Theck's MATLAB TPS analysis - WotLK/3.x

[steadypal]

just got last word tonight,


2 observations on the target dummy


without, SOV up, it does drop off quite a bit


with SOV on, it refreshes pretty much within 5 seconds each time..



so the full 5stack SOV hit, does seem to have a chance to proc it..

[theckhd]

It's probably not the SoV proc itself, but the Holy Vengeance application. Interesting that the Last Word proc would trigger off of that when other procs were explicitly nerfed (Byrntoll). Can't wait until Putricide coughs one up for me to do some testing with.

[Repartee]

Here is a WoL for Festergut I did last night using Last Word. (Click on Buffs Gained or Buffs Cast tab)

http://www.worldoflogs.com/reports/rt-q ... 900&e=5158

98.7% uptime. Also had 98.5% uptime versus Rotface.

Last Word + 2pc t10 + HotR = sexy time

[steadypal]

with seal of command

http://www.worldoflogs.com/reports/mb28 ... details/0/

with sov

http://www.worldoflogs.com/reports/r53e ... details/5/




wow look at the huge diff in how many times it applied overall,,,, hmm working as intended with SOV? and why does that work, and not mongoose QQ

[Wrathy]

Last Word + 2pc t10 + HotR = sexy time

I agree, I was giggling all night last night when I saw 12k hits splashing across my screen! Is it wrong that I have 7900 ap as a tank when the procs line up?

[theckhd]

with seal of command

http://www.worldoflogs.com/reports/mb28 ... details/0/

with sov

http://www.worldoflogs.com/reports/r53e ... details/5/




wow look at the huge diff in how many times it applied overall,,,, hmm working as intended with SOV? and why does that work, and not mongoose QQ

I'm not sure those are statistically significant differences. I'm guessing you had the SoV glyph equipped, because you have roughly 5% difference in miss rate between the two logs (12.9% with SoCom, 7.6% with SoV). Note that you have 15 dodges and 84 parries with SoCom, but 0 dodges and 58 parries with SoV.

It's still possible that SoV applications trigger it, but I don't think those two logs would be able to tell you. The way to test would be to make sure you're soft-capped on expertise with either seal and attack from behind (so there would be zero dodges and parries). Hit-capping would help as well (fewer misses). Then take data and compare the two to see if there's a noticeable difference in uptime.

[steadypal]

okay, redid them with same stats

seal of command 291 hit, and 26 expertise

http://www.worldoflogs.com/reports/fgou ... details/0/


sov with 291 hit, and 26 expertise

http://www.worldoflogs.com/reports/7iup ... details/0/



as you can see, something is fishy, only 6% diff in uptime, but holy cow look at the difference in how many times the buff was applied 397 for socomm, vs 1099 times for sov...



so ya idk, btw hope this helped you a little better theck

[theckhd]

Just checking some numbers quick:
SoComm log:
631 melee hits, 111 crits, 255 glances, 997 total swings.
897 SoCom hits, 99 crits, 996 total
224 Vindication applications (and 18 misses)
397 Blessing of Light casts (Last Word buff)

SoV log:
633 melee hits, 114 crits, 250 glances, 997 total swings.
862 SoV hits, 134 crits, 996 total
234 Vindication applications (11 misses)
997 Blood Corruption applications
599 Blood Corruption ticks
1099 Blessing of Light casts

One thing to note: In the first parse, there's fewer BoL casts than there are melee swings. So it's conceivable that the buff is only triggering off of melee swings in that data set. However, in the SoV log we have more BoL casts than we have melee swings, which means that one of the other things in that log can also trigger it.

Vindication is present in both logs, but I think we're pretty sure that it doesn't proc anything. The only way to test this for sure is to rerun the test in a spec without Vindication and see if the procrate changes.

However, there isn't a huge difference in the melee or Vindication numbers, so the difference between the logs is most likely due to one of the following hypotheses:

1. SoV procs trigger the buff but SoCom does not, or
2. HV/BC DoT ticks trigger the buff, or
3. HV/BC applications trigger the buff

I suppose it could be some combination of those effects as well. The first one seems very unlikely, but we know that the SoV proc can't miss and the SoCom one can, so it's possible they act differently. The second is almost certainly false, unless the wording of the weapon proc is very misleading ("weapon swings"). Though it'd be easy to test - put a 5-stack up on a dummy and back away. See if any of the DoT ticks trigger or refresh the buff.

I'd be curious to see a log with no seal active. I would guess it would look identical to the SoComm log. That would rule out the weapon proccing off of Seal hits (or at least SoCom).


