A Call to Arms - MoP Mechanics Testing

Yeah, it was from the character sheet.

Still would like to know why our WoG and EF went down, there was no mention of a change to them in the latest build -- although they might have changed how much pvp power rating you need per % from the previous build to this one.

Lulz. They nerfed PvP power by a factor of 2, exactly. Fitting using SP, PvP%, and the EF tick data:

Code: Select all

General model:
     f(x,y) = (508+0.0585*x)*(1+0.01*y/c)
Coefficients (with 95% confidence bounds):
       c =           2  (1.994, 2.007)

Goodness of fit:
  SSE: 3.58
  R-square: 1
  Adjusted R-square: 1
  RMSE: 0.4589


So the PvP power % displayed on the character sheet is off by a factor of 2, but otherwise the old formula for ticks holds. Now to look at EF/WoG data.

Fit results:

CS and J match tooltip exactly.

Code: Select all

-----CS-----

cs1_fit =

     General model:
     cs1_fit(x) = 0.6507*(a*x+b)
     Coefficients (with 95% confidence bounds):
       a =        1.25  (1.25, 1.25)
       b =       791.3  (791.2, 791.4)

cs1_gof =

           sse: 4.8595e-004
       rsquare: 1.0000
           dfe: 2
    adjrsquare: 1.0000
          rmse: 0.0156

-----J-----

j_fit =

     General model:
     j_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =       1.202  (1.202, 1.202)
       b =       623.1  (623, 623.2)

j_gof =

           sse: 0.0016
       rsquare: 1.0000
           dfe: 2
    adjrsquare: 1.0000
          rmse: 0.0284


WoG is still using the formula shown in EF's tooltip, modified by PvP power (which was nerfed, now gets 1% from ~530 rating).
EF matches tooltip exactly (again, modified by PvP power).

Code: Select all

-----WoG-----

     General model:
     wog1min_sfit(x,y) = (b+0.49*x)*(1+0.01*y/2)
     Coefficients (with 95% confidence bounds):
       b =        5263  (5235, 5291)


wog1min_sgof =

           sse: 1.0229e+003
       rsquare: 0.9999
           dfe: 3
    adjrsquare: 0.9999
          rmse: 18.4652


     General model:
     wog1_sfit(x,y) = (b+0.49*x)*(1+0.01*y/2)
     Coefficients (with 95% confidence bounds):
       b =        5832  (5828, 5836)


wog1_sgof =

           sse: 18.1608
       rsquare: 1.0000
           dfe: 3
    adjrsquare: 1.0000
          rmse: 2.4604

-----EF-----

     General model:
     ef1_sfit(x,y) = (5239+b+0.49*x)*(1+0.01*y/2)
     Coefficients (with 95% confidence bounds):
       b =       10.48  (-3.303, 24.26)


ef1_sgof =

           sse: 244.7959
       rsquare: 1.0000
           dfe: 3
    adjrsquare: 1.0000
          rmse: 9.0332


EF ticks also match tooltip, as shown in the previous post.

Not sure if you still need it but WoG and EF parts of #38

http://www.worldoflogs.com/reports/rt-0dxcgvq9qh3yi7jo/

Level 86 naked with Hatchet
AP 582
SP 291

Klaudandus wrote:#21
AP:602
SP:301

http://www.worldoflogs.com/reports/jss558obqq3kydkj/

No windsong procs from Holy Prism, as far as I can tell from the combat log. The numbers seem off though; Windsong only procced 12 times in 307 melee attacks.

If it's a 3 PPM enchant, then we'd expect a (3*1.9/60)=9.5% proc rate, and approximately 29 procs.
A 1 PPM enchant would give us an expectation of ~9.7 procs (3.17% proc rate), which is much closer to what we observed.

Unfortunately, at that proc rate, we can't tell if HPr is a proc trigger or not. Including it gives us an expected 10.5 procs, not enough difference from the ~10 we expect just from melees.

I'm wondering if it would make more sense to switch this test around some. Equipping a 3.6-speed (or higher) weapon and standing at range, casting only Holy Prism on the dummy. It would then take ~100 casts (34 minutes) with no proc to reasonably conclude that HPr isn't a trigger. Ideally, use no other attacks so that there's no chance of something else proccing it. Of course, if you observe a proc, you can stop earlier than 100 casts.

daishan wrote:Not sure if you still need it but WoG and EF parts of #38

http://www.worldoflogs.com/reports/rt-0dxcgvq9qh3yi7jo/

Level 86 naked with Hatchet
AP 582
SP 291

WoG definitely benefits from the 5% boost.
EF base heal not affected at all (~4990 min with and w/o SoI)
EF ticks not affected at all (485 with and w/o SoI)

ok, i'll try

I've seen a couple of ppl saying Blizz nerfed Windsong proc rate.

