[SimC] WoD Arcane SimC Discussion

[magictricks]

Updated Arcane action list to time CDs against Nithramus, every 2 minutes. Doesn't do anything for 735 ring users, but slight gain as you the ring gains itemlevels, capping at less than 4% gain even with 795 ring.

my guess is pairing is worth 5-10k at max ring level vs. not pairing

More like ~3K or less. Komma sees a ~1.5K gain with 756 ring.

Are you doing it correctly though?, so that it drops PC 6 seconds after the ring has been activated so that you get the explosion damage at the end?.

[Frosted]

That is not a DPS gain.

[Pathe]

That is not a DPS gain.

Considering the duration of Nithramus (15 secs), AP (15 secs) and PC (12 secs), is it really a DPS loss to pop AP together with Nithramus, but delay PC with 4 seconds (dont see why 6 seconds would be needed?) so Nithramus can burst into PC, while all of your ABs that go into the PC should still be AP-boosted (since you start casting the last AB right before AP expires and it will be still affected by its extra dmg).

Does that make any sense; or should we not delay PC but drop it right with AP and Nithramus?

[Curnivore]

That's what the simulations show and you can't blame that. However, you could run your simulations and see for yourself. Whatever stacks of charge you have before using the crystal the differences in total DPS aren't that significant, unless I miss something (I might have to look closely into the new simc threads to make sure).

[Komma]

There isn't anything to gain from Nithramus impacting Prismatic Crystal (Except Feast of Souls on Gorefiend). On the other hand, you don't want to waste any of Arcane Power's duration. The ideal scenario would be PC being dropped just before Nithramus is popped (or even 2 second earlier, as long as your first AB benefits from Nithramus when it finishes casting). That only works for mid fight, since raids usually don't delay the first ring usage.

[Frosted]

That is not a DPS gain.

Considering the duration of Nithramus (15 secs), AP (15 secs) and PC (12 secs), is it really a DPS loss to pop AP together with Nithramus, but delay PC with 4 seconds (dont see why 6 seconds would be needed?) so Nithramus can burst into PC, while all of your ABs that go into the PC should still be AP-boosted (since you start casting the last AB right before AP expires and it will be still affected by its extra dmg).

Does that make any sense; or should we not delay PC but drop it right with AP and Nithramus?

Yes, it is. It ended up being a trade off of AB damage for Nithramus explosion damage. You lose AB damage by delaying PC (thereby amping weaker ABs during your burn) to get the nithramus explosion on it. The small gain in nithramus damage isn't worth the loss of stronger ABs into PC (and sequentially into the ring).

[Fleks]

Nithramushit isnt amplified by PC anyways
Go to a single dummy,pop ring and cds spam DPS drop PC 2secs before ring explodes-->Ring and PC will have exactly the same dmg done so it is not like the hit from the ring gets amplified by PC

tl;dr making nithramus hit your PC is irrelevant

[skiz]

PoM
Regarding PoM I usually disable cold snap, to analyse a singel use of it, that is easy to line up with other things. With cold snap the problem with POM is usually to keep it synced. And a second use is literally always used to handle your mana.

Before I did quit with wow the following things yielded the highest result with PoM:

A) As non bloodelf:
- Use POM at the end of PC to fit in another cast
B) As bloodelf:
- Use POM to fix mana gained after cooldown poping. Use POM as every other race if you do use an instant on PC early on and it allows you to safely use your racial with that instant without mana capping. (instant = Supernova, Nethertempest)

At this very moment the following line is anoying. It basically uses Presence of Mind before PC is used.

Code: Select all

actions.conserve+=/presence_of_mind,if=buff.arcane_charge.stack<2&mana.pct>93

When will it be used?
1. Right after Pull
2. Mostly every time you do AMs before 4 AC and you are having NOT a bad luck streak ( 4 AB -> ABar is dps negative, with bad luck you will probably dont cap on mana.)

