A Call to Arms - MoP Mechanics Testing

Warning: Theorycraft inside.

Moderators: Fridmarr, Worldie, Aergis, theckhd

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Tue Jul 24, 2012 7:53 pm

WoG data is a little weird:
tooltip: http://www.wowdb.com/spells/130551-word-of-glory
(3694-4115) + 0.377*sp
fit (min and max):
Code: Select all
wog_fit =

     General model:
     wog_fit(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.5291  (0.4915, 0.5668)
       b =        3922  (3710, 4134)

wog_gof =

           sse: 9.8495e+003
       rsquare: 0.9995
           dfe: 2
    adjrsquare: 0.9992
          rmse: 70.1766


wog_fit2 =

     General model:
     wog_fit2(x) = a*x+b
     Coefficients (with 95% confidence bounds):
       a =      0.5362  (0.5001, 0.5722)
       b =        4370  (4167, 4574)

wog_gof2 =

           sse: 9.0587e+003
       rsquare: 0.9995
           dfe: 2
    adjrsquare: 0.9993
          rmse: 67.3003

There's enough error here that the base values could be right, but the scaling seems much higher than the tooltip claims. Maybe there's something obvious I'm missing, since it's late. I'll look at it agian tomorrow. <edit> and I'll add the 4-6 data.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Tue Jul 24, 2012 7:55 pm

Klaudandus wrote: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?

60 casts of ES and HPr, 8 LH. Typo on my part. Either way, ES/Hpr are going to take about an hour each because of the cooldown.

I think the fastest weapon now is a Tomahawk, so use that instead for the Colossus/Windsong tests.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Tue Jul 24, 2012 7:56 pm

what is HPr? Holy Prism? Also, shouldnt HPr actually be much faster considering that the cd is 20 secs?

Also, in Test #4, you say any gear... but you dont mention anything about hit/exp

just need to be sure before I leave the computer on for an hour
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Tue Jul 24, 2012 8:22 pm

Well, neither colossus nor windsong work on the tomahawk, only works on weapons that are 372 or better. The slowest weapon that qualifies is 2.7 (the plasma sword) but the easiest to obtain are 2.6
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Tue Jul 24, 2012 8:36 pm

I can use this mace http://mop.wowhead.com/item=80795/jade-forest-scepter if you deem it acceptable for the tests
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Wed Jul 25, 2012 3:54 am

Klaudandus wrote:what is HPr? Holy Prism? Also, shouldnt HPr actually be much faster considering that the cd is 20 secs?

Oops, yeah. HPr will only take 20 minutes, but the two 1-minute cooldown spells (ES and LH) will take an hour to get 60 casts.

Klaudandus wrote:Also, in Test #4, you say any gear... but you dont mention anything about hit/exp

just need to be sure before I leave the computer on for an hour

It specifies a low-level dummy, so you should automatically hit/exp cap.
Klaudandus wrote:I can use this mace http://mop.wowhead.com/item=80795/jade-forest-scepter if you deem it acceptable for the tests

Yeah, that should be fine. We just want something that's far away from 2.4, so 1.6 should work well enough.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Wed Jul 25, 2012 4:50 am

ok, i'll do #4 and #33 with windsong and colossus (only enchants i've been able to get) over the next few days -- got a raid tonight, so i can prolly do #4-colossus at least, maybe #4-windsong if i'm lucky
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Wed Jul 25, 2012 4:52 am

Something strange is definitely going on with the WoG data. For example:
Image
The blue line is the following fit to the "max" set:
Code: Select all
General model:
     f(x) = (a*x+b)
Coefficients (with 95% confidence bounds):
       a =      0.5371  (0.522, 0.5523)
       b =        4367  (4270, 4463)

Goodness of fit:
  SSE: 1.009e+004
  R-square: 0.9996
  Adjusted R-square: 0.9995
  RMSE: 50.23

That curve doesn't do a bad job, but note that the intercept is outside the tooltip range (but then again, so is the scaling).

If I try the same thing with the min data, I get the red curve:
Code: Select all
General model:
     f(x) = (a*x+b)
Coefficients (with 95% confidence bounds):
       a =      0.5291  (0.5134, 0.5448)
       b =        3918  (3818, 4018)

Goodness of fit:
  SSE: 1.08e+004
  R-square: 0.9995
  Adjusted R-square: 0.9994
  RMSE: 51.96


So far so good, maybe. However, if you look carefully at the lower 3 data points, something looks fishy. The errors appear systematic - in other words, both the min and the max of the naked set are too high, and both min and max of the ~3k SP data set are too low. A similar (but less significant) effect is happening for the last 3 gear sets. To illustrate that, this is what happens if you fit the lower and upper 3 data points of the min data set independently:
Image
Low:
Code: Select all
General model:
     f(x) = (a*x+b)
Coefficients (with 95% confidence bounds):
       a =      0.4825  (0.2859, 0.6791)
       b =        3995  (3592, 4398)

Goodness of fit:
  SSE: 923.6
  R-square: 0.999
  Adjusted R-square: 0.9979
  RMSE: 30.39

High:
Code: Select all
General model:
     f(x) = (a*x+b)
Coefficients (with 95% confidence bounds):
       a =      0.5405  (0.5139, 0.567)
       b =        3824  (3592, 4056)

Goodness of fit:
  SSE: 47.43
  R-square: 1
  Adjusted R-square: 1
  RMSE: 6.887


Notice that the fits get much better. Admittedly, some of this is because we have fewer data points to fit (and thus more leeway with the parameters), but it's eerie that the three data points ended up so linear. It's almost as if something changed between gear sets naked-2 and gear sets 4+.

