A Call to Arms - MoP Mechanics Testing

[Klaudandus]

aaaaaaaaaaaaaaaah~!

[theckhd]

Fitting of Klaud's latest data sets:
CS: expected 130% weapon damage + 913

Code: Select all

cs_fit =

     General model:
     cs_fit(x) = 0.6504*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =         1.3  (1.299, 1.301)
       b =       913.1  (912.1, 914.1)

cs_gof =

           sse: 0.0031
       rsquare: 1.0000
           dfe: 1
    adjrsquare: 1.0000
          rmse: 0.0558

Had to only use the first 3 data points again, due to PvP gloves, but the final 3 data points perfectly match the 5% extra multiplicative scaling. Tooltip confirmed again.

Censure: expected (126+0.11*sp)*5 = 630 + 0.55*sp

Code: Select all

cens_fit =

     General model:
     cens_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.5501  (0.5499, 0.5502)
       b =       629.9  (628.9, 630.9)

cens_gof =

           sse: 0.9455
       rsquare: 1.0000
           dfe: 4
    adjrsquare: 1.0000
          rmse: 0.4862


Tooltip confirmed.

Seal of Truth: expected 14% weapon damage (unnormalized)

Code: Select all

sot_fit =

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

sot_gof =

           sse: 0.0033
       rsquare: 1.0000
           dfe: 5
    adjrsquare: 1.0000
          rmse: 0.0255


No question here, still 14% unnormalized weapon damage


Seal of Righteousness: Expect 5% weapon damage

Code: Select all

sor_fit =

     General model:
     sor_fit(x) = a*x
     Coefficients (with 95% confidence bounds):
       a =     0.04999  (0.04997, 0.05002)

sor_gof =

           sse: 0.0161
       rsquare: 1.0000
           dfe: 6
    adjrsquare: 1.0000
          rmse: 0.0517


I had to exclude one point for this fit, since it was obviously wrong. It was considerably off of the perfect fit to the other 7 data points. I haven't looked into what caused the problem, but to be honest I don't care all that much - it's clear this is working as advertised.


Holy Prism: expected (12412 + 1.098*sp) based on tooltip

Code: Select all

hp_fit =

     General model:
     hp_fit(x) = 1.098*x+b+10000
     Coefficients (with 95% confidence bounds):
       b =        1230  (1198, 1262)

hp_gof =

           sse: 4.6415e+003
       rsquare: 1.0000
           dfe: 5
    adjrsquare: 1.0000
          rmse: 30.4679


hp_fit2 =

     General model:
     hp_fit2(x) = 1.098*x+b+10000
     Coefficients (with 95% confidence bounds):
       b =        3567  (3502, 3633)

hp_gof2 =

           sse: 1.9583e+004
       rsquare: 0.9998
           dfe: 5
    adjrsquare: 0.9998
          rmse: 62.5831

Both very good fits with the advertised scaling. Range of 11230-13567 for base damage gives an average of 12400, which is within 0.1% of the advertised 12412. Tooltip confirmed.

We can also fit the healing component when cast on an enemy:

Code: Select all

hpt_fit =

     General model:
     hpt_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.8235  (0.8212, 0.8257)
       b =        8377  (8362, 8392)

hpt_gof =

           sse: 213.9595
       rsquare: 1.0000
           dfe: 4
    adjrsquare: 1.0000
          rmse: 7.3137

hpt_fit2 =

     General model:
     hpt_fit2(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.8244  (0.8043, 0.8445)
       b =  1.018e+004  (1.005e+004, 1.031e+004)

hpt_gof2 =

           sse: 1.6786e+004
       rsquare: 0.9997
           dfe: 4
    adjrsquare: 0.9996
          rmse: 64.7811

The interesting part here is that we're getting ~(8377-10180)+0.823*SP, when we're expecting 8374+0.74*SP. However, the results we're seeing are consistent with the (8374-10234)+0.823*sp on the other half of the tooltip. In other words, the aoe healing observed matches the formula for the aoe damage we're supposed to be doing when we use Holy Prism to heal a friendly target. In my ability_model file, I already have a formula from an earlier mmo-champ database entry that is consistent with what we're observing in the logs, and I vaguely remember a mmo-champ diff showing that they swapped the formulas on the secondary effects. I'm not sure what happened - whether they swapped it and failed to update the tooltip, or whether they accidentally swapped things in the tooltip when they didn't mean to. Either way, I think this data confirms which formula we're seeing, it's just not clear whether it's a bug or not. Once we have the complementary Holy Prism data I asked for, we'll have a clearer idea of what's going on here.