Changing the APL to just use it as A) sometimes yields a dps increase over ~1000 iterations and sometimes not. (450 sec fighttimes, 20% variance.) The net gain / lose is usually somewhere between 2000 dps on a ~140k bis profile. Having such an impact on even 1000 iterations seems to be a little bit off. From my experience optimizing POM in simc is a anoying and nonrewarding task. But I am pretty sure that it just gets messed up from Prophecy of Fear, i.e. cast am on procc on low AC charges. (However, casting that one is a dps gain.)

It could be fixed by calculating the traveltime of all your AM charges, comparing it to the time on POF procc and including the amount of time you do need to get to 4 AC with AB before dumping it. Have to mention potential AM proccs. Afaik casting AB up to 3 AC should already be mana negative - so you will reduce the amount of time you do cap your mana if you go to higher AC charges.

Results:

Weaving apl = default apl + weaving.
Pom fix = just use it during PC.
AB = casting AB on lower mana % than 93 and 4 charges of AC.
weave2 = shit, no one cares about.
Weaving apl, Pom fix and AB are all inside each others error range after ~25000 sims. More iterations: weaving profile wins.

Recommended playstile: If PoF debuff is up AND you have to AM AND you have the posibility to get to 4 AC before casting AM. Then build up AC with casting AB. Keep in mind that you may get a procc of AM that screws you. Prefer to avoid that by dumping AMs before that can even happen. I.e. probably you'll do 2x AB at max before dumping your AMs at any charges into PoF debuff. Well... way to complicated anyway for almost 0 gains.

Using PoM to handle your mana afterwards isnt really rewarding but apparently this is the way to go. By the way, using PoM during PC as you're used to will probably yield a higher maximum. (Napkin math: PoM + AB with Arcane power and multipliers > PoM + Ab without any mulitpliers).

Btw. I do have those PoM profiles on a lot of sims now, just to be able to double check results from time to time.


Weaving:
This was discussed in this thread already. The following line is a decent gain. Result can be seen above. Keep in mind that PoF scales with targets, making the result even better.

Code: Select all

actions.crystal_sequence+=/arcane_missiles,if=pet.prismatic_crystal.remains>2*spell_haste+(target.distance%20),target_if=max:debuff.mark_of_doom.remains

Pooling of AM for PC & Ring burn
How to in simcraft: using a short length profile with 0 variance that definitly includes just 1 more usage of PC after opener. Using timestamps to handle pooling. It was a profile with legendary ring and it should trigger PC at roughly 120 seconds infight.

Code: Select all

actions.conserve+=/arcane_missiles,if=buff.arcane_charge.stack=4&(!talent.overpowered.enabled|cooldown.arcane_power.remains>10*spell_haste|legendary_ring.cooldown.remains>10*spell_haste)&buff.arcane_missiles.react>1&time>90 actions.conserve+=/arcane_missiles,if=buff.arcane_charge.stack=4&(!talent.overpowered.enabled|cooldown.arcane_power.remains>10*spell_haste|legendary_ring.cooldown.remains>10*spell_haste)&time<90

Both lines basically replacing the default line without timestuff & arcane missiles count.

Result:

Pool 1 = if 4 AC use AM at 2 or more charges if PC is going to be up soon. This is definitly a dps gain.
Pool 2 = if 4 AC use AM at 3 or more charges if PC is going to be up soon. This mostly results in capping AM and that will hurt you. Produces the highest possible maximum. (Weaving apl and Pool 2 apl are inside each others error range - so I cant tell you which one is better unless I do run more reps. Pretty sure it is pool 2.)

Pooling of AM on conserve for MoD procc:
You can easily pool 1 Charge of AM for it if you assume MoD will procc soon. (Soon ~20-30 seconds till next assumed procc.) I actually cant find the correct profile for it anymore but it will probably be the same as above: playing around with time. Allways pooling on this case is a dps lose because you do risk to cap.

Regarding pooling in general:
If you do not cap AM charges AND you do not let AM proccs run out: it is possible to "always" keep 1 charge of AM for any situation for higher average dps and less bad luck. Trying to keep more than 1 charge of AM increased the chance of capping way to much, but It may be rewarding if you do it before PC for ranking purposes (higher maximum).


