Leto

He appears to apply the effects of haste not to the weapon speeds, but to the number of attacks per second. It should have the same results, but I'd imagine his method is more robust for model development.

Rogue at heart.

 

Hartge

According to your spreadsheet running with 198 hit + talents, as long as you have at least a draenei, is sufficient for pve mutilate? Dropping over 100 hit from gems and enchants would free up more than enough space to stack more exp/ap/crit.

Main question: what is sufficient hit for pve mut? I thought it was 315 buffed at least for poison hit cap.

thanks

 

• Vulajin

This is not the thread for simplistic questions of that type. You are looking for Rogue: Simple Questions/Simple Answers.

 

NoPoint

edit: checking numbers, I'll update this tonight when I have more time to review it and check FA procs and so forth under different haste situations and crit figures.

 

• Vulajin

It would not have any effects on modeling. It's simply a case of applying multipliers at different times.

(edit)

Ald, I'm having a little trouble intuiting what you were trying to do with the "Envenom Categories" section where you have it broken down into (4,2), (3,2), (3,1), (2,2), (2,1), (1,1). It looks to me like the first index references the number of Mutilates "remaining" that can be performed within the cycle (assuming the cycle is based around the duration of the Rupture performed), and the second index references the number of Mutilates performed on an Envenom given that condition. It also looks like it is assumed (reasonably) that a 5-CP Rupture provides for 4 Mutilates, while a 4-CP Rupture provides for 3 Mutilates.

However, I can't see where the probabilities properly work out. Primarily, it seems that all of the formulas assume that a 2-Mut Rupture provides for 4 Mutilates, which isn't precisely accurate since it's possible to perform a 4-CP Rupture with 2 Mutilates. Also, the formula for (3,1) seems to calculate the conjunction of 2-Mut Rupture and 1-Mut initial Envenom, which shouldn't be accurate since a 2-Mut Rupture is assumed to provide a 5-CP Rupture, which would mean this formula is actually calculating (4,1). (3,1) seems like it would be calculated by the conjunction of 1-Mut Rupture and 1-Mut initial Envenom plus the conjunction of 2-Mut Rupture and 1-Mut initial Envenom and 1-Mut second Envenom.

Hopefully my interpretation isn't just way off the mark.

 

• Aldriana

Re: Expertise (and hit) calculations. To handle caps correctly, instead of adding rating, I add the equivalent amount of miss chance further down in the calculations, after caps have been evaluated. Hence, the values are changed in rows 42 through 46.

Yeah. I think the ideal would be to get some of your energy-queueing modeling into this sort of approach to try to get the best of both worlds; if anything, I think this sheet still underestimates the value of Expertise in terms of allowing more aggressive queueing.

And you're certainly correct that combat/HaT/whatever wouldn't easily fit in to the current calc sheet - there would need to be a new calc sheet for each fundamentally different spec. Fortunately, I wasn't going for generality here, just accuracy at the one that matters most to me.

The weapon speeds in line 13/14 are the base weapon speeds; as they're used to figure out instant attack damage and such things, they're just the base values. Haste is applied in line 67/68, where the number of attacks performed is counted.

The first number is the number of Mutilates since the last Envenom; the second is the number of Mutilates since then the last finisher of any sort. The first number provides an estimate for how long it's been since you wiped poison stacks, and thus allows us to estimate how many you're going to have for the envenom; the second is to figure out how big the Envenom is going to be.

So, for instance, if we're in the 1,1,2 cycle (1 mut, envenom, 1 mut, envenom, 2 muts, rupture) cycle component, the second envenom is a 1,1 envenom; the first one is either 2,1 or 3,1, depending on the nature of the rupture in the cycle before.

So, rows 115-120 are adding up the number of Mutilates of each type you expect to do during a cycle component; we then compute the cycle duration and, from there, the poison stacks. The assumption is made (for simplicity) that if you've Mutilated at least 3 times since the last Envenom, you're going to be at full stacks. This isn't of course 100% accurate, but it's pretty close - it would take a pretty significant statistical fluke for that not to be the case. So we break down only these three cases: 1,1; 2,1; and 2,2. This is what's done in lines 157 to 193.

