• Aldriana

This was actually a topic of some debate a few months back; if you search around I'm sure you can find the thread for it. The end conclusion was that all PPM effects added since the expansion seemed to have the following proc behavior:

* On instant attacks, the proc rate is determined solely by the *base* speed of the weapon.
* On autoattacks, the proc rate is determined by one's *current* weapon speed, including all haste effects - Haste Rating, Mongoose, Flurry, SnD, etc.

So, based on the assumption that if Executioner were PPM it would obey these rules, using the hasted weapon speed is correct.

Yes, it's at least 2 procs rather than exactly 2 procs, but there's no real way to track that from the data itself. However, this should A) be a fairly small effect and B) have more or less the same effect on the two data sets, in the average. my sense with analyzing proc rates is that it's usually pretty safe to ignore that particular detail; while it means that we'll measure the proc rate as .98 instead of 1.00 (or whatever), our data sets will never be large enough that both numbers won't be in the probable range anyway, whereupon we'll assume the answer is 1 (because what idiot would put in an item with a .98 PPM effect into the game?)

Testing with an OH tends to make the data analysis less precise; additionally, the Talon test was made not while dual-wielding but while using a MH but no OH. So if you want to test with neither a shield nor an OH you're welcome to do so, although frankly I'll be shocked if it makes a difference. If you *do* wish to test with an OH, please use an OH with vastly different damage range than your MH weapon, so that I'll be able to sort out which attack came from which hand based on the combat log. Or, alternately, get two weapons of the same speed and put Executioner on both, so it doesn't matter which hand procced what... although that will increase the magnitude of the aforementioned overlapping proc issue.

The ranges I'm giving are statistical 95% confidence intervals, which (more or less) means that given the data set, it is 95% likely that the true proc rate lies within the specified range. So occasionally we'll see data sets outside those ranges, but it should be pretty rare. And, with the extended data set, I think it's safe to say that this doesn't look like a statistical fluke; rather, it appears that we don't fully understand the proc mechanic yet.

I'll run a more detailed statistical analysis tomorrow, but I suspect what we're going to find, fundamentally, is that the existing data doesn't match up with any known proc mechanic, whereupon we get to start speculating wildly and trying to figure out what the heck is going on.

 

songster

In any case, even if this factor is in play, it would tend to widen rather than narrow the observed gap. If Executioner's PPM is based on unhasted speed, then Jin'rohk's PPM should be slightly higher then that for a weapon with no haste. That's the opposite of what we're seeing.

 

mysteltainn

I seem to remember reading something that showed how armor's effectiveness scaled linearily with respect to survival time; there were other articles mostly related to warrior theorycrafting, and, obviously, I forgot the site.

Thus: I'm fishing for the address in question in order to find what difference in math or analysis leads to the discrepancy in conclusions between this and the commonly-accepted "executioner scales better the less armor your target has."

 

Xerophyte

It does scale linearly in absolute survival time, however going from 30 to 25 seconds is a lesser effect than going from 15 to 10 seconds, even if both are changes of 5 seconds and thus also the exact same amount of armour. The relative effect (i.e. scaling) is always greater for lower armour.

 

Karmon

A) Hm e.g for an assumed 1PPM 15sec effect, 2.7 speed (that gives 4.5% to procc):
If you observe something in the range 16 to 30 secs it is an at least double proc.
How big is the chance to be more than two?
I say it is 1-(1-p)^(deltaT); deltaT beeing the exess time over 15 second.
(the chance to proc at least once in the first deltaT seconds of the observed duration)

With the above numbers for deltaT= 7 sec that is about 27% and with deltaT = 14sec it is about 47.5%.
I dont call that neglegible

B) agree

 

Arindelest

I did another test, ~ an hour long, with the Talon

Here's the log: Log

 

hannigaholic

That's the point of of me saying these data sets are probaby not "large enough". We can't be confident statistically that the data is accurate in the long run.

--edit--

Okay, what Aldriana said may well hold true, but I still think a larger sample size would help

I think one thing that could be tested is that haste mechanics alter the proc chance on a proc per minute enchant. When you gain Slice-and-Dice, or Mongoose itself, your proc chance falls. If it is proc per minute then using Slice-and-Dice would reduce the procs per minute slightly (since the procs from autoattack would stay the same, but those from instant attacks would fall). It would certainly help us figure what we're dealing with, mechanics wise.

