From a private message:

First, the 1850 haste rating is equivalent to 14.45% haste from gear and 17.88% haste from gear and Pathing with 3/3 Pathing. With no Pathing this requires 2290 haste rating.

From a quick skim through their posts, their desire for this haste rating seems to be able to mimic their rotations with the T11 4-set. This haste level equates to a 1.34s SS cast time. With the T11 4-set 1.8s SS cast time, about 3.0% haste from gear alone is required to achieve this cast time.

Thus, they are using this large haste amount to mimic the low haste CS cycle with the T11 4-set. Basically, then need about an additional 11.5% haste from gear to make up for the loss of the T11 4-set reduced base SS cast time. The difference here is that they give up a lot of crit and mastery to achieve this haste level relative to when they had the T11 4-set. This option is really useful if you are used to the timing of that rotation cycle and do not want to learn new timings.

The Standard Phase unhasted shot table is provided below:

CS-ASx2-SSx4 (MMM AI procs replace an AS; extra SS for focus replaces an AS)

Shot	Time	Focus	Focus2	Focus3
CS	1.00	60.72	60.72	60.72
AS	2.00	43.43	43.43	43.43
SS	3.34	58.75	58.75	58.75
SS	4.68	74.08	74.08	74.08
AS/MMM2/SS3	5.68	56.79	78.79	89.40
SS	7.02	72.12	94.12	104.73
SS	8.36	87.44	109.44	120.05
Net Focus	0.07	-12.56	9.44	20.05
Cycle Delay	-0.62	-0.64	-0.64	-0.29
Ratio		9	10	1
#Autos		4.63	4.63	4.63
Net-T12AS	10.25	-2.38	19.62	30.23
Net-T12CS	20.43	7.80	29.80	40.41

For the absence of the T12 4-set, this cycle works out pretty well from a focus point of view. It is almost focused balanced naturally and requires a 5th SS to balance focus rarely (mathematically about every 20th cycle). As such, it pretty much totally wastes the benefit of the T12 4-set. Furthermore, Termination and the addition of the KS results in additional focus wasting for unhasted rotations.

From a timing consideration, it leaves 0.64s of deadtime in the normal cycles. This means that even at 200 ms of latency, you still have some deadtime in your cycle. This haste amount seems ideal for the CS glyph and someone with around 200-250 ms latency. It does not seem ideal for players with very low latency.

Hence, this cycle does not seem to make sense at this haste level since it wastes focus and has deadtime for low latency situations. To explore other options, I decided to check out adding another AS to use the deadtime and focus. It is the following shot numbers and table:

CS-ASx3-SSx4 (MMM AI procs replace an AS; extra SS for focus replaces an AS)

Shot	Time	Focus	Focus2	Focus3
CS	1.00	60.72	60.72	60.72
AS	2.00	43.43	43.43	43.43
SS	3.34	58.75	58.75	58.75
SS	4.68	74.08	74.08	74.08
AS/MMM2/SS3	5.68	56.79	78.79	89.40
SS	7.02	72.12	94.12	104.73
SS	8.36	87.44	109.44	120.05
AS	9.36	70.16	92.16	102.77
Net Focus	-10.69	-29.84	-7.84	2.77
Cycle Delay	0.45	0.36	0.36	0.71
Ratio T12AS		1	2	1
Ratio T12CS		1	1	0
#Autos		4.82	4.82	4.99
Net-T12AS	0.00	-19.25	2.75	13.75
Net-T12CS	2.35	-8.65	13.35	24.73

From a focus point of view, this cycle is significantly focus negative without the T12 4-set. With the T12 4-set (and my conservative estimates on the proc rate), The cycle is exactly focus balanced with replacing the 3rd AS with an extra SS on every 4th cycle when all procs are consumed by AS. With all procs consumed by CS, the cycle is slightly focus positive with no extra SS cycle. Thus, the cycle is close to focus neutral with mixed AS and CS consumption of the T12 4-set proc without any extra SSs and works out ideally.

From a timing consideration, the 3rd AS results in 0.36s delay in your CS cycle, eating into a little bit of the benefit of the CS glyph. This still leaves about 0.64s of CS glyph benefit. Hence, this option is good if you have low latency with getting about half the benefit of the CS glyph at 50 ms latency. At around 200 ms latency, you have pretty much eaten up the benefit of the CS glyph and are better off using the RF glyph instead at this haste level.

Now let us see what the cycle is during RFs. With trying to keep the benefit of the CS glyph, the first cycle explored is:

CS-AIx2-SSx5 (with MMM AI procs replacing the extra SS)

Shot	Time	Focus	Focus2
CS	1.00	66.60	66.60
AI	2.39	31.33	31.33
SS	3.39	50.93	50.93
SS	4.39	70.53	70.53
SS/MMM2	5.39	90.13	81.13
AI	6.78	54.86	45.86
SS	7.78	74.46	65.46
SS	8.78	94.06	85.06
Net Focus	-11.00	-5.94	-14.94
Cycle Delay	-0.22	-0.22	-0.22
#Autos		4.48	5.48
Net-T12AI	14.46	16.46	12.46
Net-T12CS	11.47	13.77	9.17

This cycle has 0.22s deadtime. Thus, it works out to a tight cycle with about 67 ms latency when factoring in that the SS casts only takes about 0.96s. This cycle is decently focus negative without the T12 4-set, but it is focus way focus positive with the T12 4-set procs.

