I recall seeing data from Simcraft datamining that every stack past 30 gave an increasing 5% chance for the finisher proc. That backs up 40 stacks being the about average placement, with at that point it being just as likely to happen as not.
I'm not sold on that right now. If it really were 5% increasing per stack then the average proc rate should be much closer to stack 35-36 than it is 40, even at these low sample sizes. If it were .05*(stack-30), then the odds of reaching stack 40 without a proc would be .95*.9*...*.5 = 3.3%. The fact that it's an average of 40 implies something else to me.
Hopefully more people will crawl out of the woodwork soon so we can get more data to verify what is going on, I'll probably dig around public WoL parses tomorrow.
Another issue that I didn't really bring up was the way the combatlog reports which stack triggers the proc. Like many procs, there's a delay between what triggers it and the aura itself. Has anyone looked into whether or not Fury comes before or after the stack of Shadows that it corresponds to? I assumed it came right after the relevant Shadows stack, but the tooltip says it's after the 30th stack and I had some fury procs on the 30th stack and none on the 50th.
UPDATE: It doesn't necessarily require a stack of Shadows to proc. I don't know if they proc exclusive to eachother, or have a chance to proc together at the moment.
World of Logs - Real Time Raid Analysis
[19:39:38.283] Pathal gains Shadows of the Destroyer (38) from Pathal
[19:39:44.650] Pathal gains Fury of the Destroyer from Pathal
[19:39:52.001] Pathal gains Shadows of the Destroyer (2) from Pathal
To add another layer of confusion; do we know what triggers Fury? The tooltip is predictably ambiguous as to whether Fury is triggered by the Shadow proc itself or by melee attacks. And even if we did know for certain it was triggered by melee there's still the (far less significant but still relevant) question of whether a your Shadow proc has a chance to be a Fury proc instead or whether your chance to proc Fury is parallel to Shadow (in which case a melee swing might, theoretically, proc both).
I'm not sure how we would test that. I imagine it would proc much quicker than it does for Combat, however, assuming all specs had an equal proc chance if it could proc off of melees rather than the working off the Shadows proc mechanic.
It's clear now that Fury doesn't need Shadows to proc, whether that's an if/elseif setup, or if+if I don't know at the moment, though that should be testable but may be irrelevent. Does anyone have any suggestions on what they might want to know for modelling's sake? Everything suggests it procs around stack 40 on average.
I guess what I'd most like to see is the stack count when Fury procs for a large sample (ideally several hundred if not 1000+) of procs. And then I'd like to make histogram of stack count versus frequency of proccing at that stack count. As the shape of that graph will tell us an awful lot about how its implemented.
I've been storing them here so far. I will of course keep adding to it over the next couple days. It's still to chaotic to draw decent conclusions in my opinion.
EDIT: The link has been removed, if you want to see it, PM me. The resulting chart is a somewhat flat distribution, with each stack being roughly the same size.
Actually, I think that tells us plenty. In particular, it tells us what its not.
If it were a situation wherein there was a specified chance to proc on (or after) each Shadows proc, we'd expect to see a distribution that tailed off towards one end of the distribution or the other - that is, if it had a chance p to proc on each Shadows with probability p, we'd see n*p procs at 30 stacks, npq at 31, npq^2 at 32, and so on, until for 45+ stacks the odds of it triggering on that Shadows is very low. So we'd expect to see something that (in very rough terms) tales off like a normal or Poisson distribution, perhaps with a spike at 50 to make sure that it always goes off.
Similarly, if the chance to trigger scaled up, we'd see this effect but even more pronounced - i.e., if it were a 5% chance per stack past 30, we'd see a distinct peak in the distribution around 35, and it would again tail off rapidly as you got to 45+.
Notice this isn't what we see. Not at all.
What we see is an admittedly noisy distribution, but one that clearly isn't tailing off at either end of the range. So this is not a case where there is a chance to trigger Fury on each Shadows, which is checked and rechecked for each successive proc. To me, this implies that they are generating a random number between 30 and 49 (notice that there are no procs at 50 stacks), and letting it trigger after that stack occurs. And while they might be picking numbers nonuniformly out of that range, I can't imagine any reason why they would, and see no evidence that they are. So my guess is that the number of stacks at triggering is uniformly distributed from 30 to 49, inclusive. Obviously more testing to confirm this would be nice, but my guess is if you run a chi-squared analysis against the uniform distribution you'll find that it matches up pretty agreeably.
Wouldn't an exponential formula generate the same layout though if it turns out to be an even/flat distribution? It makes more sense from a programming point of view (it may or may not be easier to model with) if it's a recursive or single formula, and not a /roll 30-49 that occurs.
Defined recursively:
f(30)=.05
f(x)=f(x-1)/(1 - f(x-1))
Single calculation:
1/(50-stack_number), if stack_number>=30
Though, this might be somewhat irrelevant on the abstract level since we are mostly interested in the average, which appears to be 40.
There are different ways it could be implemented, but I don't know that it really matters that much exactly how they've implemented it. It may, however, be important what the distribution winds up looking, and not just the average. So it would be good to know if its the uniform distribution I've proposed or something slightly more exotic.
I will also note that if I'm right about it being a uniform distribution from 30 to 49, the average would be 39.5, not 40.
I have attached a text file with the stack count for 238 procs. I used an SQL query to filter the stack count number for the combat log entry exactly one before the proc happened as I didn't feel to do all that myself, so there is a slight chance something's gone wrong. But it looks plausible.
Appending the data in your sheet gives the following:
To follow up on Fury being able to proc with Shadows: I haven't observed it happening still, meaning it's probably exclusive with Shadows. I'm inclined to say that it works by doing a two roll system, with the first roll deciding if it could proc Shadows, and if true then the second roll turns that into Fury instead (post 30 stacks).