Eternal Flame: expect (4030-4491)+0.377*sp per holy power

Code: Select all

ef_fit =

     General model:
     ef_fit(x) = 3*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =      0.3771  (0.3702, 0.3839)
       b =        4050  (4007, 4093)

ef_gof =

           sse: 9.1315e+003
       rsquare: 0.9999
           dfe: 3
    adjrsquare: 0.9999
          rmse: 55.1710


ef_fit2 =

     General model:
     ef_fit2(x) = 3*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =      0.3798  (0.3716, 0.388)
       b =        4450  (4398, 4502)

ef_gof2 =

           sse: 1.3121e+004
       rsquare: 0.9999
           dfe: 3
    adjrsquare: 0.9998
          rmse: 66.1338

We don't quite reach each extreme thanks to the limited nature of the data set, but it's good enough to feel pretty confident that the tooltip isn't lying to us.

We also fit the ticks, which should be 1393*0.16*sp per holy power:

Code: Select all

eft_fit =

     General model:
     eft_fit(x) = 3*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =        0.16  (0.1599, 0.16)
       b =        1393  (1393, 1393)

eft_gof =

           sse: 0.4090
       rsquare: 1.0000
           dfe: 3
    adjrsquare: 1.0000
          rmse: 0.3692


Tooltip confirmed.


Seal of Insight: expect 0.15*ap+0.15*sp = 0.45*sp

Code: Select all

soi_fit =

     General model:
     soi_fit(x) = a*x
     Coefficients (with 95% confidence bounds):
       a =        0.45  (0.4499, 0.45)

soi_gof =

           sse: 0.8189
       rsquare: 1.0000
           dfe: 7
    adjrsquare: 1.0000
          rmse: 0.3420


Tooltip confirmed.

[Klaudandus]

The reason why Prism doesn't fit when it comes to the healing is because the prism data was from before the latest build, where it saw a nerf. Only the healing part of EF was nerfed on the latest builds (the same one where CS and Censure got hit)

The Prism data was collected 2 days before that build.

[theckhd]

Was that a stealth-nerf? Because it doesn't show up on mmo-champ's beta build diff.

<edit> nevermind, I found it in build 15851, the one before the CS nerfs.

If you don't mind posting the build # with data sets from now on, that might be helpful in avoiding this sort of confusion.

[theckhd]

Another interesting note: apparently EF, HP, and SoI's healing aren't affected by the 5% increase to healing spells granted by SoI. Not sure if that's a bug or intended (hopefully intended, since I would rather not be tanking with SoI).

[Klaudandus]

Was that a stealth-nerf? Because it doesn't show up on mmo-champ's beta build diff.

<edit> nevermind, I found it in build 15851, the one before the CS nerfs.

If you don't mind posting the build # with data sets from now on, that might be helpful in avoiding this sort of confusion.

sounds good.

[theckhd]

Blue confirmation of armor DR formula: http://us.battle.net/wow/en/forum/topic ... e=130#2593

DamageReduction = TargetArmor / (TargetArmor + 4037.5*AttackerLevel – 317117.5)

Very consistent with the values I worked out empirically based on melee data. Changes the level 93 dummy reduction factor from 0.6504 to 0.6507, which has no effect on the CS data.

[Klaudandus]

Here you go. Some WoGs were with 2HP, hope it doesnt affect much, and I was also pressing WoG on Divine Purpose proc, at least most of the time

[theckhd]

Here you go. Some WoGs were with 2HP, hope it doesnt affect much

Well, it'll throw off all of the averages and make it annoying to pick out min and max values... but I can probably still get what I need out of it.

[Klaudandus]

blah, I was doing it kinda hurried out as I have a raid (should be pulling soon)...

i can redo the logs tomorrow if you prefer not to work with these...

[Klaudandus]

I can't log on to the beta for some reason, it just hungs up

[Klaudandus]

I'm almost sure that there were no WoG with more than 1 HP in these new sets, and I created a macro that cancels Divine Purpose before casting WoG.

These can also be combined wit the previous log in order to get more data regarding Holy Prism, considering that the cd on that one is 20 secs...

Only did naked, 1, 2 and 7 for the time being... going out to jog, hope I don't get bit by Bruce again.

Once I get back, I'll do set 4 and 6

[theckhd]

Two more tests added regarding the L90 talents (#33 and #34)

[Klaudandus]

Theck, question... what do you mean by

Goal: 60 ES/LH casts, 8 LH casts.

Also, just got Colossus and Windsong -- where can I find a 1.4 weapon?

[Klaudandus]

Set 4 and 6 for WoG test