As I've gotten a question by PM on the secondary stat values, I realised while answering it that I don't think that I've actually ever explained what I mean when I say "in spite of sim results" in the guide.
As I did explain it in my reply, I am copy/pasting the reply in here (though not the questions), so it can be subject to the same peer scrutiny as everything else. (And apologies for doubleposting, but with almost 24 hours between, I think this merits a post of its own)
The reason I say "despite what sims say" is because if you analyze the effects rather than sim them, it is quite obvious that haste doesn't benefit the BM rotation, apart from breakpoints. The first breakpoint is at 33% (total) haste, where there is room for one more shot in the non FF cycle. This haste doesn't help during FF though, so it is only effective about half the time (FF should have an uptime of ~50%). The second breakpoint is 100% (total) haste with no external haste CDs, where you get a casttime of 1 second on CoS, so you can add 3 (or more realistically 2, and an extra AS due to extra focus regen).
Both these can be modified for FF or BL, but the fact that we only get 2 breakpoints and that 1 of these we have about 50% of the time anyway, haste really isn't that attractive.
As for haste with RPPM, yes RPPM does scale with haste, the effect of which is negligable, as the procs from RPPM don't (or shouldn't) contribute any significant enough damage contribution as to make haste worth enough in itself, there is a linear scaling of haste's effect on RPPM procs, but I haven't seen any procs that will be powerful enough to worry about this aspect.
As for sims and haste in general. Sims, by virtue of their nature, especially optimised sims, will value haste far more than any real situation (apart from making up for latency), as there is a continuous calculation of focus useage and focus gains and miniscule delays of instants that alla dd up to favor haste - its just not something anyone other than a robot will see any gain of, apart from air in their rotation.
If haste scaled damage linearly for all our abilities, we would get a 1% dps increase from haste, for a 1.42% increase from mastery to the pets abilities, so 0.71% overall dps increase. Now we know that haste doesn't linearly increase damage from instants, so in this case mastery is a clear winner.
Comparing to crit on the same scaling, we get the same 0.71% overall damage increase from the crit alone per 1% haste, and crits interactions feed into more damage than that through greater focus gains for the pet, since our pet is still focus "starved" - and will remain so untill the focus regen for the pet is at 16.67 fps. Through the increased amount of pet damage from crit to focus, crit feeds into mastery, increasing the benefits of mastery.
As a proponent of simc and contributor to simc (especially the Hunter module lately), I find that analysis -- which basically has no math to back it up most of the statements -- to be unconvincing. Most of the benefit from Haste comes from the pet, not the Hunter. Moreover, with ability queuing, there isn't really a situation with delays between abilities, and simc factors in lag and reaction delays as well. Beyond that -- Haste's benefit isn't really about a faster CS. In fact, when you add haste, you end up doing fewer CSs, even though they're slightly faster.
In a world of Dire Beast and Invigoration, there isn't really a solid rotation anyway, so discussion of soft breakpoints is somewhat moot. Given Dire Beast does scale with haste -- a fix I put into Simcraft a few weeks ago -- all previous simc results were undervaluing haste by a few percent. The current weights, for BM, have all secondary stats nearly equal (whereas before, Haste trailed by a decent amount).
RealPPM affects the metagem, both of our new tier bonuses, and all new trinkets except the Valor trinket. To make a blanket statement without math that the effect isn't significant is incorrect. I've tested three different gear sets -- BiS T14, T15N, and T15H, and the scaling from Simcraft is roughly:
In T14 BiS, Haste is 96% of Crit's value. In T15N, it's 110% of Crit, and in T15H it will be around 9.5% more. I think, at the minimum, it's more correct to say the stats are of roughly equal weight (and recommend people gear towards Crit > Haste > Mastery if their offspec is SV or Haste > Crit > Mastery if their offspec is MM). Mastery is terrible for SV and MM, and equal to the other stats, so eschewing it in favor of your offspec's preferred stat is probably ideal. SV mastery *might* be more arguable in AOE situations but generally AOE isn't the entire fight and single target DPS still matters.
simc and Esoth's spreadsheet are in rough agreement these days as well; I spent a considerable amount of time comparing both to in-game results. I'm certain both have bugs and perhaps some bias, but not so large as you would imply.
As a point of comparison, for my character (near BiS T14H), adding 1000 haste has the following results over a 6 minute fight (+/- 20% to simulate varying fight lengths):
300 dps from more arcane shots
360 dps from auto shot
14 less dps from cobra shot
90 dps from stormlash procs (including pets)
710 more pet dps
40 more dps from stampede
140 dps from dire beast
(plus some smaller amounts from values, rounding, etc)
That's out of around 130k dps. The 1k haste provided a total of ~1750 dps as the above breakdown summarizes.
