I've done some maths on the AB cast time, specifically how much haste is required to get AB4 to a 1 second cast time. Using the formula ((2-0.n)/h)*0.9 where n is the number of AB stacks, h is the haste acquired from gear alone and 0.9 is the 4pcT11 bonus multipler.

With the 4pcT11 bonus;

Hard cap = 39.8% or 5097 rating.
With 5% = 33.2% or 4252 rating.
With TW = 7.5% or 961 rating.
With 5% and TW = 2.4% or 308 rating.

Without the 4pcT11 bonus;

Hard cap = 55.3% or 7082 rating.
With 5% = 47.9% or 6134 rating.
With TW = 19.5% or 2498 rating.
With 5% and TW = 13.8% or 1768 rating.

I'm not one hundred percent sure how Blizzard deal with the decimal rounding so these ratings may be off by +/- 1 rating.

Edit: I've also now done the same for Trolls with the Berserking racial.

With the 4pcT11 bonus;

Hard cap = 33.1% or 4239 rating.
With 5% = 26.8% or 3432 rating.
With TW = 2.4% or 308 rating.
With 5% and TW = 0.975% or 124.86 rating.

Without the 4pcT11 bonus;

Hard cap = 47.9% or 6134 rating.
With 5% = 40.9% or 5238 rating.
With TW = 13.8% or 1768 rating.
With 5% and TW = 8.4% or 1076 rating.

Again these may be off by +/- 1 rating.

I must admit I do not know the alliance characters at all so if there are any haste cool downs they have that are not 20% haste please let me know and I shall do the numbers for those as well.

The effect is definitely there. On a test I just made looking at haste scaling graph there is a noticeable drop from about 1.5 to around 1.1 dps per stat point at the soft cap.

AceRider, what about with the goblin 1% haste racial?

I think the main reason that nobody has been able to give a definitive answer as to the exact value of haste when mastery becomes the better stat is that there maybe a variety of reasons why this occurs. There are three reasons I can think of, others may be able to add more. (these would be true for any point that mastery overtakes haste as the optimal secondary stat)

1. At around 1200 haste you may need to break AB stacks more frequently with AM or Abr to maintain a mana neutral rotation. In this case although you will be casting more spells in time period it is possible those spells are AB at low stacks or AM/Abr rather than AB3/4. therefore an increase in haste will increase dps but at a lower rate.

2. You may see a decrease in the amount of ABs that can be cast during the burn phase (not just during AP) due to less Mage armor/ replenishment etc ticks during this time and reaching 35% mana sooner. The extra haste may give an overall dps increase just at a smaller rate.

3. It is possible that you can get an extra AB cast in during one or more of the AP phases. If this occurs at x haste then any haste over that amount will be worth less than before it.

1 & 2 are also impacted upon by the amount of intellect/mana and to a lesser extent crit a Mage has. The more max mana and replenished mana a Mage has the more haste they will be able to have before it has a decrease in value. It is therefore likely if one or both of these are the cause then each Mage will have a unique value where haste is overtaken by mastery.

If 3 is the cause it will be easy to work out for yourself.

Unless it has been changed, it's my understanding that the GCD scales inversely with haste in the same way that cast times do, so that to reach a minimum of 1s GCD, with a base GCD of 1.5s, would require 50% spell haste.

Comparing the GCD affected by haste to an AB4 cast with 4pt11:

AB Casting Time = (2 – 0.1n)0.9/(1 + Z) < GCD = 1.5/(1 + Z)

with Z being proportional to haste. So, since the AB4 cast time is always less than the GCD, the limiting factor in how quickly AB4s can be cast isn't the AB4 cast time, but the GCD.

To get the GCD to a minimum of 1s as already stated would require 50% spell haste, and would leave the AB4 cast time with 4pt11 at 0.96 seconds, not 1 second.

The non-troll, modified 4pt11 haste values would then be

Hard cap = 45.6%
With 5% = 38.7%
With TW = 12.02%
With 5% and TW = 6.69%

The non 4pt11 values would be unchanged as the corresponding cast times are always above the GCD.

So any idea if Arcane Power would be better used on cooldown with 4p t12? And if so how would that affect the burn phase? Just burn with mana gem/trinket and use arcane power every time its off cool down?(assuming you've already used it once in first burn phase and once in conserve) Coincidently if ap is used in burn phase what should be done?Keep mana above 90% even with ap up? Burn til its over and flame orb/missiles back to high mana?

Still haven't gotten the 4 set to test out so pardon my endless questions, I saw it mentioned that simc shows ap used on cooldown to be a dps increase but I'm still confused as to how it would be executed. Faster cast time on ablast when burning coupled with being able to burn a bit longer with mana savings seems that we should still save ap for burn phase. Any insight or experience with those that already have the 4p?

Int is by far the best stat for arcane mages, and it isn't surprising that rawr (or any other simulator) is telling you to gem straight int. Int gives us initial mana, mana regen, spell damage, and a small amount of crit. A +40 int gem is going to increase your dps by more than a +20 int/ +20 haste gem with a +10 int socket bonus. If you're below the hit cap however, you should fill your blue sockits with int/hit gems. After int, it is very important to reach the 17% hit cap, and then run sims to figure out what to do with haste/mastery/crit.

This is all according to rawr, and what i've been noticing in game.

This leads me to MY confusion: Why is there such a large discrepency between rawr and simcraft stat weights? I recently gave up my 359 4-set for 378 gear and a ruthless helm, and have been trying to decide if the gained int is worth the lost set bonus and resil. I'm not using the Lightning Capacitor trinket Malcophant is, but simcraft still devalues haste by a considerable amount.

Here is my armory link

Rawr WPF 4.2.3 stat weights: (overall dps 30900)
Int: 4.24
Hit: 3.52
SP: 2.9
Haste: 1.55
Mastery: 1.52
Crit: 1.27

Simc-420-4 stat weights: (overall dps 31096)
Int: 5.06
Hit: 3.17
SP: 2.49
Crit: 1.34
Mastery: 1.30
Haste: .90

I made the sims as identical as possible (300 second patchwerk fights, fully raid buffed), and ran 50k iterations on simcraft. Any ideas why these two sims can come up with similar dps, but very different stat weights? Does one of these tools have a known bug in the latest releases that would explain the difference? should i re-reforge according to one or the other?

@Malcophant I, like you, have the normal version of VPLC and I'm getting the same results in SimCraft, with it favoring Crit over Mastery, while Haste is far behind. Does this mean that we should reforge into Crit to get better dps results in game?

@jeffythewise Seeing as you don't even have VPLC, it seems weird that you're getting these results, but like I said previously, I am also getting Crit > Mastery >> Haste results, even when I equip another trinket than VPLC, and re-run my sims. Can anyone explain this?

I'm currently 1486 haste rating, With TW and 5% haste buff I'm hitting the 1 second cap with no T11 bonus on AB4

From what I've tested with simc my suspicion is that a lot of the differences between rawr and simc come from the action priority list in simc. Those are usually optimized for best in slot gear, but perform a bit suboptimally on lower gear, specifically in conservation phase. As a result anything that gives value to conservation phase when executed optimally will have lower value. For example if you don't adjust your AB stacks when haste increases you'll start dropping low on mana and get lower value for it as a result. I was planning to create new more general priority lists now that simc has included conditions on non-proc max mana, I just didn't have time to get to it so far.

Don't take this for absolute truth, this is just based on my observations so far.

When I do not need it in advance for something else I prefer to use it before going into my initial burn at the start for the 30% increased crit chance on the next two ABs. You should always judge this on a per-fight basis however.