The answer is in the first post of this thread in the spreadsheets. Aswell there was a lengthy discussion in a seperate threat on this very same board.
[Raid] Executioner vs. Mongoose, preliminary numbers

Motion to add this to the first post:

Regarding Mongoose and Executioner Enchants:

Both Mongoose and Executioner enchants are very powerful for rogues. In general they are very close to one another in value, and the only thing you can really do wrong is to enchant both MH and OH with Executioner - as the added benefit from a 2nd Executioner enchant is much less, due to stacking issues. However, as a guideline, according to the best information that we have available, here are some recommendations. These recommendations are for Combat Spec'd rogues.

Karazhan/T4 geared rogue: Mongoose on both hands
T5 gear: Executioner or Mongoose MH (very close in value), Mongoose on offhand
T6 gear: Executioner MH, Mongoose OH

I suggest a note being included about the dodge from Mongoose -- sure the chances of it actually being useful (enchant being proc'd and you need to dodge something) are fairly low, but even at the very high ArP range, isn't the difference between Mongoose and Executioner pretty small? If your guild competes for WWS clears, that's one thing, but most of us just want to get through stuff that's new to us, and that sometimes involves things like Evasion-tanking Kaz'rogal... >.>

So I've been thinking about my Najentus Rotation and I'm wondering if it would be worth it to shiv during his shield. If only to build up combo points for a SND or a rupture.

I was thinking same for last couple of weeks. Basicly cycles like 3s/5r work really well with only handful amount of interupts. There isn't many fights in MH/BT (at least those I encountered already, like 5/5 MH and 4/9 BT) that have alot of interupts, maybe beside randomness on Azgalor and Archmonde, and Akama's channelers short HP. However, on those fights sometimes I find 3/5 way to long to keep 100% of its power, I can imagine that with 5/5 or so, it's gettin even worse.

I haven't tested 1/3 or other cycle combinations yet, but it would be nice to run some tests on shorter time periods and see what works better and whats not. I really hope for theorycraft EJ masterminds input on that

I've read most of the comments on Shadowstep (including the -12% estimate). I'm no math wizard, so I tend to put quite some faith in the sheet (using 2.3.2.9 atm). The weird thing is that if I put my gear in (Armory link) which is mostly T4, Heartless and Searing Sunblade OH, the sheet shows my dps with combat swords is roughly equal (1% difference, depending on OH chosen) to Shadowstep/Assassination.

So the point is: I'm really surprised by the minimal difference between combat and sub in the sheet. I know sub has been buffed a lot lately, but I hadn't expected it to be this viable. However, I've seen the comments in this thread (which come down to: ShS sucks, except for highly mobile fights) so I'm doubting the sheet now. Either the sheet is a bit too enthusiastic about ShS, or perhaps it's just a fluke having to do with my gearlevel?

PS: you can diss ShS all you want, but seeing a fellow rogue ShS behind Al'ar in phase 1: priceless

I have a question regarding rupture and adjusting cycles on the fly. I've noticed in a few places where people have mentioned that 3r and 5r receive the same AP benefit, and although I understand the AP formula behind rupture, I may be mixed up on the underlying mechanics.

Currently, if I start getting into trouble with SnD up time or energy (currently run 1s/5r) due to bad luck with CP procs, I'll blow my rupture on the target at the point where I know I can't get to 5 points but still have time to get 1 point for SnD. Generally I have 4 points on target when this happens. Is there any advantage w/ regard to rupture to placing it on the target with 3 points instead of 4? Aside from the obvious point that I'd have a little more time for energy regen, that is.

I guess I'm just not sure exactly how people are saying that a 3r receives the same bonus as a 5r, and how it applies to broken cycles.

What it means is that 3r, 4r and 5r do the same damage per tick, they just don't last as long. 1r and 2r not only last for a short time, but they also do less damage per tick. However, if you find yourself cutting a cycle short, you should still use the largest possible rupture that still leaves you enough time to get s'n'd running again.

If I equip the S2 OH, and replace my 4agi/4hit gems with 8hit gems (4x) combat swords does 1.18% more dps than Assassination/Shadowstep..... Without replacing the gems it's 1.10%. Raidbuffed percentages btw, 1311dps vs 1297dps.

Not sure why my crit is high, probably because I value agi>hit. The reason for that is that with my high agi, i'm next to unhittable when popping evasion. And I value survivability above doing 0.1% more dps

Are you sure you're modelling Hemo accurately? If you have it ticked in both parts of the spreadsheet, you'll be overstating your DPS considerably.

