Iquark

I posted this in your spreadsheet thread, and as a user of the Rod I'm relatively certain that this still holds true:

That was the data out of my Proc Watch, and if needed I may be able to find a SS of it, but I am not sure if I still have one. In over a month since having the Rod, It doesn't feel as if the chance has changed at all, still a relatively crappy proc for what seems like a cool idea.

Still a fun weapon though!

 

Dontmindme

In response to losing hit% without weapon skill...
I recall in one of the Rogue Discussion replies (or the Warrior/Rogue Weapon Skill discussion) someone making the claim that the chance to miss per weapon skill doubles if your weapon skill is not within 10 of the target. Thus, based on the 0.1% to hit per weapon skill difference (as postulated), against a level 63 (boss), the base miss chance would actually be 27% if you only have 350 weapon skill vs. 25.5% with at least 5 weapon skill. That could account for the additional miss chance you seemed to have.

 

Sarlunas

Since I regularly find myself explaining why combat daggers is no longer the top dps spec I decided to do some testing to make sure I wasn't lying about the extra procs. Obviously you get 50% more ss than bs which means more chances to proc but in addition to that: does each special with a slow mh weapon have a higher chance to proc a ppm effect than each special with a fast weapon? Since testing bs is a lot more annoying I decided to simply test ss with 2 different weapon speeds, hopefully bs doesn't have any special proccing mechanic.

Anyway I got 2 maces with no dps stats (not swords because i have sword spec and I'm paranoid), one at 2.9 and one at 1.7 speed, best I could find on ah. Also used an offhand item for no real reason. I decided to use both dragonspine and romulos at the same time to give a bit of extra data since I knew the number of hits would not be quite big enough and since there doesn't seem to be any solid consensus on whether they are actually ppm or not. I went to pick up one of Razelikhs finest and turned myself so as not to get any white hits. Logged about an hours worth of ss spam with each weapon. Used blind+bandage and the occasional vanish+eat, I turned off the logging after each vanish and started again when I had him back in position. In order to not get any hidden double haste procs I stopped the ss spam every time I got haste and resumed when it faded, I forgot once or twice but I'm fairly sure I didn't get any refreshing procs.

2.9 speed mh:
1003 ss
47 Romulo procs - 4.69%
45 Haste - 4.49%

1.7 speed mh:
899 ss
23 Romulo procs - 2.56%
24 Haste - 2.67%

WWS links:
Wow Web Stats
Wow Web Stats

If anyone wants the logs themselves I can ofc upload them somewhere.

I know the sample size is not big enough but look at the pretty numbers! They're so pretty! It certainly looks like both procs are ppm and that mh speed does indeed affect the proc rate of specials. They also seem to have very similar proc rates. In any case I at least managed to convince myself so I thought I'd share.

One interesting result I had not even considered was the fact that I got 0 procs from my Thundering meta gem during over 2h of ss spam. It did however proc on a white hit when I repulled the servant after eating, this happened twice. I guess it's safe to say that it doesn't proc from specials. On the whole procs seem a bit weird and inconsistent...

 

Wodahs

Tonights raid, with 366 weapon skill
Narjentus 26% glance
Akama 22% glance (may not be the best example due to short fight, but 2546 dps) Wodahs - WWS
Gorefiend 26% glance

previous fights
Winterchill 19%, 24%, 24%, 21%
Anetheron 25%, 25%, 24%
Kaz'rogal 22%, 22%, 27%, 21%
Azgalor 21%, 25%, 21%
Archimonde 25%, 21%, 27%


I do not think that weapon skill statistically is decreasing glancing blows. I could be misunderstanding what I am reading, but it looks to me like weapon skill is good hit and crit, but not a reduction in glancing blows.