To utilize that extra focus, I investigated an option to replace the extra SS on non-MMM AI cycles with another AI cast; however, that cycle is way focus negative even with the T12 4-set procs. But even more critical, it is not a feasible cycle since it focus dips by over 100 focus.

Summary

The haste recommendation from Paragon is a feasible option depending on your latency levels and preferences. It also keeps your shot timings about the same during the unhasted cycles as they were with the T11 4-set if you utilized the low haste option with it.

Whether it is better than my 12.93% haste recommendation is difficult to state. Assuming 3/3 Pathing, it requires about 615 more itemization in haste instead of crit and mastery. At lower T12 gear levels, this is likely to be more costly than at higher T12 gear levels. You also lose the benefit of alternative glyphs over the CS glyph and the possibility to move a point in Pathing into Frenzy.

Using Female Dwarf to do a rough and quick analysis with a 378 gear set with realizing FD does not model these rotations perfectly and at 75ms latency, the changes from my recommendation to the Paragon one ranged from -152 when all the haste came from crit to +76 when all the haste came from mastery. Considering that a decent mix of mastery and crit is needed to make up this additional haste amount, the difference between the two approaches at the 378 gear level is close to a wash and preferential. I have not done any analysis with 391 gear yet, but I anticipate that the two options will still be relatively close with the Paragon approach gaining a little at the higher gear itemization where more haste is naturally available on gear.

Correction: The above numbers were done with using the RF glyph instead of the CS glyph. With the CS glyph, the range was from -100 with using all crit to get the haste to +29 with using all mastery to get the haste. The difference between the CS and the RF glyph in FD is about 220 DPS in favor of the RF glyph. Note that this difference is a little misleading since FD does not model the rotations exactly and often cast an extra SS pushing back the CS CD from time to time.

I have a problem here, as i see everyone use arcane shot as focus dump when unhasted and boss under 90% but why does femaldwarf shows for me that using arcane is 2.2k dps loss?
With my gear when i put arcane out of rotation and put aimed to manul cast and instant i have estimated 28885 dps ,
but when i put standard roation wich everyone uses i get 26648 and thats quite a diference...
Femaledwarf used to be fairly good site but with this diferences does it mean its completly useless now?

That's because the approach you describe is probably simulated here: Simulationcraft Results (there's MM Arcane and MM non Arcane which means Aimed is your focus dump in whole the fight; I wrote probably because I am not 100% sure about the assumptions used in that simulationcraft). In FD you can't exactly simulate the "arcane strategy" you're talking about. Of course you may mark the option: "Disable Arcane Shot while Careful Aim is active" and it won't be casted during CA phase, but that's it. It still will be casted during RF and Heroism so it will be casted during second wave of RFs or at the end of your heroism if then you're already outside CA or during second heroism if in settings you type in the length of the fight longer than 10 minutes. People suggest to use arcane outside CA and when unhasted because in simulationcraft the difference - as you can see - is not any big which favours it in the fights you're forced to move a bit - so in most of them.

You can actually determine how aimed shot is used in FD by changing the "maximum speed to cast aimed shot" setting. So if you set it to 2.0 it would only cause you to use aimed shot if it was a 2.0 sec or less cast, and otherwise arcane would be used.

Femaldwarf has all options to simulate rotations and as m8 in post after yours says you can set maximum cast time for aimed shot, but what ever options i use there is always more than 2k dps loss if i use arcane instead of aimed and that realy BIG diference expecialy for min maxers XD, so what i want to ask are you guys also geting this much of a diference, and if so does that mean that femaldwarf has some major isues? Simcraft gives a litle bit better results with hardcasting aimed shot to but diference in femaldwarf is to big so i want to know who should i trust femaldwarf or simcraft...

I don't know, maybe make your arcane and aimed FD profiles publically available. I indeed wasn't using the option of maximum aimed cast speed before, but now when I test it - it shows me that aimed hardcasting gives me ~4% more dps than arcane. In your case it's ~8%. Try maybe also to check "Simulated Shot Breakdown" in both cases.

FD does have some rotation modeling issues that can affect results; however, that is true for all simulations. If you properly configure FD for each situation you are trying to model and understand its difficiencies, it can be a very useful tool.

Using my character as an example in FD using the recommended setup in the MM guide with AI when sufficiently hasted and during CA and AS when unhasted otherwise, FD gives me 31545 DPS when properly set up for this case.