I'm currently specced 11/9/41 for raiding, MHing Rod of the Sunking and OHing Tracker's Blade, and I'm usually pretty even, personal DPS wise, with our well geared Mutilate rogue, (4/5 T5, Bad Mojo, Tracker's Blade, AToL). This isn't taking into account in any way the hemo debuff however. Comparing to Combat Swords, my overall DPS (including the hemo debuff) seems to be about equal to where I was specced combat Swords (Blade of Infamy, S2 OH). The Spreadsheet also has me doing about 30dps more overall (counting debuff) than with the Combat Swords spec.

Of course, it requires Rod of the Sun King, which is a really awesome weapon, too bad its not Blade of the Sun King.

Considering it sounds like you are raiding T6 level content, what levels of DPS are you putting out? At this point no one has been able to show with any reasonable WWS that they can sustain close to combat swords level DPS with Mutilate or Shadowstep. Certain fights you might, but until we see some 2k+ Teron Parses with Shadowstep or Mutilate, you are going to have a hard time finding believers here.

Alright I have been reading the forums here for a while and have a situation now that the only thing I can find about it is to use a 1s/5r and cut the extra snd time.

The only issue I am having is that there are a lot of times that I cut a good amount of snd time and get back to a 5r well before I can apply it. This leaves me SSing until I can apply rupture and losing the CPs. If I know I am going to be running far enough ahead of the 5r time, have the CPs/Energy, and won't risk dropping SnD... Should I do a 1-3 CP finisher before building my 5r.

Something like...
1s/5r - 1s/5r (realize running ahead) - 1s/2e/5r - back to 1s/5r or continue with a short CP finisher if cycle allows.

If the 1/x/5 is worth doing what finisher should the x be? I am a combat sword spec, and don't have improved evis.

Or should I run with a shorter rupture like 1s/4r and just cut the SnD?

Well that's your choice. I just want you to understand that the minimal difference between specs that you're observing is simply due to not being properly geared/gemmed for Combat.

A cada su propio.

A couple things...
The spreadsheet has not double-dipped on Hemo DPS in a number of versions (fixed in 2.3.2.5). If Hemo Debuff Estimate is enabled, the Hemo debuff in the buffs section does nothing anymore, so having both on is ok. You can check this by selecting a Hemo build with Hemo Debuff Estimate on. Add Hemo debuff under Buffs and you will see the DPS doesn't change anymore.

The Hemo debuff amounts to a sizable amount of damage now that its up to 42 per tick or 420 per Hemo application which can be multiplied on Crits and other skill/talents/buffs which give damage multipliers while, of course, being mitigated by Armor reduction. It's not too unreasonable to estimate that the raid could be gaining about 120 dps for the debuff alone over ones own personal damage. Of course what that means for the WWS parse if the spreadsheet is showing a 30 DPS benefit is that one's personal damage is now somewhere about 90 DPS behind Combat Swords.

So just as DPS Warriors sometimes wear Solarian's Sapphire to boost raid dps (at the sacrifice of personal DPS), some rogues will be Hemo to boost raid DPS. I personally see little wrong with that. I should add that at the point where you are only about 30 DPS with the debuff ahead of Combat Swords, a 2nd Hemo rogue will probably negate just about all of the advantage in occasional debuff overlap. But if you are the only Hemo Rogue, I'd guess it's still a raid advantage.

1s/5r is going to be pretty close to the edge of sustainability. Your armory is showing 329 Hit Rating unbuffed, which means you are running at 349 buffed, which is very high, and will yield a proportionately high number of CP Procs. In addition you are using 4/5 T4 which is going to give you an additional Combo Point some of the time. The result of this is that while a normal rogue using 2/4 T4 is going to get 1 CP 60% of the time, you are going to 1 or 2 at times, and will have atleast 1 more often. This also means you are going to have occational 2 point S&D's without ever SS's after your finisher.

My recommendation without going into too much math, would be to first off migrate to some higher DPS items (T5 gloves and shoulders), and just plan on cutting and having slack time in your cycle. The DPS you will get from any 2 point finisher is going to be inferior to just SS'ing more. If you have 5 CP's, use Rupture, if Rupture is still up, wait until you have 70 energy to use it, then fire off S&D again if you get 1-2 CP's, if not SS 1 time and fire off S&D. If you get in a situation where you can't ruture due to the last one being up with Mongoose/Trinket, then wait for it to fade, refresh it, and the S&D or SS then S&D.