Burn trigger on fightend:
At this very moment robomage burns his mana to deeply running oom on some tries. I had a better picture of that one where he was actually oom. But this one should give you a rough idea:

Code: Select all

start_burn_phase,if=[b]target.time_to_die*1[/b]*(1-0.1*(talent.nether_tempest.enabled|talent.supernova.enabled))*(10%action.arcane_blast.execute_time)*1.1<mana.pct-10+(target.time_to_die*1.8%spell_haste)

I just changed target.time_to_die*0.75 -> *1. Because I am lazy. Result is a 1k dps gain after error.

3 different profiles with different values of that multiplier. Highest one was *1

Opener changes:
Fighttime is 100 sec, 0 variance. I do not want another PC to be ready.


Apls: weaving = weaving as mentioned above
Pom fix = just use pom with PC
pom fix2 = just dont use pom in the 10 first seconds after pull. (I.e. use it on pc, but do not use it to build up your first 4 AC)
opener = all those things in the quote excluding that am cast before PC. Because to much work.

Opener:
Sometimes Mark of Doom procs on the first cast, sometimes it procs a few ABs in. This leads to a priority system:

If(MoD procs on first 2 ABs): AB last 2 times, cooldowns (AP, berserking etc.), Arcane Missiles, PC as missiles land, target PC, Supernova (x2 if mark is still on boss), burn phase

If(MoD doesn't proc on first 2 ABs): PC (we're hoping MoD procs on crystal), burn phase getting all AMs possible onto crystal before it expires

I actually like the idea of changing openers to gain a better variance or even better to elimilate bad luck scenarios but I do not think that this approach is working. However, simcraft may be the wrong tool for that. Some RPPM stuff isnt modeled correctly inside of it and a shorter fighttime just spikes up errors. (i.e. Felmouth frenzy)

APl changes for opener:
http://pastebin.com/23sfw2Pb" target="_blank
copy="" -> use everything above but overwrite it with the following changes. I.e. Opener contains weaving APL, overwrites pom fixes with default apl and has a little implementation itself.

TL;Dr:
Arcane APL is pretty decent. Would have a look at the final burnphase where robomage runs oom. Weaving apl seems to be good - havent looked at potentional missplays resulting from it. I would include it. Optimizing PoM is probably to much work compared to the gain. Opener changes are meh. Pooling shouldnt be implemented in that way above because of fucking timings - and the gain is not really worth it to do it anyway. (But pooling adds fun to your rotation!)

[skiz]

Opener changes #2:

Default APL: Get somehow up to 4 charges of AC. If PoF procc = Am into it. Dump AMs before PC. Whats the problem with it? It is likely to get a PoF procc and 0 AM proccs, effectively wasting mostly the default procc. I.e. pulling without trinket.

Idea: Ensure that PoF will allways procc on PC. In simcraft PC before precasting AB doesnt work. (Robomage still targets Boss for precasts, switches after that on to PC). PC after precast works. (PoF may be able to procc from PCs cleave dmg, not sure about that behavior ingame and inside simcraft.)


Results of PC after precast. ~140 sec fighttime, 0 variance. Roughly ~1000 dps increase over 450 sec and 20% variance fighttimes. Compared to the weaving apl.

Pull2 = AB precast -> then drop PC
Pull1 = AB precast -> AM if procc -> then drop PC
Pull profiles contain weaving APL and are matched against default apl. Forget about pool1 & pool2.

Havent looked into cooldown usage during the opener with early PC. I just let the cookie cutter routine take over. I.e. pull 2 = AB -> PC -> Pop cooldowns -> damage PC. Should be worth to test if robomage should delay cooldowns for 4 AC and / or blow PoM for AB2. I would guess that using PoM for AB2 + delaying Cooldowns for AC 4 is worth it. If the precast AB can procc PoF than it may be worth to precast on PC instead of the boss.

