I have the first set of results regarding optimal cycles. This was based on level 82 character with a mix of T10 gear and quest rewards with raid buffs, but I think it should generalize pretty well.

I had to use some restrictions to get a workable state space. In particual I restricted ABar to only situations when there is less than 0.5 sec left ABar cooldown (if cooldown > 0 it would wait it out if ABar is selected in such a case) and I made spell selection not depend on exact timing left on 2T10 buff.

If anyone wants to reproduce the results at some other data point you can check dev version of Rawr4.

Top cycle is obviously AB spam. Next is AB spam with AM whenever you see it up. Third cycle is using AM whenever it is up and using ABar at either 2 or 3 stack if it is off cooldown. Last cycle is using AM whenever it is up and also waiting for ABar even with 0.4 sec left on cooldown.

It's possible my interpretations are a bit off because it's a bit hard to figure out which states actually can happen in which cycle.

Another easier to understand scenario without 2T10. I also enabled waiting for ABar cooldown in all cases since it seemed optimal at extreme regen end.

Here we get another cycle at 3rd spot which only clears with ABar at 3 stack, this was probably missing above because of 2T10 haste making it redundant. The last two cycles are a bit more clear here. First is using AB2ABar, waiting for ABar if needed and after every AM follows with AB-ABar. The last one differs in that it doesn't use AM at 0 stack but at 1 stack instead.

Actually, that is exactly the behavior you'd expect if you're using AM "whenever it is up."

In order to pre-cast AM to clear an AB stack, the proc must be available before you finish your previous AB cast. In a cast sequence involving only AB, AM and ABarr, an AM proc will never be available after your first AB cast, so the earliest you can use AM is after the second AB.

AM doesn't proc itself, so your cast sequence would be: AM - AB - AB or AM - AB - ABarr, depending on the cycle you're using. If you're clearing stacks with ABarr, the proc would become available before you start restacking AB, so the sequence would be AB - ABarr - (AM) - AB - AB.

The only way to cast AM at 1 stack is deliberately to delay using the proc.

In the first cycle when you only clear stacks with AM, the proc won't be available at 0 stacks either because you've just cast it.

Kavan's results are interesting and are giving us some insight on how the arcane mage might scale. Logix's simulator has been assuming (correctly) a data point which gives us known variables from which we can test. The simulator however shows us that the cycles are very unstable and have a high degree of RNG associated with each cycle. Further none of the cycles resulted in a mana gain, or lead us to conclude there is a theoretical mana gain cycle.

Kavan's results show a potential positive mana gain at the ABX3 /AM 1st ABr (when needed) give us negative 57.12811 mana usage. If this is reliable, it shows that our hover point is much closer to our ABspam burn cycle than we originally theorized.

Further Kavan's results show that ABspam/AMproc is sustainable for relatively long periods of time. At 278.6863 mana per second, and assuming a mana pool of approximately 69,000 mana, (Chosen for the sake of ease) we are seeing 6,967.1575 mana use every 25 seconds, which is 10.097% of the mana pool. If the arcane mage can cast ABSpam/Amproc for 100 seconds this results in 69.709% mana remaining.

At the end of this 100 second cycle, we have 20 seconds remaining before evocation and 30% mana remaining that can be used for a high dps burn period. 30% mana in our hypothetical 69,000 mana mage, is approximately 20,708 mana. Using the burn cycle of ABspam, and its mana usage of 1,492 mps we see that a burn of approximately 14 seconds can be achieved. 1492 mana per second times 14 seconds = 20,888 mana used.

This leads me to two questions:

1. Kavan can be have the stats of the level 82 mage, mana, mastery rating, crit and haste that gives you your results? I am interested in seeing if a increase in the mana pool softens the RNG penalty to a degree that our mana pools can sustain a run of bad RNG and still sustain relatively high dps cycles for long periods of time.

2. I also want to see the standard deviations that accompany the hover/slower burn cycles? It might be that we have red light green light situations where an extremely bad run of luck using this high dps cycles, could lead us to dramatically deplete our mana pools, causing us to flex to more conservative rotations on the fly. What is particularly bothersome, is that while many of the cycles Kavan showed are resulting in net mana increases, we don't know the reliability of these phases, and whether a cycle down shift can allow the mage to maintain the mana pool in the face of bad RNG. This ties back to the notion that larger mana pools may be able to sustain bad runs of RNG better than lower mana pools.

Edit clarity:

Another point. It is not clear the effect of mastery on these burn cycles. The hypothetical mid point on the ABx3 cycle was 85% mana in my examples. The hypothetical mid-point for the ABspam cycle is 55% in my example. How does this change the up shift cycling theory? Will there by a point where the mastery stat will demand that only net mana increasing cycles be used?

Last Point. What we are really having trouble with is the creation of the new Arcane Mage Primmer. Fire and Frost each have a priority system which is easy to understand, and is something that most mage players will find familiar. The arcane primmer is not yet defined. Currently I have no idea what cycle we will be using or when I will be shifting cycles. I am not sure if I need to use a gap spell from another tree. We are almost at a point where we will have to graph the mana usage versus the dps/dpm cycles individually and then determine the intersection points for each cycle. While this may be fun to theory craft, it will not create dynamic game changing fun experience, especially with the amount of randomness we are seeing.