This doesn't make sense to me though. They explicitly changed the mechanics of HV applications to prevent Bryntoll procs. So why would they be triggering Last Word procs? My only guess is that "Chance on hit" and "Your weapon swings" refer to two separate, independent trigger mechanisms. Maybe "chance on hit" is an enchant-like proc, and "weapon swings" is something else?


I actually have a hour-long combat log of me hitting a dummy with a Blade Warding weapon and SoV active that I haven't had time to upload yet. I'm curious to see if it differs from Tree's data set a few pages back in any respect.

[steadypal]

i have my marrow wpn with mongoose on it, would u like 30 min logs of socomm vs sov and mongoose uptime? ?

[theckhd]

i have my marrow wpn with mongoose on it, would u like 30 min logs of socomm vs sov and mongoose uptime? ?

Sure, if you have time. It won't hurt anything. It probably won't tell us anything we didn't know though, unless we find they've secretly reverted the Bryntoll hotfix.

I think a few simple tests can tell us what we want to know though. Here's what I'd like to see:

Test L-1: Completed - ShoR and/or Vindication do proc Last Word.
Last Word equipped, hit-capped, expertise-capped. Vindication included in spec. No seal.

1. Position yourself so that you can hit the dummy with ShoR, but cannot auto-attack it.
2. Reset Recount
3. ShoR on cooldown 25-50 times.

The proc rate on Last word should be very high, so if Vindication or ShoR procs it, then we should see some procs in this test. If we see no procs whatsoever, then neither ShoR nor Vindication procs it, and we can rule that out of our previous data set.

Provided we see 0 Blessing of Light procs in Test #1, we then perform the next few tests:

Test L-2: Invalidated by L-1.
Identical to Test L-1, but this time use Seal of Command. If we see 0 BoL procs here, then SoCom's proc can't trigger it either.

Test L-3: Completed, HV ticks do not trigger Last Word.
Identical set-up to the previous two tests, but this time with Seal of Vengeance.

1. Smack the dummy once to give it a Holy Vengeance stack and back away. Make sure you only get off one melee swing, and nothing else. Wait until the debuff falls off.
2. Repeat 25 times or so.

This will test two things: Can HV applications proc BoL, and can HV damage ticks proc BoL. If we see 2 BoL casts with any individual melee swing, then HV applications proc it. If we see a BoL proc at any point after we've backed away, then the HV damage ticks proc it.

Test L-4: Invalidated by L-1.
Identical set-up to L-3. The execution of this one is a bit trickier though.

1. Smack the dummy once to give it a HV stack, and turn so that you cannot melee, but can cast ShoR.
2. Cast ShoR on cooldown until the stack drops off.
3. Repeat 25 times or so

This is essentially test L-2 again, but with SoV. The previous test should tell us what to expect here - if only HV applications trigger it and nothing else does, then we'll see potential BoL procs when we apply the HV stacks but not from the damage ticks. If we see BoL procs coinciding with our ShoR's, then the SoV proc must be causing it (again, provided that Test #1 is successful).

[theckhd]

And while we're requesting data, I may as well re-post the tests that tlitp and I asked for about Judgement procs. I formulated my tests in terms of binary results, whereas he was looking for statistical data:

Test JK-1 (Judgement-thecK-1) Partially Complete - at least one of Judgement, JoJustice, or HotC procs Berserking
Spec something like this: 0/33, no HotC or Vindication. Equip a slow weapon (1H, 2H) with a ppm-based enchant (goose/zerk)

1. Stand far enough away from the dummy so that you can't melee it
2. Use Judgement of Justice on the dummy 100 times.

Since it might be asking a lot for people to respec for this, there's a simpler-but-almost-as-good set-up. If your off-spec is Ret, that's fine too. Just use your default ret spec with HotC and your berserking-enchanted ret weapon. The slower the speed, the better.

What we're looking for in this test is any Berserking procs. If we get 0, it tells us that:
Judgement itself doesn't proc PPM enchants (You have a 99% or higher chance to get a proc in 100 casts if it did)
The Judgement of Justice debuff application doesn't proc them either
HotC (if in spec) doesn't proc them either
Vindication (if in spec) doesn't proc them either

If we get any procs whatsoever, then we need to do some further testing (probably using the statistical tests tlitp wants).

If we get 0 procs, especially in parses with HotC and Vindication present, then we move on to..
Test JK-2 (Judgement-thecK-2) - Invalidated by JK-1
Identical to test #1, but use Judgement of Wisdom instead of Justice.

