Thanks a lot plopinou, I will take a look at the data, so maybe there is no need for additional testing :-)
The biggest difference is that you are using slam way more often than Heroic Strike, but in our dummy simulation we have them up equally. If I change the action line to only do HS during CS, I do less but harder HS, more Slam, and I gain ~1k dps. Maybe somebody can dig down a little bit deeper on this particular issue and we can thus close the gap.
Regarding the sclaing issue: One thing that could remove the gcd cap is when sudden death wouldn't proc colossus smash, because this way more haste means more colossus smash that we need to hit, i.e, even more buttons to push. Instead one could think of changing the debuff duration when a sudden death proc occurs.: E.g., every sudden death proc would give a buff that increases the duration of the next CS by 1 second. This would stack to a reasonable number. This would help out with the gcd cap, give better haste scaling, and could give interesting opportunities to make the stack big enough for longer durations to combine with other CDs.
I forgot, I'm not using Impeding victory on my runs, don't feel it will ever worth it when compared to Second Wind.
For the sudden death change, that's not a bad idea, but will it be enough ?
10% haste would cost 4246 haste rating (13 yellow rare gems).
With 0 haste Sudden Death proc once every ~15s. With 10% more haste it would proc once every ~13.5s on average. The increase is very low, and our class would need a complete revamp to make for the semi-loss of one of our hardest hitter which can procs enrage.
Also, the gameplay could become awkward, forcing us to stack a maximum rage every 20s to empty it with HS on the 6-8s the CS debuff will be up.
Haste never was a strong point of arms warriors, at least in the beginning of an expansion, even if it was better than what it is today (melee's only represents 60% of our rage generation instead of >90%).
I've given a lot of though about the arms warrior scaling problem, thinking about giving him more rage for example, by making opportunity strikes give 5-7 rage per hit, or making white melee crits give double the amount of rage. I even simulated those by hacking a little in the simcraft code, using my sparse knowledge of c++ (very sparse :p).
The result is of course a net increase of the HS count, but a net decrease in its average damage (TfB uptime is the same). Even if we buff HS base damage and decrease the gain per stack for TfB, the final result is an overpowered arms at low gear level that already do more HS's, and which need to be nerfed via Seasoned Soldier, for a very small upgrade of our scaling. It's not worth all the changes.
Doing more HS via more rage is not really viable solution for arms scaling, but still can be a nice upgrade for the mastery, at the cost of having HS do a bigger part of the total damage in later tiers.
What really is needed is a good increase in the crit value.
Arms warrior = crit, period. We want big slow crits
How to do it ?
They do seem to avoid any relation between deep wounds and crit, because of the mushing bug and all.
An easy solution would be the return of impale, with a +20% on all crits (240% instead of 200%).
But even that could not be enough to make the scaling look not so bad compared to fury warrior's.
Maybe a combination of the 2 ?
rage on our mastery proc + new impale ? That's something to test.
Or something entirely else that someone here could come up with.
I would love to test some more code changes in simcraft to try some ideas, but I don't even now how the spells infos are stocked (sc_spell_data.inc isn't readable for me), and can't manage to compile any release solution (only debug ones) ^^;
PS: Arms T14H went up by quite a bit, did the new action priority list do it ? In this state the scaling is a little less bad and a simple impale could do a lot of good (but not to fury's level)
Well during t14 I will go TG just because arms scales bad and SMF has too much dmg in execute.
Rage from opportunity strikes will increase mastery value but it will lower haste value.
Giving sudden death % chance to proc from opportunity strikes + white attacks would be some + to scale problems.
As for haste: Haste should inc deep wound ticks and lower cd for Mortal Strike, just as it does for Crusader Strike for Paladins. All solutions for Haste that makes arms gain more rage will fail just because more rage=more heroic strikes and heroic strike isnt rage dump no more. Haste have to do something else. Maybe lower GCD? Idk.
I think I will stick with SMF for the first tier, just because I think any fight with adds will generate more dps due to extra executes going off. Then again if I could get a set off 2Hs and 1Hs I would switch depending on the boss mechanics.
As for arms: A quick fix would be to make HS cost more and hit harder but at the same time make haste reduce the "ICD" of Overpower procs. Maybe a small GCD reduction from haste might be in order with that change, but it would help balance out the rage and GCD lock Arms has currently.
Regarding plopinou's numbers:
- Dragon Roar still seems to be hitting very low for Arms. I don't know if anyone checked that in SimCraft, but as I mentioned before, the spell description suggests that it should be multiplied by 1.2 for Arms. Maybe that is not working correctly. Shockwave might have the same issue by the way.
- In my logs, Heroic Strike and Bloodbath are very close to the sim'ed values you posted. Playing around with Bloodbath in SimC suggested that it should pretty much just be used on cooldown to maximize DPS, so that is what I do.
What is your exact Heroic Strike line Berthold? I tried to improve it a bit, this is what I use for Arms right now:
Build 16016 has a slight nerf to fury and a small buff for arms.