Recommend to work on the opener and implement that into simcraft afterwards. It may sound weak on average over long fights, but looking at i.e. Archimondes P1 damage checks: reflecting rng reduced rotations in simcraft may be nice. (Optimizing frost openers at the start of WoD I was getting ~+30 dps after ~400 seconds, so this stuff is kind of huge for openers)

Variance and praying for RNGJesus:

Obviously the "best" max dps is gained by placing PC after reaching 4 AC and getting PoF after your first cast on that PC. However the likelyhood of such thing is very, very little. The max dps is higher on that PC after precast profile after 10000 iterations compared to default or weaving. Using PC after building up ACs is a high risk - low reward option. (on average).


Relevant lines:

Code: Select all

# trigger burn during opener, with 0 conditions. actions.init_burn+=/start_burn_phase,if=time<5

Code: Select all

#using time > 5 = just skip those lines during opener. Do not cast AMs on Boss because they're likely to trigger PoF actions.init_crystal=call_action_list,name=conserve,if=(time>5&buff.arcane_charge.stack<4) actions.init_crystal+=/arcane_missiles,if=buff.arcane_missiles.react&t18_class_trinket&time>5

Pull profiles:
http://pastebin.com/4tba4iti" target="_blank

[Komma]

To be honest with you, I don't think it is worth publishing or even investigating PoF manipulation on pull. There is literally nobody who knows exactly what is up with RPPM behavior right now, Blizzard included. All we know is that their given formulas and "reset to 2 min TSLP" is absolutely wrong, since it suggests a ~31% proc rate on the first attack for BRF trinkets. As such, any investigation or theory using this data right now is about as accurate as the wild guesstimates of SimC developers who took liberties while adjusting RPPM proc constants.

This isn't even taking into account other issues, such as PoF trigger mechanisms within SimC to be different from that of in game. Within SimC, the code behaves as client data suggests - the debuff application triggers and explosion triggers are the same set of spells, described as "direct damage". That is to say, Unstable Magic, Prismatic Crystal, Felmouth Frenzy and many other effects are incapable of triggering Prophecy of Fear on a target. In other words, the often seen "Arcane Blast hits crystal but Mark of Doom still goes on a different mob" situation is impossible within the sim.

[Curnivore]

Blizzard included.

I bet they do. They're just not that kind of company to give raw code that easily or ever. They're more like a "get some partial info" kind of company to keep a veil of mystery on almost everything.

Another area of the game that seems to get a lot of voodoo on RNG is item drops. People often assume RNG is common RNG. But Blizzard has hinted on interviews that the RNG is often rigged.

e.g. we know some quests increase the item chance after bad luck or that raid class setups affect drops. It's likely some trinket effects might bypass pure randomness too.

[Komma]

Blizzard included.

I bet they do. They're just not that kind of company to give raw code that easily or ever. They're more like a "get some partial info" kind of company to keep a veil of mystery on almost everything.

As a matter of fact, this issue was reported to Blizzard by the SimulationCraft team and other community TCers over 8 months ago, and they've confirmed that our formulas are the same ones they use, despite in-game results being clearly different. They've promised to look into the code and investigate further, but we're entering month #9 without any new developments.

I would say it's safe to say that blue doesn't know either.

[Curnivore]

I'm still not sure there isn't some coder there that can see more clearly what's going on. I suspect there is a buffer of customer support people between players and the raw code. You may not be getting info directly from the most relevant coder.

[skiz]

To be honest with you, I don't think it is worth publishing or even investigating PoF manipulation on pull. There is literally nobody who knows exactly what is up with RPPM behavior right now, Blizzard included. All we know is that their given formulas and "reset to 2 min TSLP" is absolutely wrong, since it suggests a ~31% proc rate on the first attack for BRF trinkets. As such, any investigation or theory using this data right now is about as accurate as the wild guesstimates of SimC developers who took liberties while adjusting RPPM proc constants.

Interested in that issue. I've not looked into RPPM since the start of WoD. But I do know some ... odd behavior of it that doesnt fit the intended design. However, if PC hits to the boss are able to trigger PoF that whole stuff gets useless anyway, i.e. the chance of a correct PoF trigger should go down to 1/2 *(chance if it would not be able to trigger) . (Haven't seen that ingame so far, but I've not played PC that often.)