What I envision is a lot of confused players who will taste the bitter counter-intuitive concept of mana adept and will go back to the tried and true fire or frost trees.

Suggestion: If what we are seeing is a bland mix of cycles which punish bad RNG or luck, more than it rewards good sound play, research, and dedication, then what we need is another mana recapture mechanism. My suggestion has been to tie improved invisibility to mana regeneration ability. "For each second the mage remains invisible the mage gains 5% mana." What this does is remove the mage from dps for a fairly long period of time, (a 40% to 100% increase would require 12 seconds of invisibility) in exchange for the ability to have additional burst cycles, and the ability to recover from bad luck or game mechanics which create mana burns or depletion mechanics.

Not if you've just moved. In that case you will have cast AB1, the procc will have occured and you have had to move and had timr to see the procc occuring.

I know this is a rather stupid question... but my curiosity is too great to not ask...

While I know this would cause an immense loss of stats... would using two pieces of the old tier 3 (assuming you still had them and carried them around for 1 min evo cooldowns between trash >.>) be of any value for dps in the new arcane model? Basically what affect would having a 60 second cooldown on evocation make in the cataclysm arcane environment and would it have any value compared to the loss of stats from those pieces (say bracers and ring for example)

I'd be curious to see some results from RAWR with level 85s. That being said, how are you handling Clearcasting Kavan? Does RAWR model the actual proc, or just math it out?



That being said, I've just thrown up the new version of MMS. It has some new features that I think a few of you will really like.

A new version of Mana Made Simple is now online!
(http://manamadesimple.elementfx.com/)

This version introduces the first version of a brand new tool, the Cycle Builder!

The Cycle Builder is a powerful new tool to use with MMS! You can use it to craft, to a very precise level of detail, an entire casting strategy to use within a single simulation run. (which you can then iterate up to 200 times for statistical results).

Cycles are defined using triggers, which tell MMS when to start and stop a particular cycle. You can use a variety of events for your triggers, ranging from specifying a specific mana % point, to when a particular ability is used, to even time elapsed from the start of the fight (or time left till end of the fight). To get a better idea of the Cycle Builder, check its help section.

Along with this, there are a few other updates to MMS:

• Statistical printouts will now show the % mana left at the end of a sim.

• Support for building cycles to cast AM 'whenever available' (i.e even at no AB stack).

• Bug fixes! (thanks to all you guys that help me by finding bugs. Please keep it up, it is much appreciated!)

• A few core performance improvements (more to come soon)

Play around with it and let me know what you think. I will be posting some output from test strategies soon.

I appreciate all the work on the simulators and such you guys are doing! I thought I'd take a second to report some real world Beta results that aren't terribly scientific, but hopefully not a waste of your time.

Once I got Recount working, I spent over an hour with the test dummies in Ironforge trying to optimize my personal rotations. My goal was to be able to fight for 10-12 minutes, in order to get a second Time Warp in. So without much theorycrafting background, here is a synopsis of my results.

1) A simple rotation of ABx2, then either Missile Barrage or Arcane Barrage became the standard filler rotation. Flame Orb was cast on cooldown as well. This can be maintained indefinitely with the current mana gem and evocation cooldowns, but is not terribly high DPS.

2) Every cooldown cycle, I'd hit Mirror Image/Arcane Power/Mana Gem (Time Warp the first time) and then switch to ABx4/MB or AB for a high dps burn. After the mana burned, I'd evocate and switch to the lower DPS cycle to try not to oom.

With this strategy I regularly finished a 10-12 minute attack phase at between 7500 and 7800 DPS and well over 6 million total damage dealt.

Trying more aggressive strategies, ABx3, ABx4 it seemed like no matter how I modified the strat, I'd run out of mana rather quickly, a typical result would be 8500 DPS, but only a total of about 900,000 total damage done in about 75 seconds.

Depending on the length of the fight then, and the amount of external mana regen sources, we'll be able to dynamically adjust our rotations and a good mage will definitely put up much better numbers than a bad mage.

One aside: While questing/levelling/doing PuG dungeons the Evocation glyph that returns health and mana is invaluable for reducing downtime and helping out your healer. Being able to heal myself to full in an instance has been a lifesaver as healer mana with current gearing in Beta is an issue on every boss fight.

Another aside: Arcane Blast applying Slow is a godsend. I've levelled a priest, paladin and mage to 85 and the mage was by far the easiest.

Arril:

Real Life (Beta) information is an invaluable tool to keeping check on the simulations and the theorycrafting. Could you provide the stats which you are using? Specifically I am sorting through whether enlarging the mana pools will help the mage scale from ABx2 cycle to an ABx3 cycle. Currently I was seeing about 74,000 mana was necessary to stabilize the ABx2 cycle (using Logix' old simulator, I have not had time to test the newer version), and was finding that approximately 107,000 was necessary to stabilize the ABx3 cycle. Other than the simulators I don't have information regarding where level 85 mage will begin (freshly leveled) and where they will scale in a raid environment.

