A Call to Arms - MoP Mechanics Testing

Should be much shorter than that. Seal of Righteousness should be basically constant with a Club, so just a few swings at each gear level will be necessary. Similarly with Insight. Hammer of Wrath has a damage range, so that will take more samples (50-100 casts in each gear set, so 5-10 minutes with each gear set). Light's Hammer and Execution Sentence seem to be constant, so 2-3 casts of each should be sufficient. Holy Prism has a damage range, so that's another one we'll need multiple casts of (and that one will be painful due to the long cooldown).

Ideally we'd have Naked, fully geared, and a few data points in-between. The extreme ends (naked and fully geared) are the most important.

ok, silly question, how can I test Hammer of Wrath? I remember there used to be an undying mob in blasted lands but was removed when cata arrived.

any idea on what to do?

I think you can use the dummies at the argent tournament, if my memory is correct.

theckhd wrote:I think you can use the dummies at the argent tournament, if my memory is correct.

oh yeah! i guess it doesnt matter that they're lvl 80, right?

It shouldn't, since HoW does spell damage rather than physical.

lvl 93 dummy armor - 24835
at least according to someone on the forums

caveat emptor - it's not official

Well, that isn't consistent with my earlier estimations. That gives me an armor mitigation factor of:

Code: Select all

>> 1-24835/(24835+25050)

ans =

    0.5022


Which is a lot lower than the 0.6545 we get from the melee swing data.

Yeah, I didn't get a response from GC when I asked straight out in the beta analysis thread on the forums -- that's what I don't get about Blizz, they're asking for your analysis but they don't give out data that you guys can use to make all analysis better/easier.

It's not like it would be too hard for them to say "Lvl 93 dummy has this much armor"

I got the question regarding the prism -- what dummy should I use for that one, the lvl 93 dummy?

GC just confirmed the armor value for lvl 93 dummies

And yet, you never see Zeherah and GC anywhere at the same time. Coincidence? In any case, the number 24835 is correct.

http://us.battle.net/wow/en/forum/topic ... ge=93#1853

Klaudandus wrote:I got the question regarding the prism -- what dummy should I use for that one, the lvl 93 dummy?

Any dummy should be fine, since it does spell damage.

OK, I think I've found the problem. Looking back at the data set you took for tlitp, I get a very similar physical damage reduction factor:

Code: Select all

Linear model Poly1:
     f(x) = p1*x + p2
Coefficients (with 95% confidence bounds):
       p1 =      0.6504  (0.6502, 0.6505)
       p2 =           0  (fixed at bound)

Goodness of fit:
  SSE: 2.348
  R-square: 1
  Adjusted R-square: 1
  RMSE: 0.425

This is fitting observed max melee damage vs. theoretical max melee damage with a pure linear fit. The armor constant that tlitp came up with is what I get if I use the Cataclysm npc armor formula (which predicts 13452 armor for a level 93). GC's comment makes it clear that the old armor formula is no longer accurate.

Now that we know what the level 93 armor value is, we can work backwards to find K:
P=1-A/(A+K), with P=0.6504 and A=24835
Re-arrainging,
K=A*P/(1-P) = 46203.33.

Returning to the CS/HotR data:

Code: Select all

cs_fit =

     General model:
     cs_fit(x) = 0.6504*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =       1.718  (1.671, 1.764)
       b =        1049  (948.3, 1151)

cs_gof =

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


hotr_fit =

     General model:
     hotr_fit(x) = 0.6504*a*x
     Coefficients (with 95% confidence bounds):
       a =      0.2005  (0.2002, 0.2007)

hotr_gof =

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


HotR fits very nicely with 20% weapon damage, so that's confirmed.

CS is still sort of weird. It seems like it's doing 170% normalized weapon damage rather than 160%. The intercept falls within the bounds - if I fix that in place, I get this:

Code: Select all

General model:
     f(x) = 0.6504*(a*x+1123)
Coefficients (with 95% confidence bounds):
       a =       1.689  (1.662, 1.716)

Goodness of fit:
  SSE: 1.487e+004
  R-square: 0.9989
  Adjusted R-square: 0.9989
  RMSE: 46.09

Again, pretty consistent with about 170% normalized weapon damage. Are we sure you're not wearing any gear that buffs CS? We're seeing a ~5-6% increase in CS damage over what we're expecting. Otherwise I'd have to conclude that the tooltip is off based on this data.

*ponders*
Lemme guess, you see that deviation starting from set #4 onward, right?

I think the pvp gloves might have a 10% bonus to CS, but I can't fully recall.

Edit: just confirmed it
http://mop.wowhead.com/item=84834

I'll redo set#4 and onward, since it's from there that I started to wear that particular piece of gear.

5th data set, actually. I get a beautiful fit to the first 4 data points with 0.6504*(1.6*x+1123), but the 5th and onward are too high. I also get a beautiful fit to the 5th and later data points with 0.6504*(1.6*x+1123)*1.05. So I guess that removes the discrepancy - CS matches the tooltip exactly, and the PVP hands are still a multiplicative 5%.

Though I posted this as a response on your recent blog already Theck, I figured I might as well spread the word here too.
In this post: http://us.battle.net/wow/en/forum/topic ... ge=90#1785, GC gives us the value a for agi to dodge, and likely strength to parry as well.
The value in question would be 951.158596 to 1%(see bottom of the post).
Hope this helps and keep up the good work.

Thanks. The really impressive part is that 951.2 was my best fit to the monk data in today's blog post. So it's nice to see a confirmation that I was homing in on the correct number. I think it's a reasonable assumption that 951.158596 is the exact conversion for str->parry as well.

Klaud, if you find time, a similar diminishing returns data set would be useful for block/mastery. Naked values as well as pre- and post-DR values (mastery for pre-DR, char sheet block % for post-DR) at a variety of gear levels. The higher you can push mastery, the better, of course.

<edit> also, if you don't mind, rather than typing everything into a post here, just put it in a google docs spreadsheet. Shouldn't take you any more time (probably less) and saves me time, since I can import it into matlab instead of retyping.

Understood. I'll start working on it as soon as I get home.

https://docs.google.com/document/d/1UpO ... --Z_g/edit

I'll be editing as I go -- right now I got the mastery stuff done, but the block% tooltip is borked - it says block of 0 adds 0.00% to block, regardless of how much mastery you have.

Also, currently doing the HoW parses, noticed something weird, its proccing divine purpose

You can probably ignore the block tooltip. We should be getting 1% block from 1% mastery, and unless I'm forgetting something there's no other source of block chance.

HoW must be a bug, one of us should probably report it.

<edit> also, lol @ using a google word document for the block data. You're killin' me here Klaud. I can export spreadsheets as excel documents and import those into matlab directly. Word documents, not so much.

it was not always proccing on cast, but it would proc every now and then

Klaudandus wrote:it was not always proccing on cast, but it would proc every now and then

Yeah, I'd assume it follows the standard 15% proc rate of Divine Purpose. The fact that it procs it at all is a bug though.

Here's the block data in spreadsheet form:
https://docs.google.com/spreadsheet/ccc ... yYmc#gid=0

If you PM me your gmail address, I'll give you edit access (unless you're done, in which case don't worry about it).

i can tell u got an error in that spreadsheet, its 6646 mastery (the one below the BoM)

also, why the hell did I go with a document and a spreadsheet? oh well, c'est la vie~