Elitist Jerks Hunter DPS Analyzer (MoP edition)

08/22/12, 1:16 AM   #16
Whitefyst
Great Tiger

Orc Hunter

Draenor
With the shot nunmbers that FD provides, I just completed an updated analysis in T14 heroic gear for the tipping point where casting AI as the focus dump is better than AS. I did the analysis with the following formula:

AI + WQ = (x * AS) + (1.25 * auto) + ( (1.25 + x) * WQ)

where:
AI is AI average damage with including PS.
WQ is average WQ damage per shot taken that can proc it.
AS is AS average damage.
auto is average autoshot damage.
x is the AI cast time and fraction of ASs performed over that cast time

On the left side is 1 AI cast and the average WQ damage.
On the right side is the damage from the fractional number of ASs that can be performed during the AI cast time plus the damage from 1.25 autoshots being suppressed during the AI cast plus the average WQ damage from the ASs and autoshot.

This resulted in a 1.424s AI cast to reach the tipping point when assuming 1.25 autoshots are suppressed per AI cast.

If going with the about the best case scenario of only exactly 1 auto being suppressed during an AI cast, the result is a 1.597s AI cast time.

The haste requirements to make it a DPS gain to cast AI as the focus dump instead of AS for these two cases is shown in the following table:

 Normal T14 4P Normal* T14 4P* AI Cast Time 1.424 1.424 1.597 1.597 Bloodlust 23.84% 13.93% 10.42% 1.59% Rapid Fire 14.99% 5.79% 2.54% 0%
* Assuming exactly 1 auto suppressed per AI cast as opposed to 1.25 autos.

The following conclusions can be drawn from this:

1) MMs definitely want to be using AS as their focus dump when unhasted.
2) The AI cast time at which AI becomes better to use as the focus dump is somewhere between 1.424s and 1.597s. I suggest using either 1.6s or 1.5s as the new default value for maximum speed to cast AI setting.
3) Without the T14 4P haste bonus, it may not be beneficial to cast AI during BLs. It requires at least 10.42% (when assuming exactly 1 autoshot is suppressed during the AI cast) and possibly up to 23.84% haste from gear (when assumming 1.25 autoshots are suppressed per AI.
4) With the T14 4P, it is advised to cast AI during BL with the recommended haste amounts.

Is anyone else seeing a similar tipping point?

What is a good estimate to use for the number of autoshots suppressed during an AI?

EDIT:

I did some tests in FD with a 1 min test case during the Standard phase with BL active through most of it at the 5442 haste rating (12.80% haste) from gear that I had in my quick optimal case. AI cast time in this condition is 1.438s. I lowered the maximum cast time for AI to 1.4s, and DPS increased by over 500 DPS. Hence, in FD, it appears that a 1.4s cutoff may be better and that AI should only be cast during RFs or BL plus some haste proc.

During the CA phase with 100% AI crit chance, then it is always better to use AI as the focus dump, even with no haste on gear and no dynamic haste effects.

Also latency affects this number some since it affects the number of ASs that can be performed during an AI cast.

Last edited by Whitefyst : 08/22/12 at 9:19 AM.

08/23/12, 6:59 AM   #17
Nerec
Von Kaiser

Orc Hunter

Mannoroth (EU)
 Originally Posted by Whitefyst What is a good estimate to use for the number of autoshots suppressed during an AI?
Don't pin my down on that, but I tested the behavior of slam on an warrior alt on live servers few weeks ago and get a positive reponse from a gildmate who didn't believed it. Slam dont interrupt the swingtimer, but rather just suspend your current swing for the cast and resume to the exact % of the progressed swing. Can't check it right know, but if Aimed use the excact same behavior, to rather suspend the swing/autoshot than interrupt it, with a normalized weapon speed (2.9/3.0) you will be able to math out something like:
• haste * weaponspeed = time between autoshots
• casttime(Aimed) * time between autoshot = lost autoshots
• lost autoshot * ~dmg of autoshot = lost dmg due to casted aimed
• etc..

I know this is a big "IF", but can't test it right know, sorry!

Last edited by Nerec : 08/24/12 at 2:04 AM.