I'm going to leave beta logging auto attacks to http://www.worldoflogs.com/reports/rt-rqrxja1q77c4et4n/ for a hour or 2.

lvl 86
Naked using Obsidium Cleaver normal, speed 2.6

http://www.worldoflogs.com/reports/rt-49x9tthdgqvqyeac/

Another bunch of Windsong logs just for the hell of it.
Ranseur of Hatred normal speed 3.6 (forgot about the 252 haste on it until I'd already uploaded)

Unless my maths is miles off that gives a 5.14% proc rate from the first log and 5.37% proc rate from the 2nd log.

Just to note you can still have more than one type of proc active at once, but if you have lets say a crit proc active then get another crit proc it'll overwrite the first so you'll never gain 2 of the same proc.

The 16004 build is still not out >=/

16004 data is out for Cons and HotR on the spreadsheet

40 mins of #21 with a 3.6 weapon equipped
http://www.worldoflogs.com/reports/8zmgouclwxa9mcma/

daishan wrote:I've seen a couple of ppl saying Blizz nerfed Windsong proc rate.

I'm going to leave beta logging auto attacks to http://www.worldoflogs.com/reports/rt-rqrxja1q77c4et4n/ for a hour or 2.

lvl 86
Naked using Obsidium Cleaver normal, speed 2.6

daishan wrote:http://www.worldoflogs.com/reports/rt-49x9tthdgqvqyeac/

Another bunch of Windsong logs just for the hell of it.
Ranseur of Hatred normal speed 3.6 (forgot about the 252 haste on it until I'd already uploaded)

Unless my maths is miles off that gives a 5.14% proc rate from the first log and 5.37% proc rate from the 2nd log.

Your math is miles off.

First log:
36+43+29=108 procs from 2478 melee attacks. Proc rate of 108/2478=4.36%. 95% confidence interval is +/- 0.8%.
Expected proc rate for a 1 PPM enchant with a 2.6-speed weapon is 1*2.6/60=4.33%. Certainly seems to support a 1 PPM mechanic.

Second log:
31+30+28=89 procs from 1675 melee attacks, proc rate of 5.31%. 95% CI is +/-1.07%.
Expected proc rate from a 3.6-speed weapon is 3.6/60=6.0%. Smaller sample size makes this one less reliable, since both logs could support a flat 5% proc rate. However, our earlier tests conclusively showed that this was a PPM mechanic. We can reasonably assume that it's still a PPM mechanic, and as far as I know we've never seen a PPM enchant with half-integer value. The logical conclusion is that this is a 1 PPM enchant.

Ideally, we'd want a longer log with the 3.6-speed weapon, roughly ~5000 hits in total (though the ~1.7k we already have can count towards that total). That should be enough to rule out a flat 5% proc rate. Note that wearing haste gear is perfectly acceptable for this, and in fact is probably ideal (to collect data faster).

Klaudandus wrote:40 mins of #21 with a 3.6 weapon equipped
http://www.worldoflogs.com/reports/8zmgouclwxa9mcma/

Chance of 0 procs out of 133 triggers at 6% proc rate (1 PPM with 3.6-speed weapon): 0.0267%. It's about 0.11% if it's a 5% flat proc rate. I think we can rule out Holy Prism as a proc trigger. #21 complete.

Klaudandus wrote:16004 data is out for Cons and HotR on the spreadsheet

Fits:

Code: Select all

-----HotR(phys)----

hotr1_fit =

     General model:
     hotr1_fit(x) = 0.6507*a*x
     Coefficients (with 95% confidence bounds):
       a =         0.2  (0.1998, 0.2002)

hotr1_gof =

           sse: 0.3478
       rsquare: 1.0000
           dfe: 6
    adjrsquare: 1.0000
          rmse: 0.2407

----HotR(nova)----

nova1_fit =

     General model:
     nova1_fit(x) = a*x
     Coefficients (with 95% confidence bounds):
       a =      0.3498  (0.3496, 0.3501)

nova1_gof =

           sse: 1.2501
       rsquare: 1.0000
           dfe: 6
    adjrsquare: 1.0000
          rmse: 0.4564

----Cons----

cons1_fit =

     General model:
     cons1_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.1799  (0.1798, 0.18)
       b =       103.2  (102.7, 103.7)

cons1_gof =

           sse: 0.3965
       rsquare: 1.0000
           dfe: 5
    adjrsquare: 1.0000
          rmse: 0.2816

HotR is as the tooltip claims.