Bottomline, 1s/5r is about as good as you are going to get, if you have too much slack in that and are getting cycles shorter then rupture, and are in danger of energy capping, more Sinister Strikes will be better then anything less then a 4-5 point Eviserate (your SS is going to hit harder and won't disrupt your cycle).

I really don't think this is a gemming issue, since gems make up the equivilent of roughly an extra pice of gear and thus are not going to skew your results dramatically.

My guess is that you simply have the option "include Hemo debuff in DPS estimate" turned on so you're actually seeing your raidwide total dps rather than just your personal dps. My own 11/9/41 Shadowstep build (which is 1.9% more dps than 20/0/41 in my current gear) comes out 11% behind Combat Swords for personal dps. When I check the "include Hemo debuff in DPS estimate" box it comes out 0.5% behind Combat Swords in blues/T4 and 0.8% in my current T4/T5 gear.

Basically what that tells us is that by speccing any variant of Shadowstep we're bringing less dps to the raid as a whole than if we simply specced Combat Swords. Not by a lot, not a lot at all in fact, but still less.

Which is still based on the assumption that every charge is used every time it is applied, which simply isn't the case (example: you shadowstep back in, Hemo, and then Hemo again 1 second later, before any of the other melee gets back in, or at the very least, before they can use all the charges). The bottom line is that the very best case scenario is still worse then the worse case scenario for combat (since you normally get a disproportionately high value from cooldowns unless they come up again right as the boss dies).

Ok, so point for discussion. I decided this is probably the better forum to present this given that it has implications to mechanics as a whole. I apologize in advance for the in-depth mathematics.

This weekend I started to go through some mathematical theorycraft for some of the game mechanics starting with PPM. My goal was to arrive at a formula that properly converted a proc per unit time into % uptime.

So I began with the concept of a PPM proc that on average procs once every 50 seconds but could conceivably proc back-to-back (i.e. no cooldown) with a duration of 15 seconds. Some may recognize this as 1.2 ppm on a single weapon auto-attacking with either Mongoose or Executioner if one believes the proc rate to be 1.2.
So, to create a formula we have:
Pf = Proc Frequency (for this example 50 seconds)
Pd = Proc Duration
Up = % Uptime

So, looking at at if on average it seems a 15 second proc every 50 seconds gives 30% uptime, but this doesn't account for potential overlap. So we really know 30% is a cap.
Ov = % Overlap

So, we have:
Up = Pd/Pf*(1-Ov)


Now, what should the % overlap be? Clearly, % Overlap should equal % Uptime.
So, we now have:
Up = Pd/Pf*(1-Up)
Solving for Up we get:
Up = Pd/(Pf+Pd)

So, for the example above we get 15/(50+15) = 23.1% uptime which is pretty close to the 25% estimates I've seen thrown out.

So, let's check this calculation. First, let's assume Proc frequency approaches 0. In that case we get Pd/Pd = 1 or 100% uptime. That works!
Let's use another example. Let's assume a proc frequency of 10 seconds for a proc with duration of 10 seconds. It should be fairly obvious that because of proc overlaps you are going to have a 50% uptime. Plugging it into the equation we get... 10/(10+10) = 0.5 or 50%

So, I believe I have derived a simple formula for % uptime conversion...feel free to check the math, but it appears accurate.

Now, step 2, the dreaded Deadly Poison proc calculation. This becomes difficult because each proc adds another dose, still this formula could be used to derive % uptime (i.e % one has any combo points)
Pf can be determined by 1/(Att/sec*ProcChance)...
Aps = Attacks per second
Pr = Proc Rate
Using the formula, Up = 12/(1/(Aps*Pr)+12)
So, the downtime = 1- that value.

We now know the percent downtime or in other words the percent of time that we have no DP up. Of course this doesn't take into account skills that use up one's Poison, but is the first step in the rest of the equation.

I'll stop here for discussion, still working through and verifying my other calcs for uptime at various doses...

Thoughts? Comments?

Thanks Hanos and Hannigaholic. I can't check it now, but this must be the case.

@hanos: I'll still gem the same. The difference from the +8hit gems is so minimal (and only while fully raidbuffed) that it's not worth the loss in survivability imo. And that's all that it is, my opinion

DMM:

For Deadly Poison, I would recommend looking at post 883; it is, to my knowledge, the best known model of Deadly Poison uptime (and the one used in v0.9.3 and higher of the Rogue Gear sheet)

For Mongoose:

Well, first, let me derive the canonical formula for Mongoose Uptime.
If the proc occurs once every Pf seconds on average, that means that the chance it procs during any particular second is 1/Pf, which I will denote for the moment as p. Let q = 1-p.