I'm going to look at the prism data that's included in these data sets and see if that gives me any clue as to what's happening.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Wed Jul 25, 2012 4:55 am

i can repeat the WoG logs again if needed be. it just drives me crazy to try to always make sure that i only have one HP before casting WoG -- but at least it seems my /cancelaura Divine Purpose is working -- and do all 8 (naked + 7) sets
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Wed Jul 25, 2012 4:55 am

Klaudandus wrote:ok, i'll do #4 and #33 with windsong and colossus (only enchants i've been able to get) over the next few days -- got a raid tonight, so i can prolly do #4-colossus at least, maybe #4-windsong if i'm lucky

That's fine. As far as the L90 talents, I should amend the description slightly to note that you can stop after 10-15 casts as long as you have a few misses. I'm 99.99% sure they're spells rather than melee abilties (since that would be dumb), but what I'm after is whether they can miss, and in ES's case, does each tick roll independently.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Wed Jul 25, 2012 4:57 am

Klaudandus wrote:i can repeat the WoG logs again if needed be. it just drives me crazy to try to always make sure that i only have one HP before casting WoG -- but at least it seems my /cancelaura Divine Purpose is working -- and do all 8 (naked + 7) sets

That's fine, but let me look at the Holy Prism data first. I should have that done this morning yet, afternoon at the latest.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Wed Jul 25, 2012 5:31 am

you need #4 for all enchants listed, right? cuz like I said, only gotten 2 of those enchants listed on the AH =/

Also, question about #4, does it need exactly 60 mins or can it be more than 60 mins?
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Wed Jul 25, 2012 5:56 am

Yes, eventually.

By the way, this new Wog/HPr data is all from the current build, right? I.e. after the HPr nerfs in build 15851?

<edit> the systematic error is showing up in the HPr healing data too, and the healing done by WoG and HPr are both around 40% higher than tooltip values. Neither the error nor the 40% increase shows up in the HPr damage data. Something is definitely amiss here. I think we're going to have to re-take this data, and double check the AP/SP values of the gear sets when you do it. I'll need a full documentation of everything: AP, SP, talents, glyphs, buffs, etc. It seems unlikely that WoG and HPr healing were stealth-buffed by 40%.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

Re: A Call to Arms - MoP Mechanics Testing

Postby Klaudandus » Wed Jul 25, 2012 6:01 am

The last HPr/WoG sets were all from last night -- doh, forgot to write down the build in the txt file.

Also, I still need to know about those 60 mins...
The Element of Forum Hyperbole
Image
---
Flüttershy - Draenei Protection Paladin, Aerie Peak
Klaudandus - BE Protection Paladin, Feathermoon (Semi-retired)
User avatar
Klaudandus
 
Posts: 11168
Joined: Thu Apr 02, 2009 7:08 am
Location: Texas' Armpit

Re: A Call to Arms - MoP Mechanics Testing

Postby theckhd » Wed Jul 25, 2012 7:41 am

Time for #4 needs to exceed 60 minutes. Longer is fine, shorter is not.

Update on the WoG/HPr data. Something is really weird, it seems like the base values for WoG are 6% higher than tooltip, while the HPr base values are ~3% lower. The sp scaling coefficients are all ~40% higher though. As such, this wouldn't be consistent with an external +heal% buff, and a +sp or +healing power buff wouldn't explain the base values. At this point, I'm going to wait for an updated data set - if the results are consistent, there's a good chance that WoG and HPr got stealth changes that haven't been reflected in mmo-champion or wowhead somehow.
"Theck, Bringer of Numbers and Pounding Headaches," courtesy of Grehn|Skipjack.
MATLAB 5.x, Simcraft 6.x, Call to Arms 6.0, Talent Spec & Glyph Guide 5.x, Blog: Sacred Duty
User avatar
theckhd
Moderator
 
Posts: 7985
Joined: Thu Jul 31, 2008 3:06 pm
Location: Harrisburg, PA

PreviousNext

Return to Advanced Theorycraft and Calculations

Who is online

Users browsing this forum: Bing [Bot] and 1 guest

Who is online

In total there are 2 users online :: 1 registered, 0 hidden and 1 guest (based on users active over the past 5 minutes)
Most users ever online was 380 on Tue Oct 14, 2008 6:28 pm

Users browsing this forum: Bing [Bot] and 1 guest