A Call to Arms - Cataclysm Mechanics testing

[theckhd]

Word of Glory Testing

• spec: anything prot, so long as you note how many points you have in WoG-increasing talents (Guarded by the Light)
• glyphs: anything except Long Word or Word of Glory
• seal: Irrelevant, but Insight is probably easiest
• target: any dummy
• attack sequence: CS-WoG-CS-WoG, repeat. Cast only 1-HoPo WoGs, which means chain casting WoG if Eternal Glory procs.
• gear: 3 different sets
  
  • Naked
  • Intellect or SP gear only (same AP as Naked, different SP, in other words)
  • Tanking or Ret gear (high AP, SP)
• goal: 100 or so WoGs in each gear set up, which is about 5 minutes each.

[tlitp]

Newer research on Avalanche. The Exo test does agree with both ICD/proc chance, but the AA+SoT test doesn't. I'd like to see a retest of the latter one. Same setup : slow 2H, only AA+SoT, attack a low-lvl dummy for an hour or so.

Newer research on Vengeance. A confirmation would be handy. Keep in mind the rough value of your buffed health pool (e.g. 175k), have a look at your total AP when your Vengeance-tracking addon says it reaches a high threshold (>80%).

[theckhd]

Both very interesting. I have Avalanche on my Zin'rokh, so I'll try and get you another ~1-2 hour parse this afternoon.

The vengeance information is interesting, it means that blizzard coded it directly based on stamina instead of max health. That may have been done because threat tuning seemed right before the health modifier was bumped to 1.4, and they didn't want to have to retune anything.

[Durability]

The Vengeance oddity is really interesting - and another nail in ISH9's coffin, too.

As far as Avalanche proc weirdness w/Censure - maybe despite Censure's proc counting as a melee swing, the dot itself is considered a spell? It would explain the slightly lower than 10% proc chance via ICDs.

[theckhd]

http://www.worldoflogs.com/reports/rt-whjso6vm6jbj8890/

A little over 3 hours of dummy time with Avalanche and Zin'rokh. You might have to chop off a few minutes at the end if SoT fell off when I went AFK the last time. This was in prot gear/spec with Zin'rokh equipped.

Level 60 dummy, attacking from behind.

Relevant stats from char sheet:
Haste: 77 (0.6%)
Hit: 351 (2.92% melee @ 85, 11.43% spell)
Exp: 17 with SoT active
Weapon Speed: 3.8 base, 3.78 after haste
Damage: 3604-4552
AP: 6299
SP: 1915

[tlitp]

Theck :

• 3159 physical triggers; expected physical-based procs : mean 1000, stdev 26; assuming a normal distribution, 922-1078 procs
• 4009 spell triggers, with a 1/2.98 average rate
• "leftover" spell-based procs : 370-526

Let x be the "true" proc chance of spell triggers :

Code: Select all

true.icd=10
effective.icd=ceil(true.icd*trigger.rate)/trigger.rate=11.92
effective.pc=1/trigger.rate/(effective.icd+(1/trigger.rate/x))

370/4009=effective.pc => x.min=0.146
526/4009=effective.pc => x.max=0.276

Most probably, the "true" proc chance of spell triggers is 20%. As for the "true" ICD, the log doesn't contain (unequivocal) spell-based procs with less than ~12 sec spacing between them.

Newer research on Vengeance.

Formally confirmed.

[tlitp]

With the upcoming support for healing-based frameworks in mind, it's probably the time to ask for a Mending log :

• gear : any, with two restrictions (a. 0 (total) haste, b. a slow 2H equipped (3.6+ swing time))
• seal : Truth
• buffs : none
• target : any low-lvl dummy
• attack sequence : only autoattacks
• goal : at very least one hour (more is better) of AAs

Avoid haste procs, report base swing time, upload the CL. Thanks.

[theckhd]

I'm taking the Mending data right now. I can probably manage about an hour and a half of that and some WoG testing before I have to move on to other in-game obligations.


http://www.worldoflogs.com/reports/rt-v ... details/8/
Log in progress, should be roughly an hour and 30 minutes by the time I move over to testing WoG.

