something looks fishy.
Klaudandus
Thu Jul 05, 2012 2:59 pm

Re: A Call to Arms - MoP Mechanics Testing

A problem I had with the sets I submitted a while back is that I didn't use a white shield, I used an epic shield... The original request had them asking for club as your weapon, they didn't mention a white quality shield -- but I did use one for today's sets

It's possible I fucked up something on the previous sets -- or rather, it's almost certain -- considering I just logged in back to beta and verified that CS is giving me ~4778 damage

I'm thinking I probably equipped the epic pvp mace by accident

EDIT: Just went back and equipped the mace and the shield, and removed my str trinkets... I hit CS and almost matches the results of #5 from last week... I fail orz
tlitp
Thu Jul 05, 2012 3:39 pm

Re: A Call to Arms - MoP Mechanics Testing

Theck : Sure enough, we've got blocked criticals. Much to my own amusement, even blocked glancing blows (which weren't there in early beta, IIRC). At this point it seems they went with the easy route, simply injecting another roll for any event that happens to deal damage in the first place.


Yet another glorious example of misinformation. Gotta love the man. Hell, I wonder how many people still believe that parryhaste is gone "because GC said so". :lol:
theckhd
Fri Jul 06, 2012 4:47 am

Re: A Call to Arms - MoP Mechanics Testing

Klaudandus wrote:It's possible I fucked up something on the previous sets -- or rather, it's almost certain -- considering I just logged in back to beta and verified that CS is giving me ~4778 damage

I'm thinking I probably equipped the epic pvp mace by accident

EDIT: Just went back and equipped the mace and the shield, and removed my str trinkets... I hit CS and almost matches the results of #5 from last week... I fail orz

That's fine, as long as the AP/SP values are correct. I'll just exclude the old results from my fits for weapon-damage-based abilities. The new ones should span the space well enough to give me good results.
Klaudandus
Fri Jul 06, 2012 5:44 am

Re: A Call to Arms - MoP Mechanics Testing

good, just let me know if you need anything else, and g'luck with the moving.
theckhd
Tue Jul 10, 2012 7:29 am

Re: A Call to Arms - MoP Mechanics Testing

Just checking in - I'm not dead, the move went well, and I'm hoping to get some time to finish up the spell coefficient calculations in the next day or two. I'm over halfway done, but haven't gotten to look at them since last Thursday. So far, all of the ones I've finished match the wowdb tooltips exactly, which was more or less expected.
Klaudandus
Wed Jul 11, 2012 5:43 am

Re: A Call to Arms - MoP Mechanics Testing

X-posted because it might be relevant
Q:SOTR hits for about 1/4th of what it hits for on live. CS/HOTR also hit for much, much less.

A:If you post approximate damage numbers for those attacks along with your character level and average ilevel (just the average is fine), we can compare them to our numbers.

http://us.battle.net/wow/en/forum/topic ... 430?page=1

Some reports have prot pallies doing about 60% of a Blood DK, and (apparently due to a bug) 22% of what brewmasters do it.
theckhd
Wed Jul 11, 2012 4:30 pm

Re: A Call to Arms - MoP Mechanics Testing

Your SotR data exactly matched the tooltip, 617+0.54*AP.
Klaudandus
Wed Jul 11, 2012 4:59 pm

Re: A Call to Arms - MoP Mechanics Testing

theckhd wrote:Your SotR data exactly matched the tooltip, 617+0.54*AP.


Any idea why CS is hitting like a wet noodle?
theckhd
Wed Jul 11, 2012 5:09 pm

Re: A Call to Arms - MoP Mechanics Testing

Not yet, I'm looking at that now. That was the ability I was working on when I had to pack up, and I didn't get a chance to work out the new armor reduction factor before I did that. I was hoping that would fix the discrepancy between data and tooltip.