 

Karmon

Miss% against Hydross.
Rogue 41/20/0; 278 hit. Dualwielding daggers. No item with +dagger or Weaponsexpertise.
Only autoattacks counted

444 crits, 846 hits, 0 blocks, 487 glances, 113 misses, 105 dodges, 28 parries,
2023 total attacks
554579 damage, 274.136925 damage per swing
219753 hit damage, 259.755319 hit damage per swing
94591 glance damage, 194.232033 glance damage per swing
21.948% crits
41.819% hits
0.000% blocks
24.073% glancings
5.586% misses
5.190% dodges
1.384% parries

Miss% is way off the predicted 2.87%
Any filehoster besides filefront or savefile for uploading combatlog?

 

RPZip

The beauty of being multiplicative is that it shouldn't matter.

Ex: Rogue without Lethality crits. 100 + (100) = 200% damage.
Rogue with Lethality crits. 100 + (100 * 1.3) = 230% damage
Rogue with Lethality and Relentless Metagem crits. (100 + (100 * 1.3)) * 1.03 = 103 + (100 * 1.3) * 1.03 = 103 + 103 * 1.339 = 236.9.

No matter what way you slice it, the answer should be the same.

 

• Aldriana

Perhaps "order" wasn't quite the right word. The question is whether it stacks additively or multiplicatively. When I first saw Relentless Earthstorm Diamond, my assumption was that it applies the same way that Lethality does, i.e., a crit, instead of doing 200% damage, it would do 100 + (100 * 1.03) = 203% damage, or 100 + (100 * (1 + .3 + .03) = 233% damage with lethality. However, it could also stack multiplicatively, and do (100 + 100) * 1.03 = 206% damage, or (100 + 100 * 1.3) * 1.03 = 206.9% with lethality.

Note that one of these results yields twice the damage result of the other, so it's sort of important to the quality of the item to figure out which is correct. It should be pretty easy to test, though; just go kill rats with a low-range weapon with and without a RED equipped (but the same AP) and compare the damage ranges. I'd do it myself except I don't have a RED in any helm at the moment.

 

Pork

Tested this out on an undead mob to avoid additional modifiers (i.e. Murder, Beast Slaying):

Weapon: Blacksmith Hammer - 1-2, 2.0, .8 DPS
AP: 1638 = 117.0 DPS (used a multiple of 14 to avoid potential rounding issues)

Non-Crits: 159, 160

Projected Crits:
d + (d * m) = 322.77, 324.80
(d + d) * m = 327.54, 329.60

Actual Crits: 329, 330

Looks like the modifiers are multiplicative as far as I can tell.

 

• Aldriana

Updated the initial post with the answers that have been determined so far. There were only two I felt confident in the answer for, although there's a couple others we've seen some preliminary results for that I think we're getting close to having an answer for. Again, if you have data sets that would help, please post them. If you'd like to test one of these but don't know how, post in this thread or send me a PM, and I'll give you an idea of what sort of testing would be helpful.

 

Dampfbrumsel

Just come to think of it, how are parries and dodges handled during the data collection? I don't know if I got this straight, but this kind of data is measued in procs per minute assuming a 100% chance to hit, right? Since that can hardly be achieved against regular mobs, your hit chance is probably around 98% for the blasted lands servants. Hence ProcWatch will not give you the actual proc chance of the effect, but rather an individual value, which you will have to extrapolate depending on the amount of swings that did not hit.
So how has this been handled so far?

 

Oie

Okay, so I was going to get more data, but after my first two tests it seemed too good to be coincidence so I thought I would go ahead and post what I found now. I conducted the tests on Blasted Lands mobs and in all samples had 0 misses/dodges/paries. Using procwatch and the macro to cancel the haste buff as soon as it came up, I found the following results:

Using dual 1.3 speed weapons, with no additional haste effects:

Hits: 12194
Procs: 512
Time: 1:50:29
Hits/proc: 23.8
Procs/Min: 4.6

I repeated this using dual 2.6 speed weapons, althought these were enchanted with mongoose:

Hits: 17656
Procs: 1412
Time: 5:03:27
Hits/proc: 12.5
Procs/Min: 4.6

Now, obviously, procwatch counts each proc twice, so for dual wield autoattacking, it seems to be 2.3 ppm. I plan to do a bit more testing, but seeing as how "good" these initial results were, I wanted to get them out them quickly.