Relevant info:
STR/AGI/STA/INT/SPI: 3077/97/5082/106/117 (mix of ret and prot gear)
Melee:
Weapon is Zin'Rokh with Mending enchanted
3629-4577 character sheet damage
1079.6 DPS
6389 AP
3.80 speed
0.00% Haste (0 Rating)
5.24% Hit
8.46% crit
9 Expertise (210 rating)

Spell:
1942 SP
0% haste
14.14% hit
10.83% crit

[theckhd]

SoI/WoG/Mending data

100 WoGs cast in each config

Setup: 3/3 Divinity, 2/2 GbtL, SoI and WoG glyphs
Weapon: Zin'rokh w/ Mending
AP: 6597
SP: 2004
Crit: 2.68% melee, 4.83% spell
Haste: 0.00%
Hit: 2.92% melee, 11.43% spell
Expertise: 7
http://www.worldoflogs.com/reports/rt-t ... details/0/

Same setup, minus SoI glyph
http://www.worldoflogs.com/reports/rt-7 ... details/5/

minus WoG glyph
http://www.worldoflogs.com/reports/rt-y44kkbl6h3hkwone/

0/3 Divinity, 0/2 GbtL, neither glyph (also 0/3 Rule of Law, so fewer crits). 2/2 EG to get prot baselines
http://www.worldoflogs.com/reports/rt-ygrey77buey9mitj/

3/3 Divinity, 0/2 GbtL, neither glyph (still 0/3 RoL, 2/2 EG, no other point spent)
http://www.worldoflogs.com/reports/rt-q5cqjt8x7dvi9s9y/

At this point, I noticed that something funky is happening with the uploads. Each of these should have either 100 or 101 WoGs, but for whatever reason, it looks like WoW isn't dumping the entire log when I disable logging and force an update to the WoL server. This probably means that data sets are spilling over into the next set. I'm going to start retaking these now as one continuous parse, with a separator between each set.

[theckhd]

SoI/WoG/Mending data, round 2
Seal of Justice cast once in between each data set
http://www.worldoflogs.com/reports/rt-q ... 706&e=2970

Same stats as before

Starting with 3/3 Divinity, 2/2 EG, 0/3 RoL, neither glyph
WoL: http://www.worldoflogs.com/reports/rt-q ... 150&e=1557
Recount:
96 non-crit Wogs, 4206/4317/4455 min/avg/max
5 crit WoGs, 6372/6519/6626
91 SoI, 1449/1450/1450 (0 crits)

Finishing the prot tree to add 2/2 GbtL
WoL: http://www.worldoflogs.com/reports/rt-q ... 526&e=2025
95 non-crit WoG, 4630/4768/4909
5 crit WoG, 6982/7161/7314
86 SoI, 1449/1450/1450 (no crits)

Glyph of WoG added (still no ret talents, 3/3 Divinity 2/2 GbtL)
WoL: http://www.worldoflogs.com/reports/rt-q ... 015&e=2497
92 non-crit WoG, 5092/523/5402
8 crit WoG, 7643/7801/8011
94 SoI, 1449/1450/1450 (no crits)

Glyph of SoI added (still no ret talents, 3/3 Divinity, 2/2 GbtL, both glyphs now)
WoL: http://www.worldoflogs.com/reports/rt-q ... 514&e=2934
97 non-crit WoG, 5330/5495/5644
3 crit WoG, 8043/8094/8151
94 SoI, 1449/1450/1450 (no crits)

I should also note that I cast some Judgements in here for mana. Those may cause Mending procs depending on the parse (JotJ specced or not).

I've logged out in the gear I used, though I finished speccing the last 10 points in ret before logging out.

[benebarba]

is the list in the first post of data needed still accurate?

[theckhd]

Yes. I just removed the Mending request, since the data I collected last night should cover it.

[Rabs]

Seal of Truth: All single target attacks (including Judgement, Hammer of Wrath, Exorcism, and Templar's Verdict) can now trigger this seal.

Are you looking for PTR parses examining which abilities proc SoT or do you already have this covered?

[benebarba]

For the meta gem shield block buff - you want the lvl 80 gem, or is the 85 (eternal shadowspirit diamond) OK? (I still have an old hat with one of the 80s, and my current one has the 85). I could always do both, though the helmets they are in have different stats of course.

