A Call to Arms - MoP Mechanics Testing

[theckhd]

I may have been mixing up blue posts. This is the one I was thinking of that mentioned dodge/parry, but it looks like it was only relevant to River's Song and Colossus:
http://us.battle.net/wow/en/forum/topic ... page=8#154

[list][*]River's Song: 2PPM on melee attacks that land, or are dodged, or parried.[*]Dancing Steel: 1PPM on melee attacks that land.[*]Colossus: 3PPM on melee attacks that land, or are dodged, or parried, with a 3-second cooldown.[*]Elemental Force: 5PPM on melee damage, or non-periodic spell damage/healing, with a 0.1-second cooldown.[*]Jade Spirit: 10% chance on spell damage or healing, 50-second cooldown.[*]Windsong: 1PPM on melee damage, or non-periodic spell damage/healing, with a 1-second cooldown.

The RPPM post doesn't mention dodge/parry at all. So we'll have to wait until 5.1 to test whether that remains the case with the RPPM implementation of RS and Colossus.

I got logged off while doing the melee attack speed test. Is the data collected sufficient? If not, I'll have to try and log another hour or two Friday or Saturday.

[benebarba]

Current Parse Requests:

Test #34: Light's Hammer target cap

• Level: 90
• Gear: Any
• Target: A group of exactly N mobs. Mob level is irrelevant.
• Attack Sequence: Cast a single Light's Hammer once all mobs are in range and grouped up.
• Goal: Repeat this for N=5, 7, 9, 10, 11, 15.
• Report: build, AP/SP, upload combat log. Preferably a different log for each N.

Does it matter if they are friendly (healing) vs hostile (damage)? Or should this be tried with both?

[theckhd]

I was mostly interested in hostile targets, but testing both would probably be a good idea.

[jere]

In the hotfixes for Dec 3, there was a change listed for Holy Wrath (I think...I can't verify from here). Might be worth checking out.

[KysenMurrin]

Wording was "now scales correctly with attack power and spell power". Definately needs retesting.

[Nooska]

As opposed to how it worked with 5.1, which broke scaling from AP and SP.

[Fetzie]

My Holy Wrath (no glyphs that interact with the ability) is doing 25740-25741 (does one or the other) damage to a level 90 training dummy right now (18:29 CET, 4th Dec). It deals the same amount of damage to a boss level training dummy.

Blood Elf paladin (176 base strength)
11655 strength
23560 melee attack power
11780 spell power

If you want a combatlog, then say so

[theckhd]

In the hotfixes for Dec 3, there was a change listed for Holy Wrath (I think...I can't verify from here). Might be worth checking out.

It was just a bugfix. 5.1 accidentally removed the AP/SP scaling. They just re-instated it, the formula hasn't changed since 5.0.5. Fetzie's data is consistent with the current model (4300+0.91*AP).

[theckhd]

New test for the SOI proc rate buff:

Test #51: SoI proc rate buff

• Gear: any
• Seal: SoI
• Buffs/Talents/Glyphs/Etc: Any
• Target: Any dummy
• Attack sequence: auto-attacks only, faster weapons preferable
• Goal: Over 100 successful auto-attacks against the dummy.

[theckhd]

Answered my own question:
http://www.worldoflogs.com/reports/rt-7 ... etails/16/

200 melee attacks, 179 SoI procs, 89.5% proc rate. For such a small sample size, that's reasonably good confirmation that we're now at 20 PPM.

[Klaudandus]

Answered my own question:
http://www.worldoflogs.com/reports/rt-7 ... etails/16/

200 melee attacks, 179 SoI procs, 89.5% proc rate. For such a small sample size, that's reasonably good confirmation that we're now at 20 PPM.

I assume that's good

[theckhd]

Well, it's unquestionably a buff.

[theckhd]

While we've exhaustively tested offensive armor formulas, it seems we haven't done our due diligence on the defensive side. So far we've been using K=58370 for the armor constant of a level 90 player. Technically we use a formula which gives this number, but I don't remember testing the formula during beta (nor documenting it in this thread). My best guess is that it's from another source, either early beta datamining or it's the equation we've been using since Cataclysm.

In any event, it seems pretty clear that it's wrong. I was testing some things today and noticed that with 57030 armor, I had 55.21% damage reduction on my character sheet rather than the 49.42% predicted by A/(A+K). I took a quick data set of armor and character sheet mitigation:

Code: Select all

    0               0
5160           10.04
6494           12.31
12845           21.73
18261            28.3
25605           35.63
31277           40.34
37663           44.88
42305           47.77
45855           49.78
49724           51.81
52485           53.15
57030           55.21

Throwing that into matlab and fitting to 100*x/(x+K) gives me the following result:

Code: Select all

General model:
     f(x) = 100*x/(x+K)
Coefficients (with 95% confidence bounds):
       K =  4.626e+004  (4.626e+004, 4.626e+004)

Goodness of fit:
  SSE: 9.676e-005
  R-square: 1
  Adjusted R-square: 1
  RMSE: 0.00284


A further fit to 100*x/(x+46260+K) narrows down the last digit some:

Code: Select all

General model:
     f(x) = 100*x/(x+46260+K)
Coefficients (with 95% confidence bounds):
       K =      -1.048  (-4.834, 2.737)

Goodness of fit:
  SSE: 9.676e-005
  R-square: 1
  Adjusted R-square: 1
  RMSE: 0.00284


From this, it seems that the armor constant at L90 is actually K=46259 +/-3. I would appreciate some additional data sets from anyone who's got 5-10 minutes to spare. All I need is the character sheet armor values and mitigation percentages (from the armor tooltip). Level 90 paladins only, class/race/spec shouldn't matter. As many data points as you can submit would be great, ideally everything from naked to fully geared.

Actually, skip naked and go right to one piece of armor equipped. I don't need 100 data points to tell me that 0 armor gives 0% mitigation.

[Klaudandus]

i blame that on GC, he's the one that gave us the armor value.

[theckhd]

i blame that on GC, he's the one that gave us the armor value.

Did he? That would explain why we didn't test it (shame on us, though!).

If you could dig up a link to that post, I'd appreciate it. Once we have enough data to be convinced of this, I might write a short blog post to spread the word.