I think, to me, the bottom line is the class is so complicated at this point, with hugely varying mechanics on procs from trinkets and tier bonuses (remember, in T14, only our weapon enchant was RealPPM; now pretty much everything is), with things like stormlash and random focus procs, etc, that you just can't handwave away what tools like simc and the dps spreadsheet show. I mean, our gear is hugely different now than when MoP began; weights absolutely are different, due to mechanics and general scaling. Intuitive theorycrafting should not completely disregard what sims and analysis tell us.
Since I don't seem to have sent messages saved, I'll just quickly type up a short repetition of my counter to the above.
The heavy haste bias in sims comes not from being contrary to analysis in regards to pet contribution, but from the very way simulation works. That you set some specific conditions, and the simulation calculates this perfectly per shot - like "save focus for X" in FD - in real life you need to specifically keep track of focus and do calculations each time you hit a button, as well as adjust those depending on procs etc. The gains from haste in sims is due to the very very slight increase in passive focus regen, we are humans, not robots, so these calculated gains from haste will not manifest in reality - your example of 1k haste ratingis a difference in casttime of 69 ms on a CoS - thats a lot less than what most people experience in latency.
Your example of 1000 haste is 2.35% haste, that means we are talking about 2.82 autoshots over a 6 minute fight (122.82 versus 120 when going from 0 to 1k haste rating - the amount of increase is the same regardless of starting haste rating).
At a ~30% crit rate that means you get .085 more GftT procs from 1000 haste rating. 1000 crit rating would give you 1.67% crit, which translates into 2 GftT procs (more than before the 1k crit).
Thats the easy comparison of pet damage contribution from the hunter in regards to haste and crit.
Mastery at 1k rating results in 1.67 mastery points (which rhetorically don't exist) and this translates into 3.34% increased pet damage for all damage attributed to the pet (including KC).
The pets autoattacks; 2.35% haste increases the pets contribution via autoattacks by 2.35% (very easy). The crit inctreases the pets damage from autoattacks by 1.67% - haste is better for pet autoattacks than crit. The mastery increases the damage from pet autoattacks by 3.34%, mastery is, hands down, better than crit and haste for pet autoattacks,
**I have to go to a meeting now, so I'll come back and edit this post to continue later, unless there has been a lot of discussion at which point I migth continue in a new post instead**
...The heavy haste bias in sims comes not from being contrary to analysis in regards to pet contribution, but from the very way simulation works. That you set some specific conditions, and the simulation calculates this perfectly per shot - like "save focus for X" in FD - in real life you need to specifically keep track of focus and do calculations each time you hit a button, as well as adjust those depending on procs etc. The gains from haste in sims is due to the very very slight increase in passive focus regen, we are humans, not robots, so these calculated gains from haste will not manifest in reality - your example of 1k haste ratingis a difference in casttime of 69 ms on a CoS - thats a lot less than what most people experience in latency...
There are two issues here that I want to address, the sim one and the latency one. This is not an accurate portrayal of simc, which works very differently from FD. Simc simulates latency, reaction time, and mistakes in priority. For example, the default action list knows tha tyou plan for rapid fire and so ther'es no reaction time, but assumes that you react to bloodlust rather than know exactly when to expect it. Similarly, it knows you react to procs like LnL. Etc. It's also the case that the action list is not (at least currently) as clever as a human can be at planning ahead.
For latency, as Arison noted, the cast queuing eliminates much of the human element in latency. If it takes 69 less milliseconds to cast a cobra, then the cobra I queued up 400 ms ago will be cast 69ms faster, even for a human. Since the other haste effects are all independent of the human, haste effects are pretty well reflected in the actual game.
Nooska, there are some corrections to your numbers that affect the result. First; haste scales more than you suggest from other haste effects (it's multiplicative with flat percent bonuses from raid buffs, heroism, rapid fire, focus fire, etc). In a real 6 minute hard-limit sim (ie, no fight length variation), it's really 175.9 -> 179.8 auto shots. So 3.9 auto shots, not 2.82. Regen wise, it's 949.6 focus from GftT to 968.8 focus, which is a bit over one extra GftT proc, not 0.1 procs (not sure why you say 0.085; raid buffed, you should be around 33-36% crit for autoshot, so even your 2.82 autoshots would be one extra crit and thereby an extra proc).
