I'm gonna try to grab some time this week end and do one run with WF and no haste, another with haste stuff and no WF.
However, I can only go up to 162 haste rating. Would this be enough for a test, or should I get some badge stuff?

I know I read that post you referred to, but I can't find it.


Here's the starting point for the first "SoC scales with haste?" post I made back then, on page 69:
http://elitistjerks.com/f31/t17193-p...69/#post616404

I'm not going to link every post that was made on the subject, but here are a few highlighted posts.
http://elitistjerks.com/618483-post1722.html
http://elitistjerks.com/621149-post1764.html << good point on math model to use.


After 5 pages, we end up at
http://elitistjerks.com/625641-post1844.html

Which says:
I couldn't find the post you referred to here. I remember reading it, but the forums seem to have ate it. (I tried searching for posts by you: http://elitistjerks.com/search.php?s...8&pp=25&page=8 there's a gap between 1/25 and 2/4)


So we don't have the numbers for the 7 heroism test - but I'm gonna make some guesses - 7 heroisms is 280 seconds.

If we used a 3.8 AS weapon, 280 seconds -> 280 / 3.8 * 1.3 = 95.8 swings
expected (haste scales) SoC: 95.8 * 44.3% = 42 procs
SD = 4.9
expected (no haste scale) SoC: 95.8 * 34.1% = 32.7 procs
SD = 4.6


What was the actual # of procs? If there were, say, 35 procs, that looks like the "SoC does not scale with haste" proc #, but it's still within 2 SDs of the expected value for haste scaled SoC. (we can expect 68% of values within 1 SD, 95% within 2 SD) We still have some ambiguity.


There's one more post I found that's relevant:
http://elitistjerks.com/628413-post1882.html

Unfortunately, the WWS are gone, so we can't take a 2nd look at the raw numbers. This one is also funky because it's a low-level attacking a higher level target (As far as I can tell from the replies).



So no, I don't think the past tests were conclusive, since they were contradictory in the first place. Then consider that all those tests were done 1 major patch ago, which adds the potential that something was stealth-changed.

The main reason I want more numbers on this is that the chain proc model does not satisfactorily explain all the parses. There's a set of numbers with no haste but with WF, and we do not see the numbers suggested by the chain proc model. (ie: no hasted weapon, 100 normal swings, 30 WF, should see around 130 * .40 = 52 SoC procs; but we see 40~ SoC procs)



We "knew" WF didn't proc off SoC - but the data shows that WF proc rate is higher when SoC is used. Or can be higher, at least.

We "know" SoC doesn't scale with haste, but we have data showing PPM much higher than what we should be getting. Why? We have 3 different models, (chain proc, WF procs SoC if it was proc'd by normal attack, WF does not proc SoC but SoC scales with haste) and all seem to have problems explaining parts of the data we have.


Needs more testing. = P

EDIT: And sorry for all these walls of text. I need to organize my thoughts better for easy skimming.

I wouldn't try to analyze that WWS too much.

Somehow I got stuck with the job of swapping Shamans around for Herosims, so I wasn't pushing my cooldowns hard about halfway through the fight while I was moving people and yelling at them to blow Heroism (I really was only hitting CS and judging once in a while). I also lost Windfury for a few seconds at around the 3:30 mark because I swapped with a resto to Heroism the melee.

If you do want it though, Weapon is a Cat's Edge (3.5) with 17 passive haste rating, 2 sets of drums and 1 Heroism.

I know it's "off" (in that the PPM is less than 7).


However, if you had SoC uptime of 100%, then the amount of WF uptime is irrelevant.

You had 103 normal swings, which should contribute 7 PPM under the chain proc model.
You had 29 WF procs, which should contribute 1.4 PPM under the chain proc model.

The effective PPM is less than 7 - so either our understanding of SoC's proc mechanics is wrong, or SoC uptime was less than 100%, or some combination of both.


Personally, ever since we've started discussing this, I've found I have a habit of judging *right* before I swing - and with SoC, non-procs look just like non-sealed swings.

I went over this statement and that's actually not necessarily true under "chain proc model". It's true for Model 1, but here's what actually happens with chain proc:

Example 1 (3.8 weapon speed):

A 3.8 weapon speed would generate 60/3.8 * 1.2 = 18.95 white attacks (autoattack + WF)

Under chain rule, you'd have: 3.8 * 7 / 60 = 44.3% chance to proc SoC, which at 18.95 attacks would give you 8.4 ppm.

Example 2 (2.0 weapon speed):

A 2.0 weapon speed would generate 60/2.0 * 1.2 = 36 white attacks (autoattack + WF)

Under chain rule, you'd have: 2.0 * 7 / 60 = 23.3% chance to proc SoC, which at 36 attacks would give you 8.4 ppm.


Looks ok so far?

Now we add the chain procs:

In example 1, those 8.4 ppm SoC, would each have a 20% chance to proc WF = 1.68 more WF attacks, which would in turn possibly chain proc 0.7448 more SoC attacks (at 44.3% chance each), putting the overall proc chance at 9.1448 ppm.


In example 2, those 8.4 ppm SoC, would each have a 20% chance to proc WF = 1.68 more WF attacks, which would in turn possibly chain proc 0.392 more SoC attacks (at 23.3% chance each), putting the overall proc chance at 8.792 ppm.