[theckhd]

SoI/Mending/WoG round 2, analysis

SoI
SoI is easy. From the baseline parse in round 1, SoI triggers for 1290-1291 healing every proc. In all of the round 2 parses, it's 1449-1450. This is exactly what would be expected for multiplicative Divinity effects (1.06*1.06), and too high for additive Divinity (1.06+1.06=1.12).

Conclusions:

• Divinity is multiplicative with itself for SoI.
• GbtL, WoG glyph, SoI glyph have no effect on SoI healing.

Mending

Code: Select all

        base    Div   GbtL    WoG    SoI
min    753      857   837     847    839
mean   809.9    912   898.3   905.0  898.2
max    844      960   961     958    961


Divinity is obviously affecting Mending more than once, but there's not enough data here to definitively determine whether it's 1.06^2 or 1.12. The base sample is a little too small, and the max and min are a little closer to the mean than would be expected based on the rest of the data. Based on the first test though, it's probably a safe bet that Divinity is multiplicative here too.

Regardless, it's quite clear that GbtL and WoG don't affect it at all (nor would we expect them to), and that SoI similarly has no effect. Presumably the SoI glyph only affects heals that are cast (WoG, Holy Radiance, etc.), and not healing procs like SoI or Mending. It's not clear how it would interact with Glyph of the Long Word given this information.

Conclusions:

• Mending sees the full benefit of Divinity (1.06*1.06 or 1.12).
• SoI glyph does not affect Mending.
• GbtL and WoG glyphs don't affect Mending.

WoG

<edit> tlitp pointed out that I neglected to try SoI and WoG glyphs additive with each other, but multiplicative with everything else. This turns out to be the correct result, and explains the odd Divinity result I got. See his post on the next page for the correct conclusions.

base values taken from the 66 samples in the baseline parse.

Code: Select all

       base       Div       GbtL      WoG       SoI
min    3754       4206      4630      5092      5330
mean   3863.8     4317      4768      5230      5495
max    3976       4455      4909      5402      5644

dividing by base

min    1.0000    1.1204    1.2334    1.3564    1.4198
mean   1.0000    1.1172    1.2340    1.3535    1.4221
max    1.0000    1.1205    1.2347    1.3587    1.4195


Interestingly, this data seems to suggest that WoG sees an additive benefit from Divinity (1.120 instead of 1.1232). That's very peculiar, if true. I'm not sure whether to trust this, the sample sizes are small enough that I doubt a difference of 0.3% is statistically significant. It may just take more data.

Starting from the other end, we can clearly determine that all of the other effects are multiplicative.

If the SoI glyph was additive, we'd expect a result of 1.410 or less (1.12*1.1*1.1+0.05 = 1.4052, 1.06*1.06*1.1*1.1+0.5=1.4096). Our data clearly pegs it at closer to 1.42, which is consistent with the SoI glyph being multiplicative with the 1.356 of the previous test.

Similarly, our upper bound on the WoG glyph is 1.06*1.06*1.1+0.1=1.336, well below the observed value of 1.356, which is consistent with the WoG glyph being multiplicative with the 1.234 average in the test that preceded it.

Guarded by the Light is also multiplicative. Dividing every number in the GbtL column by the appropriate column in the Div column gives a value greater than 1.1009. Any possible scenario where GbtL is additive with divinity results in a factor less than 1.0943. This strongly suggests that GbtL is multiplicative with Divinity, but still doesn't give us much insight into whether Divinity is multiplicative with itself.

As a final note on Divinity, now that we know everything else is multiplicative, let's consider the SoI column again. The two possibilities for different versions of Divinity are:
1.12*1.1*1.1*1.05=1.4230
1.06*1.06*1.1*1.1*1.05=1.4275

Our values agree very well with the first, but the second is almost 0.05 larger. Again, it's tough to make a call based on this, because this is probably near our noise threshold. And it would be incredibly weird for Divinity to have different behavior for WoG than it does for SoI and Mending. But that's the best conclusion we can draw from this data. I think it's worth going back and performing one final set of tests for WoG to clarify this:

1) 0/3 Divinity, 2/2 GbtL, WoG and SoI glyphs
2) 3/3 Divinity, 2/2 GbtL, WoG and SoI glyphs
Roughly 200 casts in each configuration would be preferable.

Conclusions:

• SoI glyph, WoG glyph, GbtL are all multiplicative
• Divinity is possibly additive with itself (for WoG only?), but multiplicative with the other effects.