Now, lets consider the actual uptime granted by a mongoose proc that starts at T=0. We will condition on the time of the next proc. For simplicity I'm going to use the actual duration of 15 seconds, but the same technique is easily extended to general times.

If the following proc occurs at T=15 or later, the initial proc gives a full 15 seconds of uptime. However, if the following proc occurs at time T=t for t<15, the initial proc only gives t seconds of uptime; uptime after that point is counted towards the following proc.

Now, the probability that the next proc occurs at time T=t is the probability that it procs at that time (which we've defined to be p) times the probability that none of the t-1 preceding seconds contains a proc, which is easily shown to be q^{t-1}. Hence, the expected uptime gained by our proc of Mongoose is then:

p + 2pq + 3pq^2 + ... + 14pq^13+15pq^14(1 + q + q^2 + ...)
Now, 1 + q + q^2 + q^3 + ... is a well-known infinite sum; it's value is easily shown to be 1/(1-q). But 1-q = p, so our formula simplifies to

p + 2pq + 3pq^2 + ... + 14pq^13+15q^14

Let us then write 15q^14 = 14q^14 + q^14; this then combines with the preceding term as follows:
14pq^13 + 14q^14 + q^14

[top]14q^13(p+q)+q^14

14q^13 + q^14

as p+q = 1. But then applying the same trick repeatedly, we find that the entire sum telescopes down to
1+q+q^2+...+q^14

which is easily shown to be

(1-q^15)/(1-q) = (1-q^15)/p

Now, since procs by defenition occur once every 1/p seconds on average, the average *uptime* is then (1-q^15)/p seconds out of every 1/p; hence, average uptime is simply 1-q^15.

Substituting back in your variables, this becomes 1-(1-1/Pf)^Pd. This would be the estimate that is used in the Rogue Gear sheet.

Now, lets see how this compares to your formula;

Again, let 1/Pf = p, and Pd = n (for notational simplicity). hence we wish to evaluate

1-(1-p)^n

Expanding by the binomial theorem, we have (1-p)^n = 1 - np + n(n-1)/2 * p^2 + ....
If p is small, we can cross off the higher order terms, as they will be small in comparison to the first few. hence, lets approximate this as
1-(1 - np + n(n-1)/2* p^2) = np(1 - (n-1)p/2)
Substituting in your variables, we have Pd/Pf * (1 - (Pd-1)/2Pf)

Now, if (Pd-1)/2Pf is small, this is roughly equal to

Pd/Pf * (1/(1 + (Pd-1)/2Pf)) = Pd/(Pf + (Pd-1)/2)

Assuming Pd >> 1 (i.e.., Pd = 15), this becomes

Pd/(Pf+.5Pd)

Which is... certainly similar to what you came up with, but is definitely a bit different.

So, how accurate are these various models? Well, for Pd = 15 and Pf = 50, your model gives 23.08%; my approximate model gives 26.09%; and the full exact model I derived gives 26.14%.

So, if one is interested in maximum accuracy, I would go with the formula 1-(1-1/Pf)^Pd. However, if you want a simpler formula, I would use Pd/(Pf+.5Pd) rather than Pd/(Pf+Pd), as it is equally simple and much more accurate.

Edit for additional details
The problem with the approximation you performed is that while the probability that a proc occurs during an existing uptime period can reasonably be approximated by (1-Up); however, a proc occurring during an existing proc does not totally waste the uptime of that proc; on average, in only wastes half of it. Hence, the way you've defined Up and Ov, you actually have Ov = Up/2. Hence, the formula to be solving is Up = Pd/Pf * (1-Up/2), which simplifies to the Up = Pd/(Pf + .5Pd) I derive above. The difference between the approximation and the full solution is that the actual uptime lost isn't exactly half on average; given that overlap occurs, the time at which the 2nd proc occurs is not uniformly distributed over the input times, but, instead, a probability distribution determined by the exact values of Pf and Pd.

Actually, you miss the point of (1-Up). 1-Up is the average % portion of each proc which is lost. Or, on average, one loses of ones duration a percentage of it relative to uptime and only gains relative to downtime. Pd/(Pf+0.5Pd) is clearly wrong because as one's proc frequency approaches 0 you'd have a 200% uptime.

Also, the problem with your model of a Mongoose proc beginning at T=0 is the possibility that it over laps with a proc starting at T=-15 to T=0. As this is a continually modeled system, one must address that possibility as well.