A Call to Arms - Cataclysm Mechanics testing

[Staryx]

This is using Human base stats, because I don't have the updated level 80 base Draenei or BE stats. In fact, I'd like to double-check all of the racial base stats (I suspect the STA value I have for humans is wrong), so if you have a level 80 paladin of any race on the PTR, please post the naked STR/STA/AGI/INT/SPI values.

Base Blood Elf Stats @ lvl 80 (no spec chosen)

Str: 148
Agi: 92
Stam: 203
Int: 101
Spi: 103

[angelflavor]

here is a HoL run I did. I AoE tanked it like I would in live. http://www.worldoflogs.com/reports/6a6dffa0pbkx1k5r/

[Kierly]

Test #1: Holy Wrath

Done with 0/36/0 spec only missing points in imp hammer of justice

Code: Select all

 
         # of targets
sp       1     2     3
178    2262  1157   789
490    2448  1298   914
1123   2827  1581  1166
1494   3051  1750  1315
2014   3363  1983  1524


Test #2: Crusader Strike

Same Talent Build as before, using Dalaran sword on the 80 dummy.
AP / SP parse
522 / 178 http://www.worldoflogs.com/reports/l0koreniv3y8neag/details/14/
1516 / 476 http://www.worldoflogs.com/reports/6ah5m5xk64qyenva/details/21/
1664 / 934 http://www.worldoflogs.com/reports/4awsa5rag1g3q5ud/details/12/
2400 / 1155 http://www.worldoflogs.com/reports/0mygxaselsxipxqd/details/13/
2966 / 1211 http://www.worldoflogs.com/reports/zkuazchxir5pbctu/details/14/
3714 / 1146 http://www.worldoflogs.com/reports/72n898p8ckvctlgc/details/8/

[Staryx]

New requests:

Test #1: Holy Wrath
This one is easy, since the damage is a fixed amount. Find a dummy with no debuffs, and record Holy Wrath damage for various spellpower levels. I.e. naked, ~500 sp, ~1000 sp, ~1500 sp, etc. Gear is basically irrelevant, since all we need is SP and HW damage. Since HW hits for a fixed amount, you don't even need more than one data point for each SP value, so this can be done with a pencil and paper instead of WoL. Make sure you note your spec - 0/38/0 preferred, but I'll take what I can get.

Then repeat the test from a spot where you can hit 3 dummies, again, no debuffs.

The level of the dummy should be irrelevant, but post it anyway just to be safe.

Current spec on PTR: http://cata.wowhead.com/talent#sZGMhfRdkRRuc

Code: Select all

Source: lvl 80 BE
Dummy level: 80 (1 dummy)   80/70/60 (3 dummies)
SP      DMG (1 dummy)   DMG (3 dummies)
179     2262            789
281     2322            830
969     2736            1106
1018    2765            1125
1276    3065            1290
1341    3107            1317

Recorded only non-crit HW casts. 3 dummy tests were run on 1 of each level dummy (best I could get in Undercity). My apologies if the SP numbers aren't varied enough, only had Prot and Ret gear to test with and Kings/No Kings as well. I can mix up my gear more and test again if needed. Also, the 179 test is naked, if that matters for a baseline.

[Klaudandus]

0/34/0 Spec lvl 80 BE

Code: Select all

Stats
Str - 295
Agi - 92
Stamina - 895
Int - 270
Spirit - 221

Damage - 143-179
DPS - 115.6
[b]AP - 1010[/b]
Speed - 1.39
Haste - 0.52%
Hit - 1.72%
Crit - 4.8%
Exp - 3
Mastery - 9.63

[b]SP - 497[/b]
Haste - 0.52%
Hit - 2.02%
Crit - 7.33%

#Case 1+2 (2 Dummies)
http://www.worldoflogs.com/reports/1lebwz97gy5vmrsg/

#Case 1+3 (2 Dummies)
http://www.worldoflogs.com/reports/gr9r9c3cfqvcadk6/
In the last seconds of the log, I had a boomy come in and nuke the 2nd dummy T_T

Rats.. .can't get a 2000AP/1000SP without either getting a 5% all damage bonus or HotR 5% damage bonus >=/

[Arianne]

Code: Select all

Source: lvl 80 Dwarf
Dummy level: 60 (1 dummy)   80/70/60 (3 dummies)
SP      DMG (1 dummy)   DMG (3 dummies)
1956    3326 (4990C)    1499 (2250C)
2147    3441 (5162C)    1576 (2365C)
2343    3559 (5388C)    1655 (2481C)
2504    3655 (5484C)    1719 (2579C)
2781    3821 (5733C)    1829 (2745C)
3016    3963 (5945C)    1923 (2885C)
3251    4104 (6155C)    2018 (3028C)
3520    4265 (6399C)    2125 (3188C)
3703    4375 (6564C)    2199 (3298C)

0/36/0 build on the PTR with kings up.