 

• Aldriana

It's only 1-(1-p)^deltaT if p is the probability of a proc each second (as opposed to a proc each attack); hence, evaluating with deltaT = 7 seconds gives an 11% chance, and at deltaT = 14 seconds a 21% chance. Now lets look at this in terms of a real data set:

In, say, the first Talon dataset, we had 25 uptime periods, 7 in the 16-31 second range, and 2 in the 31-46 second range. So, if each of those has, say, a 15% chance to actually include an extra proc, that's an extra 1.3 procs versus the 36 we counted. Hence, the actual number of procs is, perhaps, 4% higher that we measured.

Now, I'm not going to say that 4% differences are negligible in a general sense; but given that the error on our data at the moment is still a fair amount larger than that, it can safely be ignored as being less than the random variation, particularly since it applies equally across all tests.

Grinding the numbers on this:
69 uptime periods, 92 procs, 1477 swings, hence proc rate most likely lies between 4.00 and 7.5% (1.11-1.65 PPM)

Combining it with the earlier Talon dataset, we obtain:

94 uptime periods, 128 procs, 2193 total swings, hence proc rate most likely between 4.9 and 6.8% (1.08-1.52 PPM)

So, this actually lines up decently with the Brutalizer data; based on these two data sets, it appears to be a PPM effect with a proc rate somewhere in the 1.2-1.5 PPM range. However, this doesn't work for the Jin'rohk data, so we don't understand everything that's going on yet.

Thing about using SnD is that it means you need to use SS or Hemo to generate combo points, which tends to muddy the data and reduce the effect in question. In general, I've found that one needs much larger data sets to figure out what's going on by using haste effects than you do through other means. Additionally, by changing weapon speed we can get much larger differences than we can through haste (2.7 to 1.6 is a larger difference than you can get without a heck of a lot of haste rating). So while such data would be useable, it strikes me as less immediately useful than what we're doing.

 

Nateo

Here's a ~1hr log of just Brutalizer MH, nothing in OH. So you can run the log against my other logs to see if there is some kind of weird behavior with a shield equipped. Test setup is identical to before (Blasted Lands mobs, attacking from front), just no shield equipped this time.

If there's little/no discrepancy, then I guess you can lump this in with the other Brutalizer data!

EDIT: Would be useful if I actually posted the link...
Executioner Proc Rate No Shield Test

 

• Aldriana

75 uptime periods, at least 108 procs, 2424 total swings. Proc rate most likely lies between 3.6 and 5.3%, or 1.36 and 1.97 PPM. The same? Different? Hard to say.

 

koaschten

Just noticed while testing... You can't "/cancelaura Executioner". Will upload a log later.

 

Exekutor

I read the whole thread know (i think so Oo) and i read a lot about DW warriors/shamans/rogues.
But there was _nothing_ about how those enchants effects 2h-weapons.
As a Retribution Paladin, i would really know something about that.
I wear the Lionheart CHampion (so its 3.6 speed). I'm going to get the girdle an the bracers sold for heroic badgets, so i'll have 304 passive armor piercing.
Should i (or any other 2h-guy) better go on Mongoose, Savagery or Executioner (which, besinde all effects, i'd prefer for its glow *love*. but i primary want a usable Enchant, not a i-like-to-watch-my-weapon-glow-enchant )

 

Arindelest

Where do we go from here? I assume we just need simply more testing, and I will be working on another log with the Talon. Would it be useful to have a friend with the off-hand enchanted do an offhand only test (~1.5 speed)?

 

Avyrana

Since the PPM is still unknown, its impossible to say which one of the enchants is better than the other.

One suggestion: Is it possible that executioner-proccs depend on 1Hand / 2Hand like the warriors talent unbr.wrath?
It has an increased chance to generate rage per hit, if a slow (2hand) weapon is equipped.

Maybe Blizz integrated a similar effect into the new executioner enchant, meaning that there is a dependency respectively a new formula, in which the weaponspeed is also included thus effecting the procchance?

However, i hope the testing with Jin'rokh was wrong ^^ gets a bit too complicated for me

 

koaschten

No passive haste, 2.8 MH, no off-hand, feral for healing
Free file hosting by Savefile.com

 

Zxahn

I can do another few hours with jin'rohk if more data is needed

 

beetlejuice

Some ideas to solve this fast.

Redo your tests with your weapons but use haste gear. If the ppm is different then it surely is affected from actual/current weapon speed.

To further investigate it, i would suggest, enchant executioner on a 2.5 weapon and without haste gear on, collect your data. Then use your 2.7 weapon, use your haste gear so that your actual weapon speed is 2.5, and recollect data.(2.7 2.5 is a suggestion. With 4 haste items you can achieve a -0.20 weapon speed i think. ) Then make your comparison.