At that point, we go into a second round of Envenom classification. Starting at line 196, we again see a,b notation classifying envenoms; this time, the first number is the number of Mutilates used to generate the envenom, and the second is the number of poison stacks that that envenom will have. Then, starting on line 207, we get the final round of classification; we convert to size + stacks instead of Muts + stacks. All 1-mut finishers become 4-pt finishers, and the 2-mut finishers split into 4 and 5 point based on the probabilities (though note that a 2-mut finisher will be a 5-pt finisher some 99% of the time).

Thus, after lines 207-216, we have the frequencies for every combination of stacks and CPs that is possible, which we then average out in lines 218-220. We also at that point define a no-stacks Envenom to be an Eviscerate, and work out the number of those (which works out to be "not very many").

 

NoPoint

You use base weapon speed for instant-attack damage? If that's the case, it should be at a static 1.7 for any daggers.

 

• Aldriana

Can we pretend, just for the moment, that I might possibly know what I'm talking about? If you'll note, the weapon parameters being passed in are weapon DPS and speed, hence weapon damage for instant attacks is

Weapon Speed * Weapon DPS + 1.7 * AP DPS

See that "weapon speed" term in there? So yes, even normalized attacks do depend on weapon speed. Contrary to popular believe, I'm not a *complete* moron.

 

MasterDinadan

Should the racial agility be 194 instead of 197? I only have 194 agility on my level 80 night elf rogue, so I can't imagine where the extra 3 is coming from.

 

• Aldriana

Yes, actually. Must've typed it in wrong. Also, the Rupture damage multiplier in row 280 should be 1.49 rather than 1.46 based on recent findings. I'll release an update in a few days - there's a few other changes that I'm thinking through at the moment and want to make before I make another release.

 

MasterDinadan

One other thing, I'm not sure on this as it may be due to rounding errors in game, but I'm not showing the same crit value as the spreadsheet is giving. I noticed it while I was not wearing any gear. The spreadsheet's crit matched mine, but only with the incorrect value for agility. After changing the racial agility to 194, my crit was lower than it was in game, so I wonder if you didn't overcompensate for 3 agility that we shouldn't have had to begin with. There's a "Agi-28" in the crit formula. I changed the 28 to a 25 and it was once again the right value.

 

• Vulajin

The correct formula for your crit rate is ((Agility / 83.33...) + (Crit Rating / (14 * 82/52 * 131/63)) - 0.295) / 100. The "Agi-28" part is supposed to account for that, but 28/83.33... is not equal to -0.295. 25/83.33... is also not accurate, but it is closer. You could fix the formula to be fully accurate by just removing the "-25" and instead subtracting 0.295 from the final result..

 

• Aldriana

I pulled that fomula out of an older spreadsheet I'd written - I'm not entirely sure where it came from in the first place - I think I probably wrote it sometime during the early beta, when the numbers were not precisely pinned down yet. So it wouldn't entirely surprise me if it's off by a bit.

That said: has it been confirmed that it's exactly .295? Or is that merely the best current estimate?

 

• Vulajin

I didn't publish my results but testing I did in beta at level 80 verified that, yes, -0.295 was still the intercept for crit rate. And yeah, we did linear regression at level 70 that showed it was exactly 40 agility per 1% crit with a -0.295% intercept.

 

• Aldriana

Okay. So for those of you updating your sheet at home, that's a third change you need to make. And while we're on the topic: I've found another bug. Turns out, the poison miss rate was being applied to Deadly Poison but not Instant Poison; the end effect of this is that hit below the poison hit cap is worth a lot more than initially projected. If you want to update your sheet accordingly, change B146 to =(B144+B143+B140)*0.01*(100-B44) and fill across.

So, all that said, I'd like some feedback on the Deadly Poison modeling. So, right now, the assumption in the sheet is that your only Deadly Poison downtime is immediately after an Envenom as it stacks back up; however, this assumption only works if you posit that every Envenom is performed immediately after a Deadly Poison tick, which is, of course, not practical. Setting aside for the moment the practical considerations of cycles which preclude waiting for Deadly Poison ticks in all circumstances, there's the simple fact human reactions cannot consistantly fire off Envenom at the exact instant of the Deadly Poison tick - there's going to be a delay in there. So, in order to be accurate, the sheet should model this. Thus, the question: on average, how long after the last poison tick does Envenom occur? And does the answer depend on gear or whatever?