<edit> but I can't seem to find level 93 armor values. I know I've seen them somewhere recently, but damned if I can find them now that I need them...
theckhd
Wed Jul 11, 2012 5:52 pm

Re: A Call to Arms - MoP Mechanics Testing

Ok, can't find the armor data, but I can extrapolate based on the melee data. From the max melee swings, I can infer that the physical damage reduction factor [K/(A+K), where K is the armor coefficient tlitp found earlier (25050) and A is the target armor] is around 0.6545. Fitting the CS data with
y=0.6545*(a*x+1123)

gives me:
Code: Select all
General model:
     f(x) = 0.6545*(a*x+1123)
Coefficients (with 95% confidence bounds):
       a =       1.676  (1.648, 1.704)

Goodness of fit:
  SSE: 1.591e+004
  R-square: 0.9988
  Adjusted R-square: 0.9988
  RMSE: 47.67

which isn't too bad given that my 0.6545 is just a rough estimate. We expect the coefficient to be 1.6 based on the tooltip.
Klaudandus
Wed Jul 11, 2012 5:56 pm

Re: A Call to Arms - MoP Mechanics Testing

tangentially related

Yeah, we will try to add dodge from Agi to the Agi and dodge tooltips. Currently they’re just included in the title of each of those tooltips, which are final values after summing all of the sources and applying DR. The “(before diminishing returns)” line in the dodge tooltip, refers just to the “dodge of X adds Y% dodge” line, not the total. It’s somewhat confusing right now. All of this goes for Strength / parry as well.
theckhd
Wed Jul 11, 2012 6:21 pm

Re: A Call to Arms - MoP Mechanics Testing

All right, I've validated almost all of the spell details now. Luckily, they all pretty much match the tooltip values (if ones are given). Here's the breakdown for each spell. I'm skipping melee/CS/HotR for the moment, as those three depend on the exact armor mitigation factor. Data and fit information is here: https://sites.google.com/site/matlabadi ... fit_data.m

HammerNova:
Tooltip: (840-1261) + 0.234*AP = (840-1261) + 0.468*AP
Fits (min and max):
Code: Select all
hanova_fit =

     General model:
     hanova_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.4685  (0.4676, 0.4694)
       b =        1254  (1248, 1260)

hanova_gof =

           sse: 181.8413
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 4.2643


hanova_fit2 =

     General model:
     hanova_fit2(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =       0.468  (0.4665, 0.4696)
       b =       847.9  (837.1, 858.7)

hanova_gof2 =

           sse: 576.4841
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 7.5927

Result: data seems entirely consistent with tooltip values. Tooltip confirmed.

Seal of Truth:
Tooltip: 14% weapon damage
Fit:
Code: Select all
sot_fit =

     General model:
     sot_fit(x) = a*x
     Coefficients (with 95% confidence bounds):
       a =        0.14  (0.14, 0.1401)

sot_gof =

           sse: 0.0533
       rsquare: 1.0000
           dfe: 7
    adjrsquare: 1.0000
          rmse: 0.0872

Result: Tooltip confirmed.

Censure:
Tooltip: (158+0.138*SP)*5 = 790 + 0.69*SP
Fit:
Code: Select all
cens_fit =

     General model:
     cens_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.6901  (0.69, 0.6901)
       b =       790.2  (789.6, 790.8)

cens_gof =

           sse: 1.6780
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 0.4096

Result: Tooltip confirmed

Judgment:
Tooltip: 702 + 0.397*AP + 0.635*SP = 702 + 1.429*SP
Fit:
Code: Select all
j_fit =

     General model:
     j_fit(x) = 702+a*x+b
     Coefficients (with 95% confidence bounds):
       a =       1.429  (1.429, 1.429)
       b =     -0.2869  (-1.154, 0.5804)

j_gof =

           sse: 3.7250
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 0.6103

Result: Tooltip confirmed.