For accuracy (though it becomes insignificantly small if you recurse it past the first chain proc) I'll do the 2nd chain proc recursion:

Example 1: Taking the 0.7448 additional SoC attacks, would give you 0.14896 more WF attacks which would in turn give you (at 44.3% chance) 0.066 more SoC attacks, bringing the total after 2nd recursion to: 9.21 ppm

Example 2: Taking the 0.392 additional SoC attacks, would give you 0.0784 more WF attacks which would in turn give you (at 23.3% chance) 0.01829 more SoC attacks, bringing the total after 2nd recursion to: 8.81 ppm


As you can see the variance between the first chain proc and the 2nd is very small, so it can probably be ignored in the future.

Anyway, taking 3.8 speed and 2.0 speed pretty much gives us the extremes. This means that under chain proc model, with WF, your SoC proc chance will always lie somewhere between 9.21 ppm and 8.81 ppm (not the previously claimed 8.4 ppm, I think I never actually spelled it out in math was working with proc chances rather than ppm in all my previous examples).





Yea I've seen many people link that specific post from the wow boards thread and every time I have to shake my head and say: Procwatch is a "very" bad tool for this.

How do you know what conditions those attacks were made under? How do you know he didn't have WF for some of them? Is there "any" guarantee how he ran those tests, what other variables might have influenced them?

To me a procwatch image is just someone saying "hey look the clouds are forming a bunny in the sky!", I've used it a lot back then and I know how easy it is to mess up the results. Unless the results using procwatch are in a very controlled test and coming from a first hand source you can ask about those conditions, I really tend to disregard "look someone's procwatch said this" results.


A problem a figured about all our blasted lands tests: They might be very valid to calculate proc "%" out of the number of swings, but they should not be used to calculate "ppm" straight out of "duration". The reason being are all the parries you yourself are doing (you parrying the mobs attacks), giving you a lot of "up to 50% faster" swings after those parries during that "duration" (= more attacks than you expect).



The number of procs was actually exactly 35.

I edited it in pretty late so I guess you missed it:



Well, lets not forget that, neither model will ever "satisfactorily explain all parses", since the crushing majority of the WWS logs we have here are only 6 mins or less and will suffer under RNG.

I'm not saying "this is a blanket excuse out of everything that doesn't match", but you need to always remember that what we're going from here at the moment is "correlation" of a lot of data "suggesting" the same, while fully admitting that a lot of it is messed up due to RNG.

What I mean to say by this is that "a lot of results" saying one thing or the other can strengthen (though not prove) either of our current working theories through correlation, but "one or two" results saying the opposite are not enough to disprove it (rng!).



Yea this is very true and something I always regret with blasted lands test. Make sure you use the number of attacks and procs rather than the ppm in the end, since you will have too many attacks for that attackspeed/duration to get a correct ppm due to parries.

Also noting down the number of parries can be useful.

At 162 haste rating, assuming WF scales with haste, you should notice a ~10.279% increase in SoC procs or a 7.71954 ppm. My memory on statistics isn't that fresh, so maybe someone can state how many results would be needed in order to statistically rule out RNG when distinguishing between 7.71954 ppm and 7 ppm.




(Damn these posts just keep getting longer ;/ I really hope we get a conclusive blasted lands test so we can tell heads from tails.)

In order to get rid of the issue with parries for Blasted Land mobs, might it not be a good idea to get a friend to tag along, preferably another paladin who, without being in the party keep Righteous Fury up so that the mob will be attacking him rather then the one doing the actuall test? That way you can easily make sure that you will hit the mob from behind and get rid of the problem with parries messing up the numbers.

In order to easily remove parry from the equation for a long term test, there are unkillable ogre spirits behind the area where you fight king Gordok in Diremaul north, that don't aggro. It shouldn't be too difficult to get back there with a person or two. I'd go do some testing there myself, but my paladin has zero haste rating at the moment.

Pretty much. I swap Seals around .1 or .2 on my swing timer and that usually works. The only problem (well, the problem I'm having) is when trying to fit Judgement, CS, and the occasional Consecrate.

My 1st attempt and raid seal twisting was during Brutallus the other night. It felt really... awkward.

(omg 1st EJ post)

Would seal twisting make [Libram of Divine Judgement] a better choice over [Libram of Avengement]? I'm considering trying out seal twisting tonight on Illidari Council since it seems like a pretty good fight for it seeing as there is no exorcism usage and I rarely drop consecrate to limit confusion. The only thing I'm concerned about is the possibility of going out of mana since it is such a long fight and I tend to chain haste potions. Has anyone had experience trying it out on that fight (or Illidan for that matter), or are there any suggestions of a better fight to play with it on?

Has anyone had any luck on finding an effective twist rotation to get Rank 1 SoCr worked in before your Crusader Strikes. It has a 1.5 GCD on the seal cast so seems like the challenge is how to get Seal of Command back up quick enough so that you don't miss the SoCommand Procs after you cast the SoCr and then Crusader Strike.

With seal twisting, at least as I see it, you shouldn't be doing Judgement of Command. If you could, then you could simple add an equip line to your seal macros and get to use both anyway.