Is my memory going batty, or does Consecration grant 10 ticks? If so, it just got significantly buffed (old: 690+1.2*SP, new: 1027+1.8*SP)

<edit> like, holy shit buffed. It's now our second-hardest hitting attack behind AS at 50k Vengeance, and it scales better with AP. That's going to shake some things up.

We got superglue?

Can has?

Next time you're on beta, could you just confirm for me that one Cons cast gives 10 ticks? I'm 99% sure it does, but now I'm second-guessing myself because of how large that buff is.

I only see 9 ticks
http://www.worldoflogs.com/reports/exk1p6lra597ue8t/

Im following your Blog Theck, and using your DR values and formulae to write my macros and addons, but nothing seems to line up.

Code: Select all

  baseAvoid + (1/Ca + k/A)^-1


Ca is just replaced with the appropriate stat, as is A (preDR of course)

the numbers I got off your blog:

Code: Select all

k = .855
Cb = 1.491
Cd = 65.63144
Cp = 235.5


When I run these formulae on beta my dodge is always ~.1% lower than expected, but is still higher than its PreDR value. My sample sets are likely not as good as yours, but range from 7.65% Character sheet Dodge to 9.06%. I get similar, but more profound discrepancies with Block. At 24.99% Mastery, the formula says I should have 1.416% PostDR Block, but the tooltip shows 36.65% (including Base Block).

My gut says that tooltips are wildly inaccurate, but the Block DR formula still feels too harsh (mostly considering the fact that Mastery is still quite valuable on your TDR list)

I didnt get access to the MT WoL guild, so I used my own if thats ok.

60 minute Parse, Colossus enchanted, 2.6s weapon:
http://www.worldoflogs.com/reports/rt-jtvd3j1cds3274tq/details/4/

If you need more details, I would be happy to provide them.

theckhd wrote:Your math is miles off.

Bah :p

theckhd wrote:First log:
36+43+29=108 procs from 2478 melee attacks. Proc rate of 108/2478=4.36%.

Where do you count the procs?
I look at the buffs gained http://www.worldoflogs.com/reports/rt-r ... details/0/ and see 46 haste, 43 crit, and 29 Mast = 118/2478 = 4.76%
No idea what I did to get 5.14% first time round.

Ranseur of Hatred normal speed 3.6

Todays log:
http://www.worldoflogs.com/reports/rt-x ... details/0/
106+99+79=284 procs from 4427 melee attacks, proc rate of 284/4427=6.42%.

Yesterdays 3.6 log:
http://www.worldoflogs.com/reports/rt-49x9tthdgqvqyeac/
31+30+28=89 procs from 1675 melee attacks, proc rate of 5.31%.

Total:
89+284=373 procs from 6102 melee attacks, proc rate of 373/6102=6.11%.


I'm most likely been dense but could you explain how it can be a PPM mechanic but not have a ICD, is it somehow normalized based on weapon speed? If so how are special attacks that can proc it worked in?

xstratax wrote:Im following your Blog Theck, and using your DR values and formulae to write my macros and addons, but nothing seems to line up.

Code: Select all

  baseAvoid + (1/Ca + k/A)^-1


Ca is just replaced with the appropriate stat, as is A (preDR of course)

the numbers I got off your blog:

Code: Select all

k = .855
Cb = 1.491
Cd = 65.63144
Cp = 235.5


When I run these formulae on beta my dodge is always ~.1% lower than expected, but is still higher than its PreDR value. My sample sets are likely not as good as yours, but range from 7.65% Character sheet Dodge to 9.06%. I get similar, but more profound discrepancies with Block. At 24.99% Mastery, the formula says I should have 1.416% PostDR Block, but the tooltip shows 36.65% (including Base Block).

My gut says that tooltips are wildly inaccurate, but the Block DR formula still feels too harsh (mostly considering the fact that Mastery is still quite valuable on your TDR list)

I'm guessing it's a unit consistency issue. For example, if you express Cd in percent instead of decimal, then you have to do the same with avoidance. As an example, 10% pre-DR dodge should give you:
1/(1/65.63144+0.885/10)=9.6398% post-dr dodge. Add that to the base 3.01 to get 12.65% on the character sheet.

Similarly for block. 24.99% mastery should give:
1/(1/149.1+0.885/24.99)=23.7411% post-DR. We have 13% base block (10% from GbtL, 3% base), for a total of 23.74+13=36.74%.