 

• Aldriana

The blasted lands mobs are something like -15 levels relative to the effective level (including weapon skill) of most rogues. As such, they don't tend to dodge or parry very much. In fact, in all my time testing against them, I don't think I've ever seen them do so.

In terms of how it would be handled should it happen: I don't use the PPM number generated by ProcWatch; I count the number of hits and the number of procs, and work out my own math. So I don't think this would be a particular concern even if it did happen.

30k hits.... wow. I salute your data collection abilities.

Well, lets see here. 256 procs in 12194 attacks is a 2.10% proc rate, with a standard deviation of .13%, which means the true proc rate lies between 1.85% and 2.35% with 95% confidence, which, with 1.3 speed weapons, equates to between .854 and 1.085 PPM.

Meanwhile, 706 procs in 17656 hits is a 4.00% proc rate, with a standard deviation of .15%, yielding a true proc rate between 3.71% and 4.29% (again, 95% confidence). With 2.6 speed weapons, this equates to a true proc rate between .856 and .990 PPM.

So, observations:

1) It's pretty clearly not a flat percent chance.
2) It's almost certainly a PPM
3) Right now, the best guess under the "round numbers are a good thing" rule would be 1 PPM. Under the hypothesis that it is a 1 PPM proc, the first data set is about half a standard deviation from the mean (.51, if you want to be precise), and the second data set is 2.18 standard deviations from the mean. Now, 2.18 standard deviations has a significance value of 2.96% - that is, 2.96% of the time the data will be at least 2.18 standard deviations from the mean. So that number, while outside of our ideal range, isn't totally out of the blue, either.
4) The other thing that comes to mind is that this method of testing might not totally eliminate overlapping procs; if, for instance, both weapons attacked at the same time, they could, theoretically, proc at the same time, which wouldn't show up here. So it's not inconceivable to me if these numbers are *slight* underestimates of the true proc rate, which would put the 1 PPM theory in better shape.

So, on the whole? This data points pretty strongly toward 1 PPM. Any objections to taking this as the accepted proc rate?

 

Oie

As for #4, that actually came to mind for a specific reason. During my 1.3 speed weapon tests, the attacks did not line up, something I didn't really give any thought to at first. When I came around for the 2.6 test, I noticed almost immediately that, to my naked eye, the offhand/mainhand attacks seemed perfectly in sync (made blatantly obvious because of the noticeably long 2.6 wait time in between attack sequences), and the idea of Dragonspine double procing at the same instant was something that bothered me a bit but that I dismissed, probably ill advisably now that you bring it up.

Also of note, which I didn't think of at the time I was writing my first post, most of my gear had broken by the end of my 5 hour auto-attack marathon, and I recall seeing some misses, though I wasn't paying particularly close attention (can you blame me?). I'm fairly certain the gear breaks did not start to crop up until at least 4.5 hours into the collection, so I thought it negligible at the time.

1 PPM sounds good to me. I'm still up for collecting a few more samples if you feel it would help. I have double 1.5 speed weapons, 2.8 speeds, and I could probably get something in the middle range (say, 2.0).

 

• Aldriana

I'm certainly not going to say no to more data, so if you feel like testing further, go right ahead. More data is always better than less data .

 

songster

Well, we can guesstimate it. 1 ppm on a 2.6 speed weapon comes out at a 4.33% proc rate per hit, or 0.1877% chance of a double proc. You have 17656 hits in the test, which is actually ~8828 "double hits" if we neglect the edge effects from misses, dodges, etc.