It's definitely true that 1000 mastery is more pet dps than 1000 haste. However, the 1000 haste affects the hunter's dps pretty well -- enough to make the stats equal. Mastery only affects AMoC, KC, and Dire Beast. It doesn't affect Stampede, which Haste does. Haste also affects Dire Beast, but doesn't affect AMoC or KC.
For a hard-limit 360 second fight (no variation; this is to get the exact autoshot numbers, etc), the DPS for T14H BiS is:
The 1k mastery basically gives you: 1285 pet dps, 150 stampede dps, 170 AMoC dps, and 173 dps from Dire Beast.
The 1k haste basically gives you: 343 dps from AS (2.5 extra AS at 52000 each), 367 dps from auto shot (4 extra shots at 33000 each), 123 more DPS from Dire Beast, 110 stampede dps, 755 pet dps.
The 1k crit boosts pretty much everything uniformly. The DPS delta 136854/135040 = 1.34% difference, which is quite close to the expected benefit strictly from crit at that point -- a 1.36301/1.3463 = 1.24% dps increase. The rest of the delta is from extra pet melees thanks to GftT -- a whopping 0.2 average extra hits per fight. However, really the benefit seems to come from those tiny extra hits plus occasionally topping 50 focus which doubles the next hit. The incremental focus builds up. Also, Skull Banner is a very slight boost here as well. These tiny amounts add up.
These nuanced details of raid mechanics have to be factored in. The intuitive "0% haste means one AS every 3s" is just an oversimplification and ignores so many buffs and interactions that it really isn't worth considering. You have to account for every detail of what happens in a raid if you really want to say what stats are worth. Only then can you truly get the rich interaction of abilities and estimate relative stat weights.
Instead of editing (since its been over 12 hours) I'll make a new post.
As to the 0.085 GftTprocs, thats a typu, it was supposed to be 0.85 - as for the crit rates, I just arbitrarily set a 30% crit rate, as thats where I am (was) when raid buffed in a 10 man without procs. I'll also grant that on average it would be a bit higher. Things like Skull banner and Stormlash totem are not things I want to consider in this discussion, as they are not a given (and I understand from simc discussion that 2 warriors and 2 shaman are assumed in the best-buff situation).
Just to add number sinto the mix (using FD as thats the tool I'm most comfortable with[note: and an 8 minute fight time for some reason - either default or previous setting, but all af the runs are at the same settings so that shouldn't matter for this discussion]), I did the same excercise adding 1k of each of teh stats - instead o finding some BiS profile which may or may not be BiS (the one I worked with back when MoP was new behaves oddly) I used my own gear, since a recommendation on secondary stats needs to be valid for a range of gearranges and not just a BiS (at which point the recommendations are useless anyway, since you don't get better than best).
Using my own gear means that the base dps is quite a bit lower of course.
Delta / %
∆ 929.11 / 1.05%
∆ 1116.59 / 1.26%
∆ 1276.97 / 1.44%
So far my results agree with yours in regards to total dps contribution, crit has a 37% higher dps contribution than haste for the 1k added, but in the overall contribution the difference isn't something to get upset about - ie. if only haste pieces drop you aren't totally screwed, you are just a bit below optimum, but any real life situation will leave us below optimum anyway.
Now our earlier discussion of synergy led me to repeat an earlier experiment, to see if it still held true under the changed environment, so I did the same 1k additional stat added, but split it 50/50 between 2 secondary stats and for completeness sake I made a 3-way split with crit getting the extra point.
Delta / %
∆ 1483.28 / 1.68%
∆ 1103.56 / 1.25%
∆ 1077.71 / 1.22 %
∆ 1258.7 / 1.42%
As is obvious from the deltas splitting crit/mastery is the absolute highest gain (and also what I found when I was using T14H sets when I originally did the research back in 5.0). Haste/mastery is more dps than pure haste, but less than pure mastery. haste/crit is again, more than pure haste, but less than pure crit, and even less than splitting the haste with mastery - crit and haste just doesn't synergize well. The 3-way split is better than the 2-way haste splits, but still below pure crit.
The delta contribution between crit/mastery split (the highest contributor) and pure haste (the lowest contributor) is 59.65% higher dps contribution from the same amount of stat points.
I want to clarify that I never said that haste had 0 impact, just that the impact of haste is so low as to not matter when actively gearing.
As for simulations, I know that both Zeherah and Simc tries to remove as much bias as possible, but by the very machine nature the sims can achieve and see results a human cannot react to.
Since your number show the same trend in the single stats, I will assume (for now) that the same trend would hold with the split additions. Which leaves me unconvinced.
Oh and a note about mastery - it does affect stampede, as that is pet damage, just as MoC and DB is, it is just multiplied by 0.25 either before or after the fact.