[Klaudandus]

0/34/0 lvl 80 BE

Code: Select all

Stats
Str - 875
Agi - 92
Stam - 2242
Int - 466
Spirit - 311

Melee - 244-280
Dps - 192.3
[b]AP - 2020[/b]
Speed - 1.36
Haste - 2.68%
Hit - 0.91%
Crit 14.95%

[b]SP - 981[/b]
Haste - 2.68%
Hit - 1.07%
Crit - 18.67%

Caveat - getting the 2000AP/1000AP was a paint in the ass. Ended up with the +50 AP bonus from the pvp set. Better than the +5% all damage or the +5% HotR damage. I think.

Case 1+2 (2 Dummies)
http://www.worldoflogs.com/reports/xa5boddadusynahb/

Case 1+3 (2 Dummies)
http://www.worldoflogs.com/reports/z8noluje1yi7rbf0/

Sorry for the short duration of the HotR case - the cata event started midway thru it. Will resume after event.

[theckhd]

Caveat - getting the 2000AP/1000AP was a paint in the ass. Ended up with the +50 AP bonus from the pvp set. Better than the +5% all damage or the +5% HotR damage. I think.

Maybe I should have been clearer - I don't actually need data points at the exact values of AP and SP I put in my request. Those numbers were just made up to give an example data format. I just need data points covering a large range of both AP and SP.

Thanks for all of the data everyone, I'll try and get crunching on it tonight, but I have a raid in an hour and a half, so some of it might have to wait until tomorrow. Then we can get to the bottom of these last few outstanding issues (at least until they push a new patch and change something else...).

Also, if anyone has time to do some quick tests of Hammer of Wrath or Exorcism, we're sorely lacking data for those. Similar set-up to the previous tests, just need average damage at a few SP/AP combinations.

[theckhd]

Base Blood Elf Stats @ lvl 80 (no spec chosen)

Str: 148
Agi: 92
Stam: 203
Int: 101
Spi: 103

I'm guessing you're a miner? 203 is way too high for a base stam value. But 143 would be right in line with the other races, and mining apparently gives +60 stam...

[Klaudandus]

Caveat - getting the 2000AP/1000AP was a paint in the ass. Ended up with the +50 AP bonus from the pvp set. Better than the +5% all damage or the +5% HotR damage. I think.

Maybe I should have been clearer - I don't actually need data points at the exact values of AP and SP I put in my request. Those numbers were just made up to give an example data format. I just need data points covering a large range of both AP and SP.

Thanks for all of the data everyone, I'll try and get crunching on it tonight, but I have a raid in an hour and a half, so some of it might have to wait until tomorrow. Then we can get to the bottom of these last few outstanding issues (at least until they push a new patch and change something else...).

Also, if anyone has time to do some quick tests of Hammer of Wrath or Exorcism, we're sorely lacking data for those. Similar set-up to the previous tests, just need average damage at a few SP/AP combinations.

0/34/0 lvl 80 BE
Hammer of Wrath on Grol: 20/10

Code: Select all

Stats
Str - 875
Agi - 92
Stam - 2242
Int - 466
Spirit - 311

Melee - 244-280
Dps - 192.3
AP - 2020
Speed - 1.36
Haste - 2.68%
Hit - 0.91%
Crit 14.95%

SP - 981
Haste - 2.68%
Hit - 1.07%
Crit - 18.67%

http://www.worldoflogs.com/reports/essvi3j028394xkl/

---

Hammer of Wrath on Grol: 10/5

Code: Select all

str - 395
agi - 92
sta - 770
int - 194
spi - 171

damage 156 - 194
dps - 124.8
ap - 1010
speed - 140
haste - 0
hit 2.63
crit - 3.71

sp - 421
hit - 3.09
crit - 5.79

somehow couldnt get the same set as what I had on the previous case for CS/HotR.
http://www.worldoflogs.com/reports/5dkds5mpdp1ljdsc/

----

Hammer of Wrath on Grol: Naked
Had to use Seal of Insight and Flash of Light to keep myself alive for 10 mins during the last parse. Hope it doesn't affect the numbers.
http://www.worldoflogs.com/reports/yir113iuort8dl8h/

[Staryx]

Base Blood Elf Stats @ lvl 80 (no spec chosen)

Str: 148
Agi: 92
Stam: 203
Int: 101
Spi: 103

I'm guessing you're a miner? 203 is way too high for a base stam value. But 143 would be right in line with the other races, and mining apparently gives +60 stam...

Doh! Yeah, forgot about that. Yep, got 450 mining and yeah, it's +60 Stam.

[theckhd]

Holy Wrath analysis:
The data sets from Kierly and Arianne are nice and linear, and cover the range I need. Staryx's data set seems good up until the last two points from both tests (i.e. the 1276 and 1341 SP data points for both 1 and 3 targets). Those two damage values are discontinuous and don't agree with the other two data sets:



The blue line is Staryx's data set, the red and green lines are Kierly and Arianne respectively. These are the single-target damage values, but the same effect happens in the 3-target tests. From this we have to conclude that something about those two gear sets is affecting the HW damage, and ignore those two points.