Therefore, gaining Fury while on the 35th stack would have otherwise been the 36th stack had it not been Fury.
Also, I added sp00n's log to mine, and broke the specs apart in case we need to revisit whether or not each spec works differently with Fury too. Given that the mutilate data looks pretty similar to the sub data, I doubt it though. It is slowly smoothing out though.
EDIT: Up to...
Mean
39.513
Median
39
Mode
48
Occurances
716
The list is still going to be storing the stack of Shadows at the time of gaining Fury.
Doing some H Yorsahj tonight and it seems that fury of the destroyer is sticky to the target you are on when it procs but it is redirectable . I got a proc right as a globule was reaching the boss and swapped back to Yorsahj and he didnt gain any 5 stack combo points at all until I redirected. There was enough of a delay in my redirect to make me confident it wasnt just timing.
To me it seemed that the proc sticks to the same mob unless you generate a combo point on another target (either by using an ability or redirect). Auto-attacks only didn't seem to be enough.
AR is cast every ~70-90 seconds with the legendary under optimal conditions. So if you'll get more than half of the duration of AR, since you'll be cutting at least 40 seconds off the cooldown, it should net at least neutral. This ignores the fact that you get 20% additional attack speed as well. Since you will necessarily be getting a lot more than half of the duration (even if you activate it exactly as fury procs), that means it will absolutely be optimal to hit AR while fury is up. This is especially easy now that AR is off the GCD so that you can do it mid-proc without wasting anything.
That's far from a perfect analysis, but hopefully the idea that you're sacrificing 5-20% of the duration of AR in order to cut its cooldown in half should be obvious enough to show the correct answer.
12. AR and Vitality apply additively; haste regen is then applied to that sum multiplicatively. Blade Flurry and Overkill still need to be tested. Credit: Lightshadow
I'd noticed this and done some testing myself. Given that the adjustments are the same at all levels for Overkill, Adrenaline Rush, Vitality, and Haste rating, I had gone ahead and used my Lv.70 Night Elf Rogue to do the testing.
According to what I've determined, adjustments in Energy Regen are done in the following order:
Vitality [x1.25] & Adrenaline Rush [x2.00]
Haste rating [Varies]
Overkill [x1.30] & Blade Flurry [x0.85]
So, the formula could be considered as this:
( ( ( 10.00 x Vitality[YES: 1.25, NO: 1.00] x Adrenaline Rush[YES: 2.00, NO: 1.00] ) x Haste Rating[YES: 1.00+#.####, NO: 1.00] ) x Overkill[YES: 1.30, NO: 1.00] x Blade Flurry[YES: 0.85, NO: 1.00] )
On the off chance that you have Overkill still up when switching over to Combat spec... it does not matter which is calculated first between Overkill and Blade Flurry - the numbers come out the same. This is the same way with Adrenaline Rush and Vitality when doing their calculations.
The testing I did was with Haste rating included, as I felt it necessary to show that the order mentioned is how Energy Regen is calculated.
Of course, we all know that the absolute base Energy Regen is 10.00/sec. From here, we go into the calculations. It appears that the system rounds up to the nearest second decimal place (when displaying the Energy Regen value) when reaching a certain threshold for the calculation, as evidenced below. I'm still not certain on this threshold, however.
Of course, we all know that the absolute base Energy Regen is 10.00/sec. From here, we go into the calculations. It appears that the system rounds up to the nearest second decimal place (when displaying the Energy Regen value) when reaching a certain threshold for the calculation, as evidenced below. I'm still not certain on this threshold, however.
If you really want to go down into the threshold issue, you'll probably not come around to account for the "real" haste value.
According to Combat Ratings at level 85 (Cataclysm) this is 326/(10*82/(262-3*70)) ¹ at level 70.
Although there may be another round of rounding involved in the calculation of the haste value itself (and which is different to what the character sheet says).
If you really want to go down into the threshold issue, you'll probably not come around to account for the "real" haste value.
According to Combat Ratings at level 85 (Cataclysm) this is 326/(10*82/(262-3*70)) ¹ at level 70.
Although there may be another round of rounding involved in the calculation of the haste value itself (and which is different to what the character sheet says).
Interesting. I will say that (thankfully) the calculations end up still being extremely close.
I did test on my 85 (which is only Combat however) to make sure I was correct, and the numbers in-game came out virtually the same as the manually figured numbers. I say it with such words since there's still rounding involved.
Oh yes, and I do have a new question to ask. What are the proc rates of the dagger sets' Suffering, Nightmare, and Shadows of the Destroyer buffs?
Last edited by Jinchi_Ikizu : 02/04/12 at 12:47 PM.
4. The legendary proc is currently being modeled as purely a stacking Agility effect rather than a resetting stacking buff with free finishers mixed in.
Considering this means that I cannot simply 'ask shadowcraft', how much does the value of Coup de Grace increase with the legendary? And if so, is it worth moving a point from Lethality into CdG? (And is rupture still worth using in the rotation with this spec?)
4. The legendary proc is currently being modeled as purely a stacking Agility effect rather than a resetting stacking buff with free finishers mixed in.
Considering this means that I cannot simply 'ask shadowcraft', how much does the value of Coup de Grace increase with the legendary? And if so, is it worth moving a point from Lethality into CdG? (And is rupture still worth using in the rotation with this spec?)
I'm pretty sure 3/3 CdG is superior, it was already starting to out-value Lethality by a few digits for rupture-less play before legendary.
My question now is, how much value does Ruthlessness loses? Will 3/3, 3/3 be clearly superior now?
01/25/12, 5:41 AM
Similar Threads