 08/23/12, 12:00 PM #18 Rivkah Great Tiger   Zeherah Dwarf Hunter   Stormrage It looks like the formula I'm currently using to simulate autoshot suppression (which I assume is based on aimed shot mechanics we discussed during Cata) is: (Average Aimed Shot Shot Cast Time / Weapon Speed + .5) * Aimed Shot Count = Autoshots lost I'm not sure if that's really the best formula, so I'm open to suggestions.
08/23/12, 12:31 PM   #19
Whitefyst
Great Tiger

Orc Hunter

Draenor
 Originally Posted by Nerec Don't pin my down on that, but I tested the behavior of slam on an warrior alt on live servers few weeks ago and get a positive reponse from a gildmate who didn't believed it. Slam dont interrupt the swingtimer, but rather just suspend your current swing for the cast and resume to the exact % of the progressed swing. Can't check it right know, but if Aimed use the excact same behavior, to rather suspend the swing/autoshot than interrupt it, with a normalized weapon speed (2.9/3.0) you will be able to math out something like: [*] haste * weaponspeed = time between autoshots[*] casttime(Aimed) * time between autoshot = lost autoshots[*] lost autoshot * ~dmg of autoshot = lost dmg due to casted aimed[*] etc.. I know this is a big "IF", but can't test it right know, sorry!
Thanks. I have been doing some analysis with assuming 1 autoshot lost but with using 1.25 to be conservative, especially since 1/4 of an autoshot and its expected WQ damage does not change the analysis too much. I have been suing both values to provide kind of a range.

 08/26/12, 6:48 PM #20 Rivkah Great Tiger   Zeherah Dwarf Hunter   Stormrage Just an FYI for folks who use both my site and simulationcraft- you can now get a simc profile with most of your setting data (excluding action lists and buffs) from your dps output. If you enable the debug data option the simc profile is displayed below your simulated shot breakdown. This should hopefully make it easier to transfer settings into simc (I may consider adding the ability to parse a simc profile in a future release, but that'd be a lot harder to implement).
 08/28/12, 12:21 AM #21 Rivkah Great Tiger   Zeherah Dwarf Hunter   Stormrage I added snake trap support to my site based on my initial testing on the formula for the snake attacks which indicates that it's likely a dps increase to include it in the rotation (assuming you don't spend too long fumbling with the trap). I expect they'll probably nerf it soon, but in the meantime people should be aware that snake trap now scales with haste, crit and RAP and does a noticable amount of damage (especially for SV which gets 2 extra snakes of each type). Snake trap does not benefit from BM or SV mastery. I posted the formulas I'm currently using for the snakes on the FAQ thread- note that these are at level 90 so I'm not sure if they're as effective at 85.
 08/28/12, 12:27 PM #22 Nooska King Hippo     Nooska Blood Elf Hunter   Argent Dawn (EU) Could I ask for a setting for BM to not overlap FF and RF? Also can I ask if you did something to change TotH lately - its now showing as a rather massive 3.3% increase to take DB over TotH, and a 3% increase to take Fervor - while TotH has been simming behind fervor (I know DB has changed due to other stuff), it hasn't been this far behind in previous calcs I've done, so I'm wondering if its broken/bugged currently.
08/28/12, 8:25 PM   #23
Rivkah
Great Tiger

Dwarf Hunter

Stormrage
 Originally Posted by Nooska Could I ask for a setting for BM to not overlap FF and RF? Also can I ask if you did something to change TotH lately - its now showing as a rather massive 3.3% increase to take DB over TotH, and a 3% increase to take Fervor - while TotH has been simming behind fervor (I know DB has changed due to other stuff), it hasn't been this far behind in previous calcs I've done, so I'm wondering if its broken/bugged currently.
I don't think there's been any changes to fervor for awhile. ToTH mostly has been changing as a result of arcane shot nerfs- I don't think there's been anything recently. When I was simming before, DB and Fervor were both pretty high above ToTH so I'm not sure if maybe it's just related to the config you have (or perhaps indirectly affected by the frenzy change?). DB went up significantly when I added in support for haste scaling on the DB pet.