Further in real life (Beta) run some tests that would have you hover in ABx2 for about 90-100 seconds after an evocate, and then shift to a full burn cycle until evocate comes off of cooldown and see if your dps does not increase dramatically. Your method does not take advantage of the mastery as you stay in a low mana state for the majority of the fight. Try the other way around and report your findings from recount. This would give the community some more real life data to validate the current theory. (And maybe help your game play)

@Woeful: Absolutely. I posted from work, and at lunchtime went home to double-check so it's fresh in my mind - but I'm back at work again so can't give full details:

Arril stats (self-buffed): 84K health and 74,500 mana. Am using mostly ilevel 339 blue pvp gear reforged to have enough +hit on it so I could even hit that boss dummy consistently. A couple of ilevel 333 dungeon pieces (trinket, weapon, offhand, shoulders) and the epic ilevel 359 Alchemists stone trinket. He's pretty well geared compared to what's been available - but I haven't run any heroics yet - just regular dungeons. My current mastery score gives me an 18% bonus.

As far as the rotation, I did about what you were saying - I would start by burning thru 10K of mana in Arcane Blast, then hit a macro which would use my mana gem and cast the cooldowns - then go a main burn cycle. I'd evocate after which would get me to about 80-100% full mana depending on arcane missile procs. Then I'd lower the cycle while evocate and the other cooldowns ticked back up trying to keep over 40% mana. A large part of the rotation was based on RNG, how often did I get an Arcane Missiles proc because when it procs, I was almost in a positive mana gain situation, when it didn't, I would drop mana. If evocate was almost up, I'd maybe switch to ABx3 if I was still in a high mana situation. It was very dynamic, and it was an adjustment for me because while levelling you can continue to use your Lich King rotations. You don't run out of mana on quest mobs

Since that first test I was more interested in if I could even find a rotation that would last me 10 minutes, I didn't pay as much attention to some other details. The first 20 minutes was just me trying ABx4 and ABx3 and running out of mana before I did even 2 million damage. So I'll have a better baseline going in this time. I'll probably spend some time trying to optimize for a 5 minute fight also.

I'll get more details on my gear and stats and test again this week now that I've had some practice.

Here's a rotation I use on the PTR with a premade (ilvl 245 gear) that nets me 7.5k DPS:

Arcane Blast to 4 debuff stack, then cast until you've used enough mana for the gem to restore you to 100%. Cast until you're at 20~21% mana, then cast Evocation to full mana (assuming you have Mage Armor). Finally, AB to 3 then remove the debuff using Arcane Missiles if it procced or Arcane Barrage if it didn't. Repeat when the gem is off cooldown. Potentially use PoM to generate a new gem when you've used all 3 charges.

You can supposedly get similar DPS by just doing ABx4->AM/Barrage, but this is more fun IMO.

The idea is to be at near 100% mana when you aren't spamming AB to get the most out of your mastery, while exploiting the short Evocation CD and the mana gem buff.

YouTube - Alternative 4.0.1 Arcane Mage Rotation

A few thoughts on Polymorph/Counterspell. This is mostly from the pov of running heroics, so doesn't really apply to pvp, etc.

Polymorph honestly feels like it's become one of the weaker cc's now. Running heroics with a shaman and warlock honestly made me feel jealous. Hex/Fear (glyphed for 5mans) don't break on damage, and (not as big) don't heal the target. Obviously it's not the only CC that continues to break on damage, but it's certainly become the only one with a downside. It's weird doing Grim Batol and telling your party "I don't want to use poly on this group since it'll heal the mob". (For those who haven't run it, most of the pulls in there won't be at full hp since you do a bombing run type deal beforehand). I'm jealous when I see thunderclap or consecrate (or goddamn 30 second dk diseases) not breaking fear/hex.

And honestly, looking at PvP, with the supposedly 'longer' combat times, polymorphing someone seems like a huge disadvantage as a general thing. I realize someone's going to pop in and go "just means you have to be smart about who you cc"; but consider basically every other cc can just be used as you see fit. Polymorphing a target you've been focusing really seems like a step back. I recall the old joke "polymorph is the single most effecient heal in the game".

Moving on to Counterspell, it still feels really limited as a 24 second cd. Yes it has the longest lockout, but (especially for pve) quantity of interrupts is usually far more important then duration of the lockout. It'd be nice if the performance of this was evaluated. Running Throne of the Tides, two bosses had abilities that would either gib the tank or a random party member unless interrupted. I felt pretty useless with a shaman and Dk in the group on that particular run as you can tell.

I had honestly hoped we'd see Spellsteal turned into another interrupt that would function like Dark Simulacrum (before they actually gave DS to dk's, once they got that I figured the jig was up) just to give mages a little something (even if on a long-ish cooldown, two interrupts would be big).