Yeah, I thought I'd removed the comment about stampede and mastery; I thought they didn't interact but when my results with simc clearly showed they did, I went back and edited, but missed that reference.
Your results on stat splitting definitely are true. It more or less comes down to (1 + x) ** 3 being greater than 1 + 3x, ie, splitting the budget among different stats produces a higher number than dumping completely in one. It's a small difference in many cases but it's there. Really though, for pretty much every hunter item in the game, ilvl trumps everything else right now, and stat weights are really just about reforging, enchanting, and gemming -- rarely gear choice (except some cases with weapons, thanks to MM being almost-viable now).
FD has some odd behaviors with haste; for me, if you trust it's weights, haste is better than Agility. That said, the 1k stat delta is pretty close to what I get with simc. One thing FD seems to be doing is reducing the number of KCs when 1k extra haste is added, which should never happen; otherwise the haste number would be a bit higher. I'll discuss that in the other thread.
My point all along really has been that "Agi>hit==exp (to cap)>>mastery>~crit>>haste" is bad guidance in late T14 and especially as T15 progresses. RPPM is a very real change to how gearing has ever worked.
I think a more pragmatic view is basically that all stats for BM are pretty close to equal and that Haste will improve over the tier, but gearing/enchanting/gemming choices should be made based on whatever your offspec is rather than BM itself. Most hunters have SV as an offspec for AOE-intensive fights, which pushes Crit pretty high (mastery, arguably, though it's so weak that it takes a lot to compensate for the lower single target dps).
Really though what I think matters most -- since RPPM will vary for each hunter based on their gear acquisition rate -- is that stat guidelines need to be verified on a per-hunter basis. There's no cookie cutter answer, but it's definitely not true that there really are major breakpoints or rotational shifts that make haste a non-smooth scaling stat.
With haste being the index of 100%, crit is worth 137% and mastery is worth 120% and an even split between crit and mastery is worth 160% of haste, per the simulations. Nota bene to the simulations is taht if anything they favor the value of haste. The above calculations were done with my own current gear but match up with the simulations I did early MoP, and as an aside, with every simulation where I've analyzed the result for secondary stat worth.
Have to say after reading this all I am is more confused. I prefer survival but have been wanting to try out bm for the possibility of pushing dps higher but trying to figure out how to go about regemming/reforging has been a nightmare even more so after reading the 2 of you go back and forth. If simcraft really isn't giving us dead on stat values then what method should we use to figure out reforge strats? Arison I actually looked at your logs of 25 normal tot and comparing you and beastgirl where she has higher mastery vs your higher haste and both of you have seemingly high crit. On fight where you both were bm she outdid your dps with the higher mastery setup. If I read anything wrong please let me know my only goal is to try to understand the better layout to try and increase my own dps by changing specs if possible.
In T14, going mastery to increase BM was feasible and even encouraged. In the mechanics-heavy T15 content, it's inadvisable to go mastery-heavy due to the utility and aoe of SV (Horridon adds, Council, Tortos bats/turtles, Ji'Kun fledgling, Primordius blobs, and Lei Shen adds) and not wanting to gimp the spec. Additionally, crit-mastery gear is rarer within the tier and crit-haste gear is much easier to come by.
To be more versatile in the T15 content, you should be going more haste than mastery, in general. It's not only BiS in both simcraft and the spreadsheet, it's also better for alternate specs that you're going to want to bring to the Throne of Thunder.
Reforging should be easy. Grab Reforgelite addon, press calculate, then show, and follow its suggestions for the default Survival profile. It will get you close to the hit and expertise caps and try to maximize crit and then haste. Take what it has and import it into femaledwarf.com and tweak it until you get the best dps numbers when you have time after raid. The most important part of the new tier is learning the mechanics. If you can eek out 2k more dps great, but unlike T14, the encounters are much more involved and everyone staying alive until the end is more important (especially on first kills when you generally don't overgear the content).
I am still lacking any backup, whatsoever, that haste is superior to mastery or crit - even in the hastebiased simulations, haste loses to crit and mastery (not to mention a crit/mastery combination), by not insignificant (relative) amounts.
Also, the more mechanic heavy a fight is, the more mastery is better than crit or haste, because only when the pet is not on the boss, will mastery not be beneficial. As for movement heavy fights, tmovement makes no difference after the change to CoS being castable while moving, you can continue your rotation while moving, as long as you make sure you don't have your back turned to your target.