Now, if one makes *no* effort to synchronize Envenoms versus Deadly Poison ticks, the answer is simply going to be 1.5 - DP ticks are every 3 seconds, and you're randomly somewhere between them. On the other end of the spectrum, if you assume that you can always go perfectly after a tick, the answer is limited only by human reaction time + latency, and is probably down around .2 seconds. So the question, then, is where in between these extremes does the answer lie? Does anyone have any sense of what's actually feasible in practice?

Additionally, note that in the case where you're timing off ticks, a dodge will cost you a second of Deadly Poison uptime; hence, the model will also be updated to add 1 second to this delay for missed/dodged envenoms. I suppose theoretically one could queue for another 2 seconds and go at the same time relative to the next tick, but my assumption would be that dodges are disruptive enough to one's cycle for other reasons that doing so is seldom practical. Does this seem reasonable to people?

 

Clemenza

On the deadly poison versus versus envenom timing: what I personally would like to see in the sheet is a variable solution. A field on the sheet where one could enter a value between 0 and 1.5. So one could see how much impact this part of the play style has on the dps.

If perfect envenom timing brings me 75 dps, I will do it. If it brings me 5 dps, I will probably focus more on other things like optimal timing in refreshing hunger for blood, optimizing rupture uptime versus the need to envenom eventually to keep slice and dice up, or simply on surviving.

The reasoning behind this is, that I am convinced, that one can not focus on every factor that has an impact on dps, since there are too many of them. (for example the idea of queuing energy to make optimal use of trinket procs, that you have commented on somewhere else.) So I like to know what brings me the most and to focus on that.

I haven't had a look at the sheet in detail. So I unfortunately do not have an idea, if this would be a lot of additional work.

 

• Neto-

I too agree with having the time between Envenoms and the DP tick as a variable.

It makes more sense, as it would be a pretty big thing to pay attention to, in addition to our HfB, Rupture and SnD timers *and* having to move out of fire; and even then, I suspect the large majority wouldn't bother doing that, but I also know that there are people such as myself who would do it anyway even if it is a 5 DPS gain, and if it does come out to be a low number, I could use my discretion to decide if it is worth doing or not for certain fights. It definitely sounds like something interesting.

 

• Vulajin

Timing off ticks seems unlikely additionally because there is simply no consistent method to do so using the current UI. A target debuff timer is only so useful since DP starts ticking the first time you apply it and ticks every 3 seconds no matter what weird intervals you apply further stacks or refresh the full stack. Thus, effectively the only way to currently do it is to watch for a tick on SCT or Parrot, which hardly strikes me as conducive to precise timing.

Of course, having said this, I'm sure someone will now go write an add-on to display precise timing for Deadly Poison ticks, rendering my points moot and enabling precise timing off ticks.

(edit) Minor error found, cell B164, which prorates the Envenom buff across white attacks performed during its duration, account for attacks not landed or failed procs, does not factor in the dodge rate. B165 has the same problem with respect to Mutilate, as do the corresponding cells B175, B176, B186, and B187.

The net result should be a very slight increase in the value of expertise.

(edit 2) More thoughts - would it be useful to track situations where you Envenom at 4 CP (i.e. a 1-Mut Envenom) with 5 DP stacks up? In this situation you'd still have a stack remaining afterwards, which would help contribute towards your stacks for the next Envenom. This would primarily matter to cycles involving multiple successive 1-Mut Envenoms.

Actually, on second thought, this would result in a circular calculation...might not be worth the required effort.

 

• Feist-Mok

An addon for timing poison ticks already exists.

TickToxin, a module of cutup, a fairly common timer for Slice and Dice, Hunger, rupture, etc. is a module for timing deadly poison on a target - the bar 'pulses' when the poison ticks.

 

• Aldriana

On timing off ticks: even if no so mod currently existed (and it sounds like one does), I can't imagine that it would be that hard to create an energywatch style bar that starts ticking whenever you apply Deadly Poison and has a nice timer bar to make it easy to time the envenom. This would actually be preferable to something that simply indicates when a tick as occurred, as you would be able to time the usage of the ability rather than relying on your reflexes.