If I go to a case that uses AI all of the time without changing any other settings besides removing AS from the shot priority, I lose about 400 DPS. This is not real though since I didn't correctly adjust other settings such as the one that controls when I hardcast AI. When I fix the AI hardcast setting, I am still at about 80 DPS less because of how FD by default will not fire AI until it has saved up the 90 focus to be able to cast CS right after the AI even if CS is on CD for a while yet and you will be casting SSs before it to regen. Turning off that setting to allow AI to cast earlier, ends up being about a 150 DPS increase over the original recommended case. So AI is indeed better theoretical DPS; however, if you have 1 or 2 AIs interrupted, it is now worse DPS. Of course, the AI case can be even more improved upon by adjusting haste levels appriopriately to tighten up the CS cycles and improve focus consumption. But even with those improvements, you are looking at a few hundred DPS difference that can be overcome by a couple interrupted AIs.

Starting with the recommended case again, if I instead remove hardcasting AIs at all times, it is a 1300 DPS loss.

The above shows that if you do not configure FD properly you will get erroneous results with misleading conclusions. To see how well FD is modeling what you want it to do, look at how it performs its shot selections in the shot simulation. If it is not performing anywhere like what you expected, then you probably have some bad settings, but even with proper settings, it will never do a 100% perfect shot selection for a repeatable rotation.

Please check the recommended FD settings section of the guide for some setting recommendations in FD.

The difference with t11 4pc between AS and AiS is around 500-600 dps, when FD is correctly set up. That is consistent with Simulationcraft. (check my public profiles in FD, search for user "pichu")

The difference comes when jumping into t12 4pc, beacause it is hard as hell to simulate the procc use, and even more hard for FD to simulate the 2 or 3 possible different cycles in the same combat. AiS hardcast simplificates this task a lot, reduces the cycles differences, hence it makes the simulation a lot more accurate.

Even then, the difference with 4pc t12 is not higher than 1k dps in favor of AiS, and this is consistent again with Simulationcraft. The great advantage that SC has over FD is that you can set up every single shot, not only a mere priority list.

Where you are off is that the CS-ASx3-SSx4 shot selection is recommended for those without the CS glyph. With the CS glyph and essentially one less GCD, you have to remove an AS since you are constrained to have the CS and 4 SSs to maintain ISS. This results in a shot selection of CS-ASx2-SSx4. Including MMM AI procs, this rotation is almost focused balanced without the T12 4-set. With the T12 4-set, this cycle wastes about 15 focus per cycle with an “average” of the proc being consumed sometimes by CS and sometimes by AS when implementing the shots in a minimum focus dipping cycle of CS-SS-SS-AS-AS-SS-SS. It is partly because of this high amount of focus being wasted that I recommend using a glyph other than the CS glyph, although it is still a viable option.

Concerning the 0.4s deadtime in the cycle, this is due to the 12.93% haste case being designed to be able to cast 5 SSs in 7s for the cycles where an extra SS instead of AS is needed to balance focus. The extra SS is more applicable when you do not have the T12 4-set, when you have the T12 4-set, the extra SS is not needed as often and haste can be reduced some. However, the nice thing about the 12.93% haste case when you are just casting 4 SSs is that it makes up for exactly up to 100 ms of latency. If you operate at less than 100 ms latency, then you can reduce the amount of haste some to both reduce the deadtime and get more crit and mastery with just realizing that you will have a small delay in the CS cycle on cycles where an extra SS is required.

If you are using the CS glyph and have low latency, then the lower haste case is probably better since it results in a tighter 4 SS CS cycle and wastes a little bit less focus with the T12 4-set (only about 11 per cycle).

Yes, the averaging of proc abilities is one difficiency of FD. If you want to evaluate the benefit of the proc in FD, the best thing to do is empty the trinket slot and hand adjust the static amount in. Running FD here will provide the results for when the proc is not in effect. Then you can hand adjust in the proc amount and run FD again to see the benefit of the proc. If you want to see how it affect different phases of the fight, then you have to adjust settings accordingly. Thuis is obviously a cumbersome situation, which is even more cumbersome when you have two procs you do not want to average.

Ta da! Just in case people weren't aware, FD has changed the way in which haste procs are modeled. It seemed pertinent to the last couple of comments in this thread, here's the link for anyone who wants.

Hunter DPS Analyzer (post #706)

The reason I recommend using a RF pair during the KS phase is the combination of factors:

1) Full Termination, especially if you have the T12 4-set, really only has its best benefit when you are hasted and using AI as your focus dump. Only a single point in Termination is needed in the Termination phase if not hasted to have a focused balanced CS cycle without the T12 procs. Even a single point in Termination is mostly wasted in the KS phase if you are not hasted.

2) A single KS replacement of an AS per CS cycle (not glyphed for KS) in the KS phase already balances focus without the T12 4-set and without Termination, with any combination of the T12 4-set, Termination, and the KS glyph, you are wasting focus when unhasted in the KS phase.
These factors combined make it very worthwhile to be hasted during the KS phase so that you can cast AI as your focus dump. The additional focus from Termination and casting focus free KSs empowers more AIs per cycle on average than you can achieve with RFs during the Standard phase, or even the short Termination only phase.

The DPS gain could be potentially higher with KC replacements of AS for T12 4-set procs, but FD does not currently model this correctly. I am informing Zeherah of this.