Fitting the combined data set (<3 cftool) gives us the following results:

Code: Select all

Linear model Poly1:
       f(x) = p1*x + p2
Coefficients (with 95% confidence bounds):
       p1 =      0.5995  (0.5991, 0.6)
       p2 =        2154  (2153, 2155)

Goodness of fit:
  SSE: 6.555
  R-square: 1
  Adjusted R-square: 1
  RMSE: 0.7391

Ergo, the formula of Holy Wrath damage should be (2154+0.6*SP) for a single target. Note that this is twice the SP coefficient in the model, which is a little weird (model was 2420+0.3*SP).

Repeating for the 3-target data gives:

Code: Select all

Linear model Poly1:
       f(x) = p1*x + p2
Coefficients (with 95% confidence bounds):
       p1 =      0.3999  (0.3995, 0.4002)
       p2 =       717.6  (716.9, 718.4)

Goodness of fit:
  SSE: 4.078
  R-square: 1
  Adjusted R-square: 1
  RMSE: 0.583


In other words, for 3 targets the damage is (717.6+0.4*SP). Interesting. Base damage is round(2154/3) for 3 targets, but the SP coefficient isn't just 0.6/3.

Fitting Kierly's data for 2 targets gives us a base damage of 2154/2 and a coefficient of 0.45.

So the coefficient is decreasing as number of targets increases, but not by a straight factor of 1/N. On a whim, I threw this scaling into cftool and tried fitting it with a 1/x form, since it looks roughly like that. Sure enough, here's what it spit out:

Code: Select all

General model:
       f(x) = a/x+b
Coefficients (with 95% confidence bounds):
       a =         0.3  (0.3, 0.3)
       b =         0.3  (0.3, 0.3)

Goodness of fit:
  SSE: 7.452e-020
  R-square: 1
  Adjusted R-square: 1
  RMSE: 2.73e-010

With only 3 data points, it's hard to be sure, but it's a reasonably good bet that the new formula @ level 80 is:

damage=2154/N+(0.3/N+0.3)*SP

What does this mean? Personally, I think it's a bug. The formula before was 2154/N+0.3*SP, which has the weird "feature" that it calculates damage per mob from the base and then applies the spellpower bonus damage to each mob. I expected them to fix it by changing the formula to (2154+0.3*SP)/N; in other words, calculate total damage first, then divide it all by the number of mobs to determine damage per mob. That makes perfect sense.

Instead, we now have (2154+0.3*SP)/N + 0.3*SP. In other words, we have both versions of the formula simultaneously. My guess is that they tried to change the spell to split the spellpower bonus across all mobs, but didn't actually fix whatever bug is causing the extra spellpower damage to be applied to each mob individually.

In any event, right now we know how it acts. We'll have to check this again after a future patch to see if it's fixed.

If anyone wants to help confirm this, just find an area of dummies where you can hit a large number of targets and see what your damage is on 1, 2, 3, 4, ... dummies. If we keep getting the 0.3*(1+1/N) scaling, we should expect to see the total damage increase by 0.3*SP for each additional dummy you hit.

[Klaudandus]

Out of curiosity Theck, does my HW data fit the linear equation as well? I was always hitting 2 dummy targets as well

[theckhd]

Out of curiosity Theck, does my HW data fit the linear equation as well? I was always hitting 2 dummy targets as well

I haven't had time to look, since I was going to check yours when I was looking at the CS/HotR data. But I have the new model incorporated into the code now, so when I get to that, it should be pretty apparent. If you could annotate your posts with the number of dummys you were hitting that would help save some time, otherwise I have to use the Log Browser to check that.

<edit> obviously you just said how many dummies, but it'll be easier for me to remember it tomorrow if it's in the post with the data links.

[theckhd]

Kierly's CS data yields no correlation with SP (as expected) and the following with AP:

Code: Select all

Linear model Poly1:
       f(x) = p1*x + p2
Coefficients (with 95% confidence bounds):
       p1 =      0.2018  (0.2, 0.2035)
       p2 =       70.58  (66.47, 74.68)

Goodness of fit:
  SSE: 10.05
  R-square: 1
  Adjusted R-square: 1
  RMSE: 1.585

71 is the base damage you'd get from a dalaran sword with 0 AP in that talent spec against an 80 dummy, so that's good.

The formula should be
1.2*(base+X*ap/14)*mdf.phdmg*mdf.Crus*(1+2*mdf.WotL)
where X is either 2.4 (for normalized attacks) or the player's swing speed (for non-normalized attacks, 1.4 for dalaran sword). 1.2*mdf.phdmf*(1+2*mdf.WotL0 works out to be 1.1717.

Thus, p1=1.1717*X/14 -> X=14*p1/1.1717=2.4112. Close enough to be sure that CS is still a normalized attack (the exact coefficient would be 0.2009, which is within the 95% confidence interval).

So CS is coded properly already, which means that something else is causing the weird effects we were seeing earlier. It might be a sample size issue or an interfering debuff. I'll continue checking the rest of the CS parses submitted today to make sure that none of them are seeing the erroneous behavior.