On the minor bug: I agree that B164, B175, and B186 need to be updated, and will make this change; I don't think it's necessary for B165, B176, and B187, though, as even if the Mutilate is dodged you're going to reapply. Basically, this is the total number of procs gained during 1 Mutilate, which thus by definition must include the procs generated by the Mutilate.

On the DP lasting through a 4-pt envenom: I actually do this for the 2,2 table, as the only time you get a 2,2 Envenom is in the cycle elements 1,2,1 and 1,2,2; as such, you perform the rupture in the previous cycle (giving time to get to 5 stacks), followed by a 1-Mutilate (and hence 4 point) envenom, leaving you with one stack as you prepare to make the following Envenom. As such, it is assumed that you have one lingering stack in the table for that (which is why rows 189 through 193 offer no "0 stacks" option.

In the more general case: I was actually thinking about this some last night, and I believe with some cleverness a better estimate is possible, but there's still some details of it that I'm working through. At the very least I'd like to get it added to the DP DPS calculations, but I think it should be possible to get some improvement even for the Envenom stacks modeling by making suitable simplifying assumptions.

Edit: On second thought, B176 and B187 to need to be updated; however, making these changes actually causes the value of Expertise to go *down*, which is very strange and makes me suspect another bug; hence, I'm going to look at this further before I release again.

Edit #2: Found the problem. The damage contribution of Eviscerate was being significantly overestimated. Still playing with a few other tweaks, but I should have something up later this evening.

 

• Aldriana

Okay, new version of the sheet up in the first post. There's still some lingering issues in trying to model surviving poison stacks after an Envenom, so I haven't made those changes; however, there's a pretty decent stack of tweaks and improvements nevertheless:

• Corrected base agility to be accurate for an NE.
• Properly applied spell resist chance to Instant Poison
• Updated Rupture damage formula to reflect that it gets 9% from Find Weakness rather than 6%.
• Updated Eviscerate Damage to use .07 AP per CP rather than .05, to match recent testing.
• Fixed bug with Eviscerate calculations that was causing it to roughly 20 times too much damage.
• Adjusted handling of hit calculations to give correct EP weights for players with very low hit rating.
• Added settings page to allow user to select whether there is a Draenei in their group or not.
• Added settings for timing Envenoms against DP ticks.
• Correctly apply dodge rate to poison stacking calculations.
• Slightly adjusted attack counts for 2,1- and 2,2-Envenom poison stacking calculations to better reflect reality.
• Updated Agi to Crit conversion offset to match in-game values.

There might be 1 or 2 other minor ones, but those are the immediate ones that come to mind. From a practical perspective: the value of expertise rating nosed up a bit more, and the value of hit below the poison cap increased significantly. The end effect on gear is that Strong Handed Ring passes Ring of Invincibility for the second ring slot, which has been reflected in the gear on page 1. Note, however, that if one does not consistently have a Draenei in one's group, it may be worth replacing 1-2 Deadly Monarch Topazes with Glinting so as to make the spell hit cap.

 

• Aldriana

In one of the various discussions I've had with people about optimizing rogue DPS and analyzing WWS parses, it came to my attention that I missed a buff in this sheet; Imp Scorch (or equivalent) is not factored in. If you want to correct your sheet at home, just replace the "-3" in line 54 with a "+7" instead. This seems to have no effects of significance on any stat weightings - the optimal gear cited in the most recent version appears to still be optimal.

I'll have an updated version with the fix in out at some point, but probably not until after I've had a chance to tune up the Deadly Poison model in the sheet, which may take a couple of days and/or weeks.

 

Loot

Ald, have you made some prediction how much dps goes down in the optimal gear with the new mutilate fix?

 

Genre

By applying the changes to line 143 and 144 discussed in another thread (replacing the 2 for a 1), we get a drop of 2.02% in DPS. From 5122.99 to 5019.13; please correct me if I am wrong.

 

Almehym

To be fair, that would only be if all you did was fix the bug and not swap your poisons/weapons (depending on which weapons you're using). Because this bug would no longer apply, we would be switching back to putting our fastest weapon (with IP) on the main hand in order to proc poisons off of finishers. There will still be a dps loss, obviously, but it isn't as cut and dry as just removing the bug, as we'd be adjusting our playstyles to fit.