Haste remains the dump stat, and you should aim for 0 haste rating if at all possible (haste may feel good because it shortens casttimes and thus make you feel more respponsive. It also helps for hesitation, as you need to be very tight to take full advantage of the "perfect" lineup of CoS, AS and KC in a 0 haste environment).
The 10% attack speed raid buff will get you over the first CoS haste threshold with room to spare, and you really don't need anything else.
As an aside, if we got a mechanic like paladins Sanctity of Battle, that made our GCD and cooldown abilities(specifically KC) scale linearly with haste, it would be a different matter and haste would definately become a very good stat as it would actually let you "do more" - as is, haste doesn't let you "do more" - or rather it does but crit is better at it than haste, and mastery is a greater damage contribution than haste, so haste loses on both fronts (which is also part of the reason of the strong synergy between haste and mastery).
That haste is prevalent on gear and that our optimal gear is rare, doesn't make haste more attractive (quite the contrary, as you will alreadye have more than you need, and no way to get enough for the next plateau)
I am still lacking any backup, whatsoever, that haste is superior to mastery or crit - even in the hastebiased simulations, haste loses to crit and mastery (not to mention a crit/mastery combination), by not insignificant (relative) amounts. ...
This sounds reasonable however there's clearly synergy between stats that is extremely hard to evaluate. Simc works very hard to accurately and usefully model the game, including latencies and response times. Thus, it should not be and I know of no evidence that it is "haste biased". For context, I'm not doing heroic raids at the moment and have no stat or gear preference. The simc T15H profile is reforged for crit>haste>mastery because that provided the highest dps at the time for the given gear set (I think because of the increased number of hits during stampede). It's always worth revisiting, considering how fight mechanics affect the priority, etc. Based on the above conversation, I decided to try a reforge-plot. That trades a set amount of one stat for the other to produce data that could go into a graph. With just two stats it's actually pretty fast (especially given some recent simc performance improvements). I simmed at every 500 points of reforge from haste to mastery. Of course you can't reforge in 500 point increments, but this gives a good picture across a wide range of gear and reforging for this crit level. The command line was:
The profile starts with 8835 haste rating from gear (20.79%). Two observations from the data: Reducing haste reduces dps. Increasing haste is flat until +3k haste, which gets into the "more attacks from our dire beast" territory.
I also compared crit vs. haste: reducing haste in favor of crit lost dps. Indeed, it's a (small) gain to reforge crit to haste up to +7000. Reforge of 7K crit to haste yields +500dps (which seems not worth the effort, but interesting nonetheless). And for completeness I did crit vs. mastery, and the current T15H is sitting at a wide, flat peak where any reforge there will be flat or lower dps.
To actually figure out *why* these changes lead to improvements will require more digging into the resulting action sequences. For example, it might simply be that the burst trinkets line up better with stampede. I'm working on a different priority list (from the community) that does result in better trinket stacking on average, resulting in +1500 dps. That's enough of a difference to be worth doing if it's not otherwise difficult to play.
EDIT: I should add that for my lower-end, current gearset, reforge plot shows Crit > Mastery > Haste, so the priority is strongly gear dependent.
For your current item level, mastery may still make sense because so much of your damage is based on ranged attack power and not weapon damage, but with the increase base weapon dmg in this tier's weapons and the consequent trend of hunter dmg now eclipsing pet dmg in BM (even in the custom profile I made for you and linked above, the hunter dmg is now 58% of total dmg when maximizing mastery), it's not just that the "optimal mastery" gear is rare. It's that haste is eclipsing mastery in this tier. Any hunter currently gearing for normal or heroic Tier 15 content cannot get away from haste and shouldn't want to.
Even with the switch to haste in gearing, BM is still outdpsing SV in the simulations. I notice that in your breakdowns, you have not included tradeoffs in weapon dmg with haste/crit/mastery. You may want to revisit that if you are really wanting to understand the relationship between haste and mastery and why haste is now better for all three specs of hunter for this tier. This relationship between haste and base weapon damage is much more clear now that weapon speeds are normalized to 3.0, and it's worth noting that mastery is orthogonal to weapon dmg (everything is based off RAP (agi) and mastery) whereas the value of haste does scale as weapon damage scales.
Thank you Lokrick, I will be looking mor einto that in the coming days where I shoudld have time to do so.
In regards to the "BiS"BM profile from the max dps thread, that is not conducive, nor conclusive to anyreal theorycrafting. That thread (liek its predecessors) have one aim, finding the exact set of circumstances that theoretically give you the best dps under e predefined set of variables. Since hunters reading a guide will be of various races, and at various gear levels, what contributes the most to the final dps in the very theoretical setup isn't applicable.