Edit: Btw. sadly the result of having a "bad" PoF implementation is that one : http://altered-time.com/forum/viewtopic.php?f=3&t=2119" target="_blank . But maybe this is also an ingame issue.

[Komma]

However, if PC hits to the boss are able to trigger PoF that whole stuff gets useless anyway, i.e. the chance of a correct PoF trigger should go down to 1/2 *(chance if it would not be able to trigger) . (Haven't seen that ingame so far, but I've not played PC that often.)

[Searix]

glad you're looking into it!

[Frosted]

APL Adjusted to take into account RPPM fix.

[Ogo]

APL Adjusted to take into account RPPM fix.

Please elaborate?

[Norrinir]

APL Adjusted to take into account RPPM fix.

Please elaborate?

See 1st commit and 2nd commit.

Anyways, I was looking into the arcane opener with the RPPM fixes. The default APL has rather big inefficiency with PoF. Arcane Blast precast can trigger PoF on the boss which is then absolutely wasted since the crystal_sequence action list doesn't take it into account. There are two ways to fix this, some numbers first:

Using the default mythic profile, 25k iterations with

Code: Select all

deterministic=1 seed=1337 max_time=18 vary_combat_length=0.0 fixed_time=1

results in 633081 dps.

We can remove the Arcane Blast precast to "guarantee" PoF proc on crystal. Simply commenting out the line

Code: Select all

actions.precombat+=/arcane_blast

gives 647946 dps.

The better way is to dump AM into whatever target currently has Mark of Doom on it. Very simple way is to do this is what one of the previous posts mentioned:

Code: Select all

#Replace: #actions.crystal_sequence+=/arcane_missiles,if=pet.prismatic_crystal.remains>2*spell_haste+(target.distance%20) #with: actions.crystal_sequence+=/arcane_missiles,if=pet.prismatic_crystal.remains>2*spell_haste+(target.distance%20),target_if=max:target.debuff.mark_of_doom.remains

This change gives 659803 dps on the opener (and also boosts the other crystal burst sequences).

Of course, an ideal action would fire AM into whatever gives more PoF explosions and prefer crystal in case the number of explosions is equal. That one would be... rather hard to express (especially since simc doesn't support min/max functions in expression and the equivalent min(a,b) := a * (a < b) + b * (a >= b) would get very large very quickly). And using an approximation such as the following

Code: Select all

# fire AM into crystal if you can guarantee 5 explosions no matter the duration of MoD on the boss actions.crystal_sequence+=/arcane_missiles,if=pet.prismatic_crystal.remains>2*spell_haste+(target.distance%20)&debuff.mark_of_doom.remains>2*spell_haste+(target.distance%20) actions.crystal_sequence+=/arcane_missiles,if=pet.prismatic_crystal.remains>2*spell_haste+(target.distance%20),target_if=max:target.debuff.mark_of_doom.remains

shows no change on opener burst and very slight dps increase on non-opener bursts. I'm getting amazing 8 dps increase with max_time=300 (remember that deterministic=1 and the only change is what we target, so it is an actual increase).

So yeah, the simple target_if version of "weaving" is probably good enough and the ~4.2% dps change in opener probably means it deserves a spot in the default APL.

[Komma]

The precast is a known issue. As mentioned in the RPPM post, we won't be going out of our way to model this. It is pointless, because the Mark of Doom trigger flags in game simply don't match that within the sim. For example, Prismatic Crystal pulses and Unstable Magic are incapable of triggering MoD in sims, which means that you are guaranteed a 100% first proc on PC if you do PC0 within the sim. That isn't the case in game, where PC/UM both have 30% chances to proc MoD whenever AB does not apply it directly.

--------------------------------------------------------------------------------

Fixed a really obscure bug with Nithramus contribution filtering, due to spell targets. It is somewhat significant, because it causes spells not "centered" at PC but impacting PC to contribute towards Nithramus, effectively double dipping on the explosion.

Yes, the bug has a rather huge effect on the opener modelling, so previous sim-based opener modelling is once again invalidated. Not that it had been valid previously, anyway.