If you incorrectly use 1.491 and 24.99 in the formulas simultaneously, you get the 1.416% result.

I usually do everything in decimals, i.e.:

Code: Select all

Cd=0.65631440;  %Diminishing Returns C coefficient - dodge
Cp=2.32;        %Diminishing Returns C coefficient - parry
Cb=1.491;       %Diminishing Returns C coefficient - block

But either way works as long as you stay consistent. Not sure why the character sheet is saying 36.65% instead of 36.74%; it's possible they've slightly tweaked the block DR coefficient since we last tested it. That sort of thing doesn't usually make it into patch notes. It's easy enough for me to re-fit it if someone gives me data.

daishan wrote:

theckhd wrote:First log:
36+43+29=108 procs from 2478 melee attacks. Proc rate of 108/2478=4.36%.

Where do you count the procs?
I look at the buffs gained http://www.worldoflogs.com/reports/rt-r ... details/0/ and see 46 haste, 43 crit, and 29 Mast = 118/2478 = 4.76%
No idea what I did to get 5.14% first time round.

Same place you did. But apparently I typo-ed 46 into 36. Fixing that gives me the same 4.76% you got. Still no idea how you got 5.14%, but it was likely a similar error.

daishan wrote:Ranseur of Hatred normal speed 3.6

Todays log:
http://www.worldoflogs.com/reports/rt-x ... details/0/
106+99+79=284 procs from 4427 melee attacks, proc rate of 284/4427=6.42%.

Yesterdays 3.6 log:
http://www.worldoflogs.com/reports/rt-49x9tthdgqvqyeac/
31+30+28=89 procs from 1675 melee attacks, proc rate of 5.31%.

Total:
89+284=373 procs from 6102 melee attacks, proc rate of 373/6102=6.11%.


I'm most likely been dense but could you explain how it can be a PPM mechanic but not have a ICD, is it somehow normalized based on weapon speed? If so how are special attacks that can proc it worked in?

PPM mechanics don't have an ICD. They use your weapon speed to determine a proc rate in order to normalize for weapon speed.

For example, if I have a 3.6-speed weapon, the proc rate for a 1 PPM enchant is
PPM*weapon_speed/seconds_per_minute = 1*3.6/60 = 0.06 = 6%.

A 3 PPM enchant would have a proc rate three times higher, or 3*3.6/60=18%. If we changed to a 2.6-speed weapon, the 3PPM proc rate would drop accordingly: 3*2.6/60 = 13%.

Special attacks that can proc it are a free bonus - they use exactly the same proc rate as white attacks. So while we generally use only melee attacks to determine the proc rate, in practice you'll get significantly more procs per minute than the given PPM value. In the case of Ret, you might get 3x as many (because their swing time is much longer than the GCD). Prot usually only gets 1.5-2x as many because more of our attacks are spells and our swing timer isn't as long compared to the GCD.

In any case, the combined data sets you provided solidify Windsong as 1 PPM. Proc rate of 0.0611 with a 3.6-speed weapon. The 95% confidence interval is 1.96*sqrt(0.0611*(1-0.0611)/6102) = 0.006 = 0.6%. So we feel pretty confident the proc rate is between 5.51% and 6.71%, which rules out the flat 5% proc hypothesis.

How many Mastery points do you need to get an accurate model of the coefficient? And how much of a difference should there be, I think I could possibly get a bunch of plot points between my Ret, PvP, and Tank items.

xstratax wrote:I didnt get access to the MT WoL guild, so I used my own if thats ok.

60 minute Parse, Colossus enchanted, 2.6s weapon:
http://www.worldoflogs.com/reports/rt-jtvd3j1cds3274tq/details/4/

If you need more details, I would be happy to provide them.

153 procs in (1180+97+84)=1361 attacks. Proc rate of 11.24%, 95% confidence interval of +/-1.68%.

This one's a little troubling, because 2 PPM would be 8.67% and 3 PPM would be 13%. It's possible that we're under-counting the procs though, because in my experience WoL tends to ignore refreshes in the buff count tally. Still, it seems unlikely that we're missing almost 24 procs.

There's also the possibility that it's using normalized weapon speed, which would give a proc rate of 3*2.4/60=12%. That also seems unlikely, given that I don't think any weapon enchant has done that since.. Wrath? BC? I can't even remember.

We'll probably need more data to nail this one down. Unfortunately, to determine it to the required accuracy (+/- 0.5% would be ideal), it's going to take 15k melee swings...