Holy Wrath:
Tooltip: 4300+0.91*SP
Fit:
Code: Select all
hw_fit =

     General model:
     hw_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =        0.91  (0.91, 0.9101)
       b =        4301  (4300, 4301)

hw_gof =

           sse: 1.3406
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 0.3661

Result: Tooltip confirmed.

Consecration:
Tooltip: No scaling info provided
Fit:
Code: Select all
cons_fit =

     General model:
     cos_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =        0.12  (0.12, 0.12)
       b =       68.98  (68.89, 69.07)

cons_gof =

           sse: 0.0400
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 0.0632

Result: Cons is 69+0.12*SP damage per tick, total of 690+1.2*SP per cast.

Avenger's Shield:
Tooltip: (4039-4937)+0.21*SP+0.545*AP = (4039-4937)+1.30*SP
Fits (min and max):
Code: Select all
as_fit =

     General model:
     as_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =       1.299  (1.296, 1.303)
       b =        4059  (4033, 4085)

as_gof =

           sse: 3.2986e+003
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 18.1621


as_fit2 =

     General model:
     as_fit2(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =       1.302  (1.297, 1.307)
       b =        4895  (4861, 4930)

as_gof2 =

           sse: 5.9482e+003
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 24.3889

Result: Data seems entirely consistent with tooltip (to 95% CI). Tooltip confirmed.

Shield of the Righteous
Tooltip: 617 + 0.54* AP = 617 + 1.08*SP
Fit:
Code: Select all
sotr_fit =

     General model:
     sotr_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =        1.08  (1.08, 1.08)
       b =       616.8  (616.1, 617.5)

sotr_gof =

           sse: 2.6125
       rsquare: 1.0000
           dfe: 10
    adjrsquare: 1.0000
          rmse: 0.5111

Result: Tooltip confirmed.
theckhd
Wed Jul 11, 2012 6:24 pm

Re: A Call to Arms - MoP Mechanics Testing

By the way, here's what I have for melee, CS, and HotR (phys). I'm using the melee max data to determine the armor coefficient, and using that to fit CS and HotR:
Code: Select all
melee_fit =

     General model:
     melee_fit(x) = a*x
     Coefficients (with 95% confidence bounds):
       a =      0.6545  (0.6453, 0.6637)

melee_gof =

           sse: 2.5220e+003
       rsquare: 0.9991
           dfe: 7
    adjrsquare: 0.9991
          rmse: 18.9814


cs_fit =

     General model:
     cs_fit(x) = 0.6545*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =       1.707  (1.661, 1.753)
       b =        1043  (942.4, 1143)

cs_gof =

           sse: 9.7370e+003
       rsquare: 0.9993
           dfe: 6
    adjrsquare: 0.9991
          rmse: 40.2844


hotr_fit =

     General model:
     hotr_fit(x) = 0.6545*a*x
     Coefficients (with 95% confidence bounds):
       a =      0.1992  (0.1989, 0.1995)

hotr_gof =

           sse: 0.8558
       rsquare: 1.0000
           dfe: 7
    adjrsquare: 1.0000
          rmse: 0.3496


CS is a bit high at 1.7x weapon damage instead of 1.6x. Note that CS is based off of normalized weapon damage, while HotR is not. Once I have level 93 armor values, I'll use the exact values and see if I can do better.
theckhd
Wed Jul 11, 2012 6:30 pm

Re: A Call to Arms - MoP Mechanics Testing

By the way, I wouldn't mind a data set with which to test Seal of Righteousness/Insight, Hammer of Wrath, Holy Prism, Execution Sentence, and Light's Hammer. Bonus points if you can test Sacred Shield / Eternal Flame somehow.
Klaudandus
Wed Jul 11, 2012 6:35 pm

Re: A Call to Arms - MoP Mechanics Testing

Ok, I'll work on friday, or maybe tomorrow afternoon. (Got a raid in 20 mins =P)
Should I do the Naked + 7 for each Seal? Would 10 minutes of data per seal be enough or should I do 15 at least?