With these comparisons, you can figure out at least how this proc is affected from weapon speed and if haste affects the proc.

PS. Seems executioner procs as often as mongoose, wouldn't be bad idea to have both enchants on your dual weapons even on trash or arena fights, at least for enhancement shamans

 

• Aldriana

70 uptime periods, at least 102 procs, 1630 hits. Proc rate most likely lies between 5.1% and 7.4%, or 1.09 and 1.59 PPM.

So, I see three options off the top of my head:

1) 1H/Shield, 1H/no OH, and 2H all have different proc behaviors; for 1H/no OH, the proc rate is a flat ~5%. This is possible, but seems pretty unlikely to me. I have a hard time believing 1H/Shield and 1H/no OH have different proc behaviors, and, even past that, the proc ranges for the 1H/no OH tests only barely overlap, so, statistically, this is pretty unlikely.

2) 1H and 2H may have different proc behaviors; the 2H proc rate is undetermined as yet, but the 1H proc rate appears to be a PPM with a proc rate of roughly one proc per 40 seconds (1.33 PPM); this is outside the range tested for Jin'rohk, so the 2H proc rate appears to be somewhat lower.

3) Something totally different than any known proc.

Now, to me, option #2 seems the most plausible. I'm not saying it's for certain correct; but it does seem like the most likely of the options to me, purely because #1 is unlikely for several reasons, and I have a hard time believing blizzard concocted some sort of bizarre new proc mechanic just for Executioner.

 

Wodahs

Have any of the rogues frequenting this thread changed to a mh executioner / oh mongoose. I am tempted, and I understand this thread is for gathering hard numbers, I am just wondering if practically speaking, it worked well for anyone.

 

• Aldriana

If theory #2 above is correct, MH Exec/OH Mong should be double Mongoose by about 20 AP-equivalent, or on the order of 5-10 DPS. Hence, while it is most likely the best option, it is also true that anecdotal evidence may have a hard time telling the difference.

 

nelalas

Does this include Blessing of Kings, which would increase the agility contribution of Mongoose? Moreover, I think the relative value of these enchants will be interesting to look at under various buff and gear environments (i.e. a spreadsheet) once the proc behavior is confirmed and can be included. The choice for a MH enchant may end up being more personal than a global X is better than Y.

 

• Aldriana

Yes, it includes BoK. The 20 AP number was pulled from a modified version of the 0.8 Rogue Gear Spreadsheet, with the default/gear buffs it uses. I will be adding Executioner to the next version of that spreadsheet with the best estimate we have at that point for the proc rate.

 

Neckface

Since armor penetration is more than additive, how would one compute the inflection point where Executioner > Mongoose for a rogue's MH? It seems like the inflection point, if it's greater than zero, would be at an actual armor value for your target, as opposed to one's own armor penetration. Maybe start with armor value of 840 and see if Executioner is better in the best case scenario. If so, you could work backwards from there and for a given rogue's gear setup figure out which bosses favor which enchants.

 

Aten

Executioner Test - ~2H20

The tests were run with 3 mods working at same:

Procmeter - counts procs only when it gains
Procwatch - counts proc when it Gains and when it Fades
Uptimemeter - tells you % of time it was up and runing


I made for tests of ~35m each, all 4 in battle ST and only auto-attack.


Test 1 - only 1 weapon(talon 2.7) with Executioner used and in MH






Test 2 - only 1 weapon(talon 2.7) with Executioner used and in OH

This test also includes the hit from my unarmed MH






Test 3 - MH(dragonstrike 2.7) with Moongose and OH(Talon 2.7) with Executioner






Test 4 - MH(Talon 2.7) with Executioner and OH(Fist S2 1.5) with Mongoose

Dont have a slower weapon with mongoose to put on OH to this test, but the fist also dont have the high haste proc from dragonstrike.







My tests show that the uptime is around 9 minutes (average) in the 35 minutes of the test.

the PPM is the same on MH as is on OH - 1.8
(except on my first test that is 1.4, this may be cause of less Flurry uptime on this test for sure)

Im not a theorycrafter by nature by is pretty obvious that it doesnt matter in what hand the weapon is MH/OH to have the same PPM.

u can see the test screenies for yourself and debate freely.


PS: i dunno i to get the Combat logs u require, but hope this helps a bit.

Edit: My build is 17/44 with precison 2/3 and WM 2/2, also i was on my crap PvP gear for more HP

 

Borodin

Throwing this in here on a pure hunch: if there is a new proc rate mechanic based on weapon speed what if it's based off the normalized speeds (1.7 for Daggers, 2.4 for 1H and 3.3 for 2H?).