As far as I am concerned, that's good enough and all three specs will switch order based on the encounter design.
However, Blizz will need to re-balance it for the next tier as fury and especially Arms will then be gcd-capped and can only spend more rage via very inefficient Heroic Strikes, but this tier should work out very well.
I got another ~@200 dps out of it. I'm now using HS only if CS up as you suggested, unless it's a 4+ stack that is about to run out. Lowering the rage threshold to 75 increased DPS a bit as well, basically burning some rage with HS while CS is up instead of letting it go to waste.
Whereas the priority list has WS*>RB in all cases.
That's not true, WS* is only > RB/CS if there is 1 second or less left on Bloodthirst's cooldown. This happens when you are using e.g. RB followed by WS*. Instead of waiting and doing nothing for 0.5 seconds, another WS* should be used and BT should be delayed by 0.5 seconds. However, only WS* should be used to fill such a gap, never an ability that has a 1.5 second GCD. That's why there is a wait in the line following that one.
I added those two lines because it resulted in higher overall DPS. The "regular" WS* line is lower, under CS and RB.
hmm changing CD and HS to your suggestion gives me a loss of 60 dps (for the latest build). Let's take a look at it later on:-) Anyone caring to prepare some 14N or preRaid profiles? I would assume that the HS strike would be different for those as we do not have that much rage to spend.
Hm, yea, I must have mixed something up there. I did get a slight DPS boost by lowering the rage threshold for Deadly Calm a bit further, but this only increased DPS for the heroic T14 profile. The raid dummy profile would do less. Probably requires way more rules regarding rage management, so lets just stick to what you already have for now.
I did change a few other things that should yield a bit more DPS like moving up Dragon Roar in the priority list. That is also a DPS increase for SMF by the way, probably because it gets a damage bonus on DR as well.
I was reading again the the different stats of the arms warrior simulation, and I saw something that should be changed.
It's not about the priorty list, but is about a lag thing.
On the T14H profile, which is sim'ed with a 0.1s default lag, MS have an average interval of 6.16s, but it should be < 6.05 in the Beta (6.01 or 6.02).
The explanation is simple : CD's are now started client side before the server even received the information, and so we can make full use of the spell queue even as melee.
On live :
client does MS -> info going to the server -> server receive info -> server acknowledge -> info going to the client -> client receive info and start the 4.5s CD.
So the true CD of the MS on live can never be 4.5s, but more 4.7-4.8s (because even with 30ms ping I never was able to do an MS every 4.6s as simulated with the 0.1s default_world_lag).
On Beta :
client does MS -> gcd is instantly triggered -> info going to the server which returns the acknowledgement -> gcd can continue.
Even with 300ms ping on the beta, I can do 10 BT in exactly 45s (not even 45.3 or 45.5), not counting the one at 0s of course. Even with other skills used between them, I still get 45s, showing that the custom lag tolerance is doing a great job.
If it doesn't change anything for a lot of class, those with a main CD that beat the time like MS for arms and BT for fury will have a welcomed small dps increase.
Jewelcrafting which is currently used as a profession in Simcraft profiles isn't the best option as long as the crit SEP is between 0.5 and 1. The two jewelcrafting gems only offer 320 more stats regardless if it is a primary or a secondary one. On the other hand we are using pure crit sockets because non jewelcrafting gems offer 320 crit vs. 160 str which is in favor of the crit one. As jewelcrafter we have two options:
1) take crit jewelcrafting gems which results in 320 more crit as a profession bonus
2) take str ones which results in 640 more str for 640 less crit
Both cases are worse than the typical 320 str profession bonus.
As for SMF with T14H BiS gear it seems like Blacksmithing>Skinning>Engineering>Tailoring>Enchanting/Leatherworking/Inscription/Alchemy>Jewelcrafting>Herbalism>Mining
Simulationcraft takes skinning as a 320 crit bonus while on wowhead it says 480 crit. If wowhead is wrong then skinning is just as good as jewelcrafting.
Also in current fury profiles RB isn't used sub 20% which should be done as long as RB damage is higher than a third of one Exe.
Now to simplify, we use Heroic Strike if x&y&z=true, where
X is the big expression in brackets.
Y is if Target Health % >20
Z is if CS debuff is on the target
Now my question is, if CS debuff is not on the target, and if we are about to rage cap, this would actually make us waste Rage and not Heroic Strike, correct? Because even though the condition (buff.deadly_calm.up|rage>110) is satisfied, Z is not satisfied and hence we do not use HS, but we rage cap and may lose extra Rage. Please correct me if I'm wrong, but this is what it seems to me.
Is this simming higher DPS than having an X&Y|Z equation?
Also about this one for CS : colossus_smash,if=debuff.colossus_smash.remains<=1.5
What if there was no Rage available, and no spells available on cooldown? Do we simply not press CS merely because losing the extra seconds of the CS debuff is less DPS than delaying CS? Also taking into account possible Enrage procs for extra Rage and physical damage dealt.
Also a minor error for those who wish to test with Bladestorm instead of Avatar/Bloodbath :