Test JK-3 (Judgement-thecK-3) - Invalidated by JK-1
This time, respec prot (and into Judgements of the Just). Otherwise, it's identical to test #1. This might take a while, as you won't have Judgements of the Wise for mana regen. Just DP on cooldown and chug pots, if necessary (or wait, I guess, it just means it'll bore you to tears for longer).


Now for the more statistical tests. Note that I've changed the spec tlitp asked for - he said that the spec should have JotJ, but the 42/7 spec he linked doesn't go deep enough in the tree to pick it up. I'm assuming he meant that you should pick it up (and that the 42/7 was just to guide you away from picking certain prot talents), but I'd hate for someone to have wasted their time doing a bunch of testing with that exact 42/7 spec. If I'm wrong on this, I'm sure he'll stop by and correct me.

Test JT-1 (Judgement-Tlitp-1)
Spec something like this: 0/33, no HotC or Vindication, but with JotW. Equip a slow weapon (1H, 2H) with a ppm-based enchant (goose/zerk). Use Seal of Righteousness.

1. Position yourself so that you can't melee the dummy.
2. Judge one of the level 60 dummies 250+ times

Test JT-2 (Judgement-Tlitp-2)
Identical to JT-1, but this time spec prot and pick up JotJ (some variation of 47/7, again, no Vind or HotR)
Use SoR.

1. Auto-attack the dummy while Judging on cooldown (every 8s).
2. Do this until you've Judged at least 250+ times

In this spec, we'll need the auto-attacks to keep DP up (at least for long data sets like this, 250 Judgements is around 34 mins).

Test JT-3 (Judgement-Tlitp-3)
Identical to JT-2, but with SoCommand. Don't let SoCom cleave to another dummy (make sure of this before you start logging, and don't move once you've found a good spot).

I think there are three distinct tests to be carried out :
1. J/debuff (W/L)
2. J/JotJ/debuff
3. J/SoComm_J/JotJ/debuff

As for the methodology itself :
1.
- a slow weapon (1H, 2H) with a ppm-based enchant (goose/zerk)
- a variation of 0/33 (do not get HotC and Vindication, do get JotW)
- SoR
- hit one of the level 60 dummies only with J
2.
- slow weapon and ppm-based effect
- a variation of 42/7
- SoR
- AA+J
3.
- slow weapon and ppm-based effect
- a variation of 42/7
- SoComm
- AA+J (do not allow SoComm to cleave)

Please keep in mind that we need samples of at (very) least 250 J casts each, in order to reduce the impact of statistical fluctuations.

[theckhd]

By the way, I updated the weapon analysis plots. They now properly reflect 100% uptime on Last Word.

[steadypal]

really hoping i did what u wanted, LOL


exp cap, hit cap, used seal of command, and judge of justice...

http://www.worldoflogs.com/reports/035j ... details/0/

spec has hotc, but does not have vindication though...



k, just did the shield slam test, hit capped, exp capped, no seal only 25 shield slams though

http://www.worldoflogs.com/reports/7n3j ... details/0/

[theckhd]

exp cap, hit cap, used seal of command, and judge of justice...

http://www.worldoflogs.com/reports/035j ... details/0/

spec has hotc, but does not have vindication though...

Ok, so 81 Judgement casts, 81 JoJustice debuffs applied, and 81 HotC debuffs applied. We also had 36 Righteous Vengeance ticks.

Berserking procced 4 times. So at least one of those things triggers it. This throws JK-2 and JK-3 out the window, because we have to filter down this set of parameters first. It looks like we really will need a larger data set taken in the 0/33 spec (so no RV and no HotC). I guess we'll need to run tests JT-1 through JT-3.

k, just did the shield slam test, hit capped, exp capped, no seal only 25 shield slams though

http://www.worldoflogs.com/reports/7n3j ... details/0/

This is very interesting. This was test L-1, which I expected to give 0 BoL procs. However, in 25 ShoR's and 6 Vindication procs, we got 11 BoL procs.
This definitely means that ShoR procs it, which I hadn't expected (though I guess it does make sense). We can't rule out Vindication either, sadly. This throws tests L-2 and L-4 out the window. Test L-3 could still be useful though (confirming that HV dot ticks don't trigger it).

I'll have to think a bit, but I believe this means we'll need to make some large data runs to get enough statistical data to sort this out. Null tests using ShoR aren't going to be helpful if ShoR itself procs BoL.

[steadypal]

pretty sure the sov tick does NOT proc it, didnt log it, but i was observing it for a while, and hmm, recount said i did 100 judgements... idk whats up



tell me what i can test w/o respeccing,



ret spec does not have vindi, but has hotc

prot is the normal 53/18