With regards to FF and RF, it occurs to me that I probably should just setup FF to be limited by the "minimum focus shot speed to activate haste cooldowns" setting. Right now if you use that, RF won't activate during FF (assuming the speed is set appropriately), but I don't think FF is configured to limit itself in the other direction. Would that address the issue?

 08/29/12, 4:12 AM #24 Nooska King Hippo     Nooska Blood Elf Hunter   Argent Dawn (EU) It would, Its mainly a problem of prioritizing RF anfd FF, and since they only affect CoS (and Auto) they are right above and below eachother, with no difference regardless of which is first (which doesn't make a whole lot of sense if you want to spread them out)
 09/01/12, 6:45 PM #25 Rivkah Great Tiger   Zeherah Dwarf Hunter   Stormrage I've made a couple tweaks that folks might want to know about: - At Nooska's request, focus fire is now restricted by the haste cooldown restriction (i.e. it won't cast it if your current focus shot cast speed is below the configured amount) - I've modified my support of autoshots so they are directly implemented in the simulation (although I didn't include them in the simulation report). Previously they were estimated based on haste data. This means autoshot counts should more accurately reflect the stacking of haste buffs at various points in the fight and I was able to more accurately reflect the autoshot suppression from aimed shot (which resets the timer and suppresses until the shot goes off) and barrage (which does not reset the timer but prevents autoshot from going off until the channel is over) - Because I now support autoshots directly in the simulation, I was able to change my support for go for the throat. Previously go for the throat procs were estimated before the simulation and applied as passive pet regen. Now they are tracked directly in the shot simulation. This will give more accurate representation of the peaks and valleys of pet regen during the fight - While fixing other things, I noticed that barrage and powershot weren't being scaled with haste in my simulation. I've fixed this. It doesn't seem to have made either perform better than glaive toss, but the gap is lower - Due to the autoshot changes I was also able to more accurately reflect wild quiver proc rates. I've also corrected which shots are actually proccing wild quiver and applied the 1/6 chance to proc for each tick of barrage and the 2 separate tick chances (one for each hit) for glaive toss From what I could tell, most of these changes didn't have a significant effect on dps, but I wanted everyone to be aware of the status of how things are implemented. Next on the list is to add support for stormlash. I'll also probably modify pet melee attacks to be supported in the simulation directly instead of estimated at some point, but that's a lower priority.
 09/02/12, 4:08 AM #26 Nooska King Hippo     Nooska Blood Elf Hunter   Argent Dawn (EU) I have another request now that I've played around a bit more. I notice that for BM SrS actually falls off quite a bit. Saving focus for it is a dps loss (simulated of course), but in a real world situation, I would always advocate not letting it fall off. Would it be possible to have a checkmark setting that forces a CoS (after KC) if SrS has less than 7 seconds to live (like the ISS setting really, but not so complicated in choice). Also, Sanity check, could you see if you have affected agility's coding in any way? Suddenly crit is consistently showing as being worth more than agi, per point, and changing out agi for crit is showing a dps increase when looking at sane amounts (like hybrid gemming over straight agi). Edit; actually crits value now seems to fluctuate as much as Agi's - are the autoshot crits averaged or calculated per, per crit chance? If averaged, then I guess we will see the same fluctuation with crit as with haster - where crit is worth significatly more when it adds 1 more crit AS, then worth almost nothing untill the next crit percent with another AS crit. Last edited by Nooska : 09/02/12 at 4:17 AM.
09/02/12, 10:46 AM   #27
Whitefyst
Great Tiger

Orc Hunter

Draenor
 Originally Posted by Nooska Also, Sanity check, could you see if you have affected agility's coding in any way? Suddenly crit is consistently showing as being worth more than agi, per point, and changing out agi for crit is showing a dps increase when looking at sane amounts (like hybrid gemming over straight agi). Edit; actually crits value now seems to fluctuate as much as Agi's - are the autoshot crits averaged or calculated per, per crit chance? If averaged, then I guess we will see the same fluctuation with crit as with haster - where crit is worth significatly more when it adds 1 more crit AS, then worth almost nothing untill the next crit percent with another AS crit.
I will first preface this reply with the caveat that I know very little about BM. I have not tested yet on live and have done very little theorycrafting concerning it.

First, at least for the BM setting from Zeherah, I see agi well abive crit at 5.8 to 1.9.