That means the expected number of double procs is 16.57, under the assumption that it is indeed 1 PPM. That gives a "correction factor" of 722/706 to the values calculated in #37, indicating that the 95% limits on the true proc rate are .875 and 1.01 PPM. Looking even more promising.

 

• tsigo

Here's some 4pc T5 data. Used the macro to cancel Coup de Grace instantly.

Kamien - WWS

36 procs in 2094 landed melee attacks.

Definitely no internal cooldown.

And just for fun you can see some Skyfire Swiftness procs there too.

 

bredar

I heard there's a cap to +haste. Something along the lines you can't go below a certain weapon speed that's hard coded in the game. Is this true or false?

 

Telani

False, unless the cap is extremely low. I've seen weapon speeds as fast as .19.

 

Delore

I've been looking around for the hit cap for rogues is and I found it. To my knowledge it it 308 hit rating if you have 5/5 Precision and 2/2 Weapon Expertise.

Now doesn't it make sense to go for this hit cap first to max out the potential of your crit, AP, and procs like trinkets and combat potency(you never miss)? How come high end rogues, for example, in Nihilum aren't even close to the cap and waving around 250? Can anyone explain please?

 

• Aldriana

As a general rule, any statement of the form "you should go for X before worrying about Y" is wrong.

Yes, hit is a powerful stat, and yes, it is generally desirable to be up near the hit cap. However, it's all a matter of tradeoffs, and if you get enough AP, Crit, or other stats to compensate the loss of hit, it can be worth it. In particular, T5 and above gear tends to be light on +hit relative to Karazhan + T4 level gear; as such, any rogue approaching this level of gear will tend to find their hit dropping to around 250, but picking up huge gobs of other stats that more than compensates for the loss.

 

• Aldriana

That's the weapon speed for this test?

 

Delore

What about for gem selection?
If you take a look at Cloze <Nihilum>
The Armory

He clearly has some hit gems but he also could reach closer to the hit cap but chooses not to. Is this because you don't need to close your miss rate to 0% and you could leave yourself opened to miss? Why doesn't 8 hit rating yield more dps than 8 agil when your not even close to the hit cap?

 

• Aldriana

Answer? It does. Why he doesn't use them? I have no idea, why don't you ask him?

Long story short? Just because someone is in an advanced guild doesn't mean they know what they're doing from a theorycrafting point of view. I've seen any number of people from top guilds with totally nonsensical gear and spec decisions.

Now, can we abandon this thread derailment and get back to testing stuff, please?

 

Delore

Sorry, but thank you I really appreciate your answers!

 

• Aldriana

So, Roq sent me a couple of data sets relating to the proc rate of TSD. If you want the raw data, PM me and I'll send it to you. On to the conclusions:

1) No two procs (out of 98 procs accross two tests) were less than 40.172 seconds apart. As such, let us assume the previously-hypothesized 40 sec hidden cooldown is correct.

2) Therefore, I wrote a short program to count the number of attacks that occurred at least 40 seconds after the last proc - that is, at a time when it's theoretically possible for it to proc, assuming the 40 sec hidden cooldown.

In data set one - made with 2.6 speed weapons - out of 2619 total attacks, 998 were elgigible for procs, and there were 53 procs, yielding a 95% confidence range of 3.92% to 6.70%.

In data set two - made with 1.3 speed weapons - out of 5518 total attacks, 2626 were eligible for procs, and there were 45 procs, yielding a 95% confidence range of 1.22% to 2.21%.

Observation 1: those probablity ranges don't even come close to overlapping, so it's not a flat %.

Observation 2: converting them to PPM, we find that they equate to ranges of .56 to 1.02 PPM (for the 1.3 speed weapon) or .90 to 1.55 PPM (for the 2.6 speed weapon). Now, admittedly, they don't overlap by a lot, so more data might be in order - but, from the perspective of these data sets, it appears to plausibly be a 1 PPM enchant with a 40 sec hidden cooldown.