However, from a theoretical point of view, I can see how crit is very important to BMs and changes to it can cause crit to be greatly valued in the simulation. Besides each crit increasing hunter and pet damage, the biggest impact for crit that I can see for BM is the interaction of GftT with WH affecting your pet's basic attacks.

In FD, the pet's basic attack was a little over 10K DPS out of about 89K total DPS. Hence, the pet's basic attack accounted for about 11.2% of the overall DPS. That is a large amount for an attack whose damage level is do finicky currently and which is not only affected by crits, but also WH and attack rate.

In MoP, pets are generally very focus starved. This holds true for BM pets too, although the situation is not as bad with GftT and FF. Despite those two talents helping pet focus and basic attacks, the BM pet still had its basic attack delayed due to being focus starved on 24 of 130 attacks or about 18% of the time. Compare this to my MM pet whose basic attacks are focus starved 70% of the time with the only times it is not focus starved is during BL and RFs. If I would have stacked BL over RF on the run I had, then the MM pet would have been focused starved on even more of its basic attacks.

The items that increase the rate at which the BM pet is not focus starved over the MM pet are:
1) GftT procs providing 15 focus.
2) FF providing 30 focus
3) FF increasing pet focus regen

I will not fully discuss the FF benefits here since they are not affected by crit and is off the topic. But I will mention a few items. First is its synergy with GftT procs. FF occurred 9 times for 270 focus directly returned powering about 10.8 pet basic attacks but also resulting in additional focus from the increased focus regen rate. When under no other dynamic haste effects, FF increases pet regen in this case by 1.577 FPS, which results in an estimate of 284 pet focus regened from FF, enough focus for 11.4 pet basic attacks. Each of the 9 FF by itself provides 30 instant focus, resulting in the pet not being focus starved. Depending on when the FF occurs, depends on how it affects the pet basic attack rate. 2 of the FFs occured when the pet already had enough focus to use its basic attack so did not immediately affect its basic attack rate. For the other 7 FF, how long after the previosu basic attack that the FF occurred affected how well it saved a basic attack from being focus starved.

Over the run, GftT occurred 52 times restoring 780 focus, which is enough for 31.2 pet basic attacks. In the FD sim, the time between GftT procs does vary considerably with an average of about 8.6s between them with a range from 5.7s to 13s. Obviously, the quicker rates tended to be at times when under dynamic haste effects with increased autoshot frequency. This is where FDs new autoshot modeling really helps to be simulate the situation. The slower rates are when no dynamic haste effects are in place. Considering that there is kind of a pattern between GftT procs when under no dynamic haste effects, it appears that FD is modeling GftT procs in some predictable manner where a proc occurs after so many autoshots based on crit rate.

Another benefit of GftT (and FF) beyond increased pet basic attack rate is the number of basic attacks when WH is active. For the BM sim in FD, 32 of the 130 basic attacks (24.6%) were under the effects of WH hunt, increasing the damage of the pet attack and increasing it even more if that pet attack was a crit. Compare that to the pet basic attacks for my MM pet which only had 4 WHs (3 at the start and one during a RF later) for only about 3.5% of its basic attacks being affected by WH.

Thus, considering that the BM pet's basic attacks did about 11.2% of the over all BM hunter damage (as compared to about 2.5% of the DPS for a MM hunter) and that GftT procs from autoshot crits was a major factor (along with FF) in increasing both the number of pet basic attacks from 114 for the MM hunter to 132 for the BM hunter (despite the BM hunter having a lower base pet focus regen rate by 0.4 FPS due to the BM hunter having about 40% of the haste rating on gear than the MM hunter has) and the number of those attacks affected by WH from 4 for the MM hunter to 32 for the BM hunter, it is quite evident that crit has a large impact on BM hunter DPS over and beyond the double damage on crit component.

Then also considering how changing crit rate can affect the number of GftT procs over the duration of the sim and when in the sim that GftT procs occur relative to focus levels and can result in additions or losses of pet basic attacks and pet basic attacks affected by WH, it is easy to see how at certain points where you gain basic attacks or more basic attacks affected by WH why crit would be temporarily highly valued. With a pet basic attack for a BM averaging about 34K, getting 1 aditional basic attack can increase DPS by 76 on average without WH and 152 DPS with WH.

So this is essentially similar to the situations where we have seen haste rating temporarily valued very high due to haste allowing additional attacks over the course of a fight. This is the situation where higher crit results results in more pet basic attacks. We did not see this effect in Cataclysm much since out pets were generally rarely focus starved and almost always performed their basic attacks off CD if we were talented correctly. The crit benefit was more to increased WH uptime as oppossed to increased number of attacks.

EDIT: Now that I have has a little time to digest these findings, I must say that I am really concerned about pet focus levels for all specs, much mostly for MM and SV, which do not have GftT and FF and for which Fervor is currently not the best overall DPS option.

For MM and SV hunters, pet basic attacks are delayed by the pet being focus starved about 70 to 75% of the time, with the times that they are not focus starved being during RFs and BLs. Furthermore, WH might as well be a BM only ability since MM and SV pets rarely see it currently.

With the pet basic attack costing 25 focus and a base pet regen of 5 FPS, with no haste on gear or dynamic haste effects, it takes 5s to regen enough focus for a pet basic attack. This is about 51% longer than the ideal pet basic attack rate of 3.3s. Even with MM's recommended haste amount of 12.93% haste from gear from rotation analysis, the pet regen rate is only 5.65 FPS, still requiring 4.43s to regen enough pet focus for a basic attack. This is still 34% longer than the ideal pet basic attack rate and results in the loss of many pet basic attacks. Over each minute of a fight, this results in a loss of over 4.6 pet basic attacks per minute, which is 23 over a 5 min fight. The situation is even worse for SV hunters who have even lower amounts of haste on gear.

For a MM under BL, the per focus regen rate is 7.34. Even this rate is not enough for preventing focus starving and pet basic attacks being delayed since it still takes over 3.4s to regen enough focus for a basic attack.

For a MM under RF, the per focus regen rate is 7.91. This is barely enough to keep the pet from being focus starved with generating about 26.1 focus over the 3.3s time between pet attacks. Hence, a measly 1.1 extra focus in genereated every 3.3s during a RF for a whole net gain of 5 pet focus during a RF. This is not enough extra focus for a pet toget into the WH range. The situation is even worse once again for SV pets.

I believe that this is a major factor for why I have seen FD favoring haste levels beyond the 12.93% recommended for MMs.

I really think that this is something that Blizzard needs to address. Options include:
- Reducing the focus cost of the basic attack. Even with making a pet basic attack only cost 20 focus, that still requires 21.21% haste from gear and applicable haste effects to make MM and SV pets not be focus starved. Hence, even at this level MM and SV pets would still be focus starved most of the time but would be experiencing less of a delay in pet basic attacks but would be at full performance during BL and RFs with generating a little extra focus over the periods where we might actually see WH once in awhile. Lowering the pet basic attack cost to 15 focus would probably be too low though since then pets would never be focus starved even with 0% haste effects. Although that would bring WH into play more often for SV and MMs.
- Provide MMs and SVs something similar to GftT where they can periodically get pet focus boosts.

Last edited by Whitefyst : 09/02/12 at 12:25 PM.

 09/02/12, 6:48 PM #28 Rivkah Great Tiger   Zeherah Dwarf Hunter   Stormrage I think Whitefyst did a good job of explaining why you're seeing the crit fluctuations due to the changes in my autoshot support. Before, the go for the throat procs were being averaged over the course of the fight, so the pet focus was flatter. Now you can see the effect of peaks caused by go for the throat, but it will create more issues related to shot shifting. Simc would probably be a better source for actual dps per stat values as those things can be averaged over a longer period. I'll give some thought to the serpent sting option- I need to get a better idea first of whether it's actually worth doing that in game before I add in special code for it. Setting it up in simc may give some idea. I did just add support for SV to suppress multishot except for when serpent sting needs applying. This should mean that you can replace serpent sting with multishot in the shot priority to simulate if it's worth using that instead of serpent sting to apply your sting. Since you get the damage from improved serpent sting and the damage from multishot, it's a better option unless that 15 focus difference is going to cost you more elsewhere. This wasn't as attractive in Cata due to weaker focus regen and sting from multi having a shorter duration, but now that it's a full 15 seconds it seems to math out as a win. I also added in support for stormlash totem based on the known implementation data. This isn't a perfect representation as it doesn't perfectly reflect the 0.1sec ICD on it, but it gives a decent idea. There is a field in the buffs now where you can specify the number of shamans who are casting stormlash totem in your group. It will assume that they get cast in series, so essentially if you specify 3 shamans, stormlash totem will be up for 30 seconds (duration * 3) every 5 minutes. Based on my tests, stormlash totem does not stack, so I think this is the best way to represent it. By default I am not enabling stormlash when you turn on raid buffs, as I have no reasonable way to guess how many shamans will be in your group, so you will need to turn it on manually. On the topic of pet focus regen, I raised the issue previously but Blizz doesn't seem to have a problem with it, and it looks like pet dps as a percentage of hunter dps for MM/SV hasn't really changed much, so I guess it's not a huge deal. I am concerned about being able to cast things like dash, which we don't dare leave on autocast anymore. But we don't have other buffs on us as MM/SV that rely on pet casting the basic attack on time so it's not as big a problem now. I was thinking about suggesting they remove WH for non-BM specs, but it's actually still of use on stampede since the stampede pet initial casts will use it, which may mean they get in a higher amount of total damage before despawning depending on regen.
09/03/12, 2:47 AM   #29
Lokrick
Piston Honda

Orc Hunter

Steamwheedle Cartel
 Originally Posted by Whitefyst Thanks. I have been doing some analysis with assuming 1 autoshot lost but with using 1.25 to be conservative, especially since 1/4 of an autoshot and its expected WQ damage does not change the analysis too much. I have been suing both values to provide kind of a range.
I realized I don't understand where your 1.25 comes from, and since some of your other recommendations rely on it, it's worth reviewing. I think the correct number should be derived similar to:

 Originally Posted by Rivkah It looks like the formula I'm currently using to simulate autoshot suppression (which I assume is based on aimed shot mechanics we discussed during Cata) is: (Average Aimed Shot Shot Cast Time / Weapon Speed + .5) * Aimed Shot Count = Autoshots lost
Basically when aimed shot starts, it may be interrupting autoshot anywhere in its progression. Since it resets the swing timer, the cast so far is simply lost. Thus on average you will lose 50% of an autoshot. You further lose the time while casting the aimed shot. Since aimed and autoshot are affected identically by haste, all that matters is their ratio, as above. Since all the relevant weapons are 3.0 second cast time, that leads to:
`0.5 + 2.9 / 3.0 = 1.47`
So there's 1.47 lost autoshots per aimed shot, independent of haste level.

09/03/12, 10:04 AM   #30
Whitefyst
Great Tiger

Orc Hunter

Draenor
 Originally Posted by Lokrick I realized I don't understand where your 1.25 comes from, and since some of your other recommendations rely on it, it's worth reviewing. I think the correct number should be derived similar to:
Thanks for the information.

For clarification on my part, I fully know the affect of weapon speed on the analysis, but since it affects the analysis very little for most weapons that are between 2.8 and 3.0s speed, I choose to ignore it and keep things simpler since it only has a 3% impact on autoshot to AI duration. Since autoshot damage is a small portion of the analysis, 3% difference in it is really noise. That amount is important for yopur simulation accuracy, but it does not make a big deal in the analysis on whether to use AI or AS as the focus dump.

Concerning the autoshot behavior, I was not on beta so originally used estimates of 1.5 and 1.25 in my analysis for to give using AI the benefit of the doubt. Then someone on beta said that AI behavior was improved where is was only suppression and not reset. Thus, I changed my analysis to use 1 and 1.25. I can see now from testing it myself on live that it is indeed both suppression and reset. Thus, I will update the analysis when I do the 5.1.x version. However, since autoshot damage is a small part of the analysis, it does not change any of the recommendations much. AI is still a poor option to use outside the CA Phase unless its cast speed is severely hasted.

I will go back to using 1.5 in my analysis so it better covers 2.8s speed weapons too.

 Elitist Jerks Hunter